Environmental Monitoring Series
CONTINUED RESEARCH IN MESOSCALE AIR
        POLLUTION SIMULATION MODELING:
Volume II  - Refinements in the Treatment
of Chemistry, Meteorology, and Numerical
                     Integration Procedures
                        Environmental Sciences Research Laboratory
                            Office of Research and Development
                           U.S. Environmental Protection Agency
                       Research Triangle Park, North Carolina 27711

-------
                RESEARCH REPORTING SERIES

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

     1.    Environmental Health Effects Research
     2.    Environmental Protection Technology
     3.    Ecological Research
     4.    Environmental Monitoring
     5.    Socioeconomic Environmental Studies

This report has been  assigned to the ENVIRONMENTAL MONITORING series.
This series describes  research conducted to develop new or improved methods
and  instrumentation for the  identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
 This document is available to the public through the National Technical Informa-
 tion Service. Springfield, Virginia 22161.

-------
                                   EPA 600/4-76-016
                                   May 1976
     CONTINUED RESEARCH IN MESOSCALE AIR
       POLLUTION SIMULATION MODELING:
  VOLUME II - REFINEMENTS IN THE TREATMENT
       OF CHEMISTRY, METEOROLOGY, AND
      NUMERICAL INTEGRATION PROCEDURES
               S. D. Reynolds
                   J. Ames
                 T. A. Hecht
                 J. P. Meyer
                D. C. Whitney
                 M. A. Yocke

     Systems Applications, Incorporated
             950 Northgate Drive
        San Rafael, California  94903
                 68-02-1237
               Project Officer

            Kenneth L. Demerjian
     Meteorology and Assessment Division
 Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina  27711
    U.S. ENVIRONMENTAL PROTECTION AGENCY
     OFFICE OF RESEARCH AND DEVELOPMENT
 ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, NORTH CAROLINA  27711

-------
                          DISCLAIMER
     This report has  been  reviewed by the Office of Research and
Development,  U.S.  Environmental Protection Agency, and approved
for publication.   Mention  of  trade names or commercial products
does not constitute endorsement or recommendation for use.

-------
                                                                     m
                               CONTENTS


DISCLAIMER	ii

LIST OF AUTHORS	vi

LIST OF ILLUSTRATIONS	vii

LIST OF TABLES .  .  .  *	    x

LIST OF EXHIBITS		xii

 I  INTRODUCTION  	 .    1

II  CHEMISTRY-RELATED DEVELOPMENT STUDIES	    4

    A.   Development of an Automatic Computer Program  for the
        Evaluation  of Kinetic Mechanisms  	    5

        1.  Treatment of Chamber Effects		    6
        2.  Computational Considerations  .  	  .  	    8
        3.  Ease  of Changing Reactions  .	    9

    B.   Development of an Improved Kinetic  Mechanism  for
        Incorporation in Photochemical  Dispersion  Models 	   11

        1.  General Considerations in .the Design of a
            Suitable  Mechanism 	   11
        2.  Elimination of Unimportant  Reactions in the
            General Kinetic Mechanism	14
        3.  Further Modifications To Reduce Computing
            Requirements .,	29
        4.  The Present Status of the Mechanism	32

    C.   Development of a Kinetic Mechanism  Describing S02
        Reactions and Sulfuric Acid Formation	35

        1.  The State of the Art of Gas Phase S02  Kinetics	36
        2.  The State of the Art Regarding  the Oxidation
            of S02  in Solution	'	41
        3.  Efforts To Test the Gas Phase Reaction Mechanism
            for S02	51
        4.  Future  Examinations of S09  Chemistry  	   52
                                     c.
    D.   Special  Considerations  Regarding the Treatment of
        Temperature,  Water,  and Hydrogen Peroxide in
        the Airshed Model	53

-------
                                                                     IV
 II   CHEMISTRY-RELATED  DEVELOPMENT  STUDIES  (Continued)

         1.   The Predicted  Effects  of  Changes  in Temperature
             and Water  Concentration on  Smog Kinetics	56
         2.   Specification  of  the  Initial Concentration  of  FLCL.  .  .   62
         3.   Spatial  and  Temporal  Variations in Temperature
             and Water  Concentration in  the South  Coast
             Air Basin	63

     E.   Treatment of Organics in  the  Airshed  Model	73

     F.   Introduction of  the  Improved  Kinetic  Mechanism
         into the Airshed Model	77

III   METEOROLOGY-RELATED  DEVELOPMENT ACTIVITIES	87

     A.   Model  Sensitivity  to  the  Inclusion of Wind  Shear	87

         1.   Wind Velocity  Profile	89
         2.   Implementation of the  Wind  Velocity Profile	90
         3.   Computer Coding	91
         4.   Description  of the Experiment	91
         5.   Discussion of  the Results	92

     B.   Treatment of Wind  Shear in the  Airshed Model.	101

     C.   Examination  of an  Algorithm for Deriving  Mass-
         Consistent Wind  Fields	101

         1.   The Governing  Equations	103
         2.   Tests of the Model	106
         3.   Discussion of  the Results	110

     D.   Adoption of  an Improved Algorithm  for Estimating
         Turbulent Diffusivities 	  .  	  121

     E.   Modified Treatment of the  Inversion Layer in
         the Airshed  Model	123

 IV   EVALUATION OF ALTERNATIVE TECHNIQUES FOR  INTEGRATING
     THE SPECIES CONTINUITY EQUATIONS	126

     A.   Introduction	126

     B.   Available Methods	130

         1.   The Price  Scheme	131
         2.   The Crowley  Second- and Fourth-Order  Methods	132
         3.   The SHASTA Method	133
         4.   The Galerkin Method	135
         5.   Particle-in-Cel 1  Techniques	140
         6.   The Method of  Egan and Mahoney	141

     C.   A Test Problem	142

-------
 IV  EVALUATION OF ALTERNATIVE  TECHNIQUES  FOR INTEGRATING
     THE SPECIES CONTINUITY  EQUATIONS  (Continued)

     D.   Results	146

         1.   The Price Scheme	148
         2.   The Crowley Second-  and  Fourth-Order  Methods	148
         3.   The SHASTA Method	149
         4.   The Galerkin Method	149
         5.   Particle-in-Cell Methods	149
         6.   The Method of Egan and Mahoney	150
         7.   Computational Time	189

     E.   Conclusions	190

  V  AIRSHED MODEL MODIFICATION FOR MULTIDAY  SIMULATION	192

     A.   Introduction	192

     B.   Model  Refinements	194

         1.   Treatment of Photochemistry at Night	194
         2.   Definition of the  Modeling Region	196
         3.   Use of a Grid with Variable Resolution	197
         4.   Modification of the  Finite Difference Equations  .  .  .   198
         5.   Modification of the  Computer Codes	200

     C.   Multiday Simulation of the Los Angeles  Basin	201

         I.   Preparation of ^missions  and Meteorological
             Results	201
         2.   Discussion of the  Multiday Simulation Results  ....   203

     D.   Recommendations for Future Work	216

APPENDIX:  A USER'S GUIDE TO MODKIN	218

REFERENCES	283

FORM 2220-1	288

-------
                           AUTHORS
CHAPTER I    -  Steven D.  Reynolds
CHAPTER II   -  Thomas A.  Hecht,  David  C.  Whitney, Jody Ames,
                Steven D.  Reynolds
CHAPTER III  -  Steven D.  Reynolds,  Mark  A.  Yocke, Jody Ames
CHAPTER IV   -  James P.  Meyer
CHAPTER V    -  Steven D.  Reynolds,  Jody Ames
APPENDIX     -  David C.  Whitney

-------
ILLUSTRATIONS
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

Concentration-Time Profiles for NO, N0?, Oo, and Propylene
at 15°C and 35°C 	 L 	
Predicted Concentration-Time Profiles for NO, N02, 03, and
Propylene at 0, 50, and 100 Percent Relative Humidity 	
Locations of Temperature and Relative Humidity
Monitoring Sites 	
Distribution of the Temperature Aloft Above Rialto on
26-27 July 1973 	 	 	
Temporal Variations in Water Concentration at Five Locations
in the Los Angeles Basin 	 	 	
Distribution of the Water Concentration Aloft Above Rialto
on 26-27 July 1973 	
Predicted and Measured Concentrations for La Habra Using the
15- and 31-Step Kinetic Mechanisms 	
Predicted and Measured Concentrations for Anaheim Using the
15- and 31-Step Kinetic Mechanisms 	 	 	 	
Predicted and Measured Concentrations for Pomona Using the
15- and 31-Step Kinetic Mechanisms 	 . 	
Predicted and Measured Concentrations for Pasadena Using the 15-
and 31-Step Kinetic Mechanisms 	
Predicted and Measured Concentrations for Downtown Los Angeles
Using the 15- and 31-Step Mechanisms 	 	 	
Predicted and Measured Concentrations for West Los Angeles
Using the 15- and 31-Step Mechanisms 	
The Effect— Expressed as Average Deviation— of Variations in
Vertical Wind Shear on NO and N02 . .' 	
The Effect— Expressed as Percentage Deviation— of Variations
in Vertical Wind Shear on NO and N02 	 • 	
The Effect— Expressed as Maximum Deviation— of Variations
in Vertical Wind Shear on NO and NO? 	

60

. 61

67

. 69

71

. 72

81

82

83

84

85

86

, 93

. 94

, 95

-------
                                                                           vm
16    The Effect—Expressed as Percentage Maximum Deviation—of
      Variations in Vertical Wind Shear on NO and NO?	96

17    The Effect—Expressed as Average Deviation--of Variations
      in Vertical Wind Shear on CO and 63	97

18    The Effect—Expressed as Percentage Deviation—of Variations in
      Vertical Wind Shear on CO and 03	98

19    The Effect-Expressed as Maximum Deviation—of Variations in
      Vertical Wind Shear on CO and 03	  .  99

20    The Effect-Expressed as Percentage Maximum Deviation—of
      Variations in Vertical Wind Shear on CO and 03	»,...- 100

21    Distribution of 03 Aloft Between Brackett and Rialto
      During the Morning of 11 July 1973	124

22    Concentration as a Function of Downw.ind Distance for the
      Explicit Price Scheme	151

23    Concentration as a Function of Downwind Distance for the
      Implicit Price Scheme .....  	 156

24    Concentration as a Function of Downwind Distance for the
      Crowley Second-Order Scheme 	 161

25    Concentration as a Function of Downwind Distance for the
      Crowley Fourth-Order Scheme 	 166

26    Concentration as a Function of Downwind Distance for the
      SHASTA Method 	  ......  	 171

27    Concentration as a Function of Downwind Distance for the
      Galerkin Method	176

28    Concentration as a Function of Downwind Distance for the
      Particle-in-Cell (Smoothed) Methods  	 ... 181

29    Concentration as a Function of Downwind Distance for the Egan
      and Mahoney Method	186

30    Comparison of Predicted and Measured Hourly Averaged
      CO Concentrations at Downwind Los Angeles 	 205

31    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at Long Beach	206

32    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at West Los  Angeles	207

-------
33    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at Burbank 	 208

34    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at Reseda  	  ........ 209

35    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at Whittier  ................... 210

36    Comparison of Predicted and Measured Hourly Averaged CO
      Concentrations at Azusa 	 ...... 	 211

-------
                                                                         X
                                 TABLES
 1    The Reactions Ranked by Amount of Uncertainty  ...  	   18
 2    Individual Area and Sensitivity Indices	23
 3    The Reactions Ranked by Sensitivity  	   24
 4    Characteristics of the Smog Chamber Runs	27
 5    Values of T, H, and M Before and After Removal  of the
      Six Reactions	29
 6    A Lumped Kinetic Mechanism for Photochemical  Smog  	   33
 7    Type of Mathematical Representation Required  to Predict
      Concentrations of Species in the General  Mechanism 	   35
 8    The Effect of Different Catalysts on S02 Oxidation 	   47
 9    Activation Energies of Reactions in the General Mechanism  ....   57
10    Ground-Level Air Temperatures in the Los Angeles Basin on 28-30
      June 1974	   65
11    Ground-Level Relative Humidities in the Los Angeles  Basin on
      28-30 June 1974	   66
12    Rate Constants for 0, OH, and 03 Attack on  Various Hydrocarbons  .   74
13    Hourly Averaged Wind Speed and Direction in the Los  Angeles
      Basin on 29 September 1969 at 6:00 a.m. PST	107
14    Hourly Averaged Wind Speed and Direction in the Los  Angeles  Basin
      on 29 September 1969 at 3:00 p.m. PST	108
15    Mixing Depths in the Los Angeles Basin on 29  September 1969
      at 6:00 a.m. and 3:00 p.m. PST	109
16    Predicted Changes in Wind Speed and Direction for Case 1  	  Ill
17    Predicted Changes in Wind Speed and Direction for Case 2  	  112
18    Predicted Changes in Wind Speed and Direction for Case 3  	  113
19    Predicted Changes in Wind Speed and Direction for Case 4  	  114

-------
                                                                               XI
 20    Predicted Changes in Wind Speed and Direction for Case 5 	    115

 21    Predicted Changes in Wind Speed and Direction for Case 6 	    116

 22    Predicted Changes in Wind Speed and Direction for Case 7 	    117

 23    Predicted Changes in Wind Speed and Direction for Case 8 	    118

 24    Conditions Represented in Tables 16 through 23 	    119

 25    Values of Diffusivity and Peclet Number for Three Case Studies  .  .    147

 26    Kinetic Constants for Each Case	    147

 27    Computing Time Required for Alternative Solution Methods 	    189

 28    Organization of Multiday Input 	    202

 29    Predicted and Measured Hourly Averaged CO Concentrations at the
       End of the 29 to 30 September Nighttime Period	    213

 30    Multiday Ground-Level  CO Concentration Map at 5 a.m.  PST on
       30 September 1969	    214

 31    Single-Day Ground-Level CO Concentration Map at 5 a.m. PST on  30
       September 1969	    215

 32    Predicted and Measured Hourly Averaged CO Concentrations for
       the Last Hour of the Multiday Simulation	    216

A-l    Input Card Format for MODKIN	  .  .  .    221

-------
                                                                         XI
                               EXHIBITS
A-l    Listing of Main Program MODKIN	   236
A-2    Listing of Subroutine  LMPCAL  	   250
A-3    Listing of Subroutine  DIFSUB  	   253
A-4    Listing of Subroutine  DIFFUN  	   262
A-5    Listing of Subroutine  MATINV  	   266
A-6    Listing of Subroutine  PEDERV  	   268
A-7    Listing of Subroutine  PLOT	   269
A-8    Sample MODKIN Input  	   273
A-9    Sample MODKIN Output—Selected  Pages	   275

-------
                         I   INTRODUCTION

     The SAI urban airshed model  was originally developed  for the Environmental
Protection Agency (EPA) under Contracts CPA  70-148 and  68-02-0339.   Two  series
of reports, entitled "Development of a Simulation  Model  for  Estimating Ground-
Level Concentrations of Photochemical  Pollutants"  and  "Further Development
and Evaluation of a Model  for Estimating Ground-Level  Concentrations of  Photo-
chemical Pollutants," describe our models development  and  evaluation studies.
In concept, the model formulation was  general,  based on  mass conservation rela-
tionships for a reactive species  in a  turbulent fluid.   To implement the model,
however, we assumed that we could do the following:

     >  Use the gradient transport hypothesis to represent pollutant trans-
        port by turbulence.
     >  Neglect turbulence influences  on the net rate  of chemical  reactions.
     >  Neglect subgrid-scale concentration  variations  and their  effect  on
        reaction rates.

     Volume III discusses  this threefold assumption.  In addition,  we made
several  assumptions with regard to the treatment of various  parameters in the
model.  For example, we assumed that a 15-step  kinetic mechanism  could be used
to represent the chemical  reaction processes.  The nature  of these  assumptions
reflects not only the time and funding constraints on  our  work then, but also
the current understanding  of the  physical and chemical  processes  that occur  in
the urban atmosphere.  In  this Volume, we discuss  efforts  carried out under  the
present  contract to refine further various aspects of  the  model and its  usage.

     Basically, our model  refinement activities have focused on four areas:

     >  Chemistry-related  model  development  activities.
     >  Meteorology-related model  development activities.
     >  An evaluation of alternative techniques for integrating the species
        continuity equations.
     >  Airshed model modification for multiday simulation.

-------
Each of these areas is the subject of a chapter in this  volume.

     Chapter II discusses our efforts to improve the treatment of chemical
parameters in the model.   Specifically, we began with a  39-step general-
ized kinetic mechanism arid, by eliminating unimportant reactions, by
invoking the steady-state assumption, and by combining reaction steps  we
derived a 31-step mechanism suitable for incorporation in the airshed  model.
In addition, we examined  S(L chemistry and developed an  interim 10-step
reaction mechanism for describing both homogeneous and heterogeneous reac-
tions.  Although this mechanism has yet to be validated  using smog chamber
data, it does provide a starting point for treating S02  chemistry in the
airshed model.  We also determined the sensitivity of the kinetic model
predictions to variations in temperature, water concentration, and hLO^
concentration.  These results provide guidance with regard to the appro-
priate treatment of the spatial and temporal variations  of these parameters
in the airshed model.  Finally, the chapter describes our experience to
date in using the new 31-step mechanism in an actual simulation of a smoggy
day in the Los Angeles basin.

     Chapter III describes our efforts to improve the treatment of meteo-
rological parameters in the model.  We examined the impact on the. model
predictions of wind shear—an effect previously neglected in the model.
Upon finding that wind shear has a significant influence, we extended  the
capabilities of the model to treat this parameter.  In addition, we devel-
oped a methodology to derive improved diffusivity relationships (discussed
more fully in Volume III) and examined an algorithm for rendering three-
dimensional wind fields mass consistent.  We gave special consideration to
the treatment of elevated temperature inversions, especially with respect
to possible importance of pollutant exchange between the stable inversion
layer and the turbulent mixed layer as the inversion is  eroded by surface
heating.

     Chapter IV presents  our evaluation of alternative techniques for  inte-
grating the species continuity equations.  Because the governing equations
of the photochemical dispersion model are nonlinear, numerical techniques

-------
must be employed to obtain approximate solutions.   Since we must attempt to
solve large systems of coupled, nonlinear partial  differential  equations,
we have to be careful  to choose an appropriate numerical procedure.   The
two most important concerns influencing this choice are the following:

     >  Accurate representation of the horizontal  advective transport
        of pollutants.
     >  Efficient solution of large systems of nonlinear equations.

For this contract effort, we restricted our attention to the first of these
areas.  We carried out a comparative study of various alternative techniques
that have appeared in  the literature and that, if implemented in the airshed
model, would represent a means for minimizing truncation error propagation
effects.  The methods  examined include finite difference, particle-in-cell ,
and finite element schemes.  We applied each method to the same test prob-
lem, and we compared the numerical results with analytical solutions.

     Chapter V summarizes our efforts to modify the airshed model for multi-
day simulations.  In previous photochemical modeling studies, multiday simu-
lations have been ignored.  Model  applications have usually been limited to
the simulation of daytime conditions.  For example, a model run might start
at some point in the morning preceding the rush hour and extend into the
afternoon to model the buildup of CL.  Accurate nighttime simulations are
hindered by the typically small size of wind speeds then and the lack of
available measurements  aloft.  However, multiple-day simulations may prove
to be extremely useful.   For example, in the evaluation of an emission con-
trol strategy that is  to be carried out in some future year, the model  user
must carefully choose  the initial  pollutant concentration distribution to be
employed in the simulation.  If a multiple-day run is made, the influence of
the initial concentrations on the predictions for the second and subsequent
days will  not be as pronounced as it is on the first day.  Furthermore, mul-
tiday simulations may  uncover errors in the treatment of emission, meteoro-
logical, or chemical  parameters that would otherwise remain unnoticed in a
relatively short term  simulation.   Chapter'V concludes with a presentation
of the results of a 34-hour simulation of the Los  Angeles basin for CO using
the SAI model.

-------
        II   CHEMISTRY-RELATED  DEVELOPMENT  STUDIES

                          Thomas A. Hecht
         ,                David  C.  Whitney
                             Jody  Ames
                        Steven  D.  Reynolds

     One of the distinguishing  characteristics of models capable of estima-
ting photochemical  pollutant  concentrations  is that chemical reaction pro-
cesses must be represented accurately.  Two  pollutants treated in such
models for which air quality  standards have  been established, namely NCL
and Oo, are not emitted  from  sources  in appreciable quantities.  These pol-
lutants are formed  in the atmosphere  as the  products of numerous reactions
involving NO,  hydrocarbons, and a  variety of free radical species.  Because
of the inherent complexity of the  overall reaction processes, care must be
exercised to incorporate in a photochemical  dispersion model a tractable
kinetic mechanism which  embodies as much chemical reaction as possible.

     In this chapter we  discuss efforts to improve the treatment of the
atmospheric chemical  reaction processes in the SAI airshed model.  Previ-
ously, these processes were represented by a 15-step mechanism developed by
Hecht and Seinfeld  (1971).  Since  this mechanism was developed, however,
additional  efforts  have  been  undertaken to design improved mechanisms.  One
of the most promising mechanisms to appear in the literature is that reported
by Hecht et al. (1973).   This mechanism consists of 39 reaction steps and
treats four classes of hydrocarbons (paraffins, olefins, aromatics, and alde-
hydes).  In general,  this new mechanism seems to represent an advance of such
importance as  to warrant its  incorporation in the airshed model.

     In the course  of reviewing modeling needs with respect to atmospheric
chemistry,  a number of issues were raised.   These topics, which we address in
this chapter,  include the following:

    >   Development of a computer  program to facilitate the evaluation.
        of kinetic  mechanisms.
    >   Compaction  of the 39-step  hydrocarbon/NO /0~ mechanism to reduce
                                               A  O
        its impact  on computing time  in the  airshed model.

-------
     >  Development of an interim mechanism for S0? reactions and
        sulfuric acid formation suitable for incorporation in the
        airshed model .
     >  Examination of spatial and temporal variations in temper-
        ature, water, and H?CL concentration in an urban airshed
        such as the South Coast air basin to provide guidance with
        respect to the treatment of these parameters in regional
        models.
     >  Evaluation of alternative means for treating organics in
        the airshed model .
     >  Experience gained in the use of the new kinetic mechanism
        to perform an actual airshed simulation of the Los Angeles
        Basin.

Each of these issues is discussed in the 'succeeding sections of this chapter.

A.   DEVELOPMENT OF AN AUTOMATIC COMPUTER PROGRAM
     FOR THE EVALUATION OF KINETIC MECHANISMS

     Once a set of reactions for the formation of photochemical  smog has been
proposed, it is necessary to demonstrate that the mechanism is correct; i.e.,
that it is able to account for, within experimental error, the actual concen-
tration of each species present in the reaction mixture at any point during
the time span of the reaction.  In its simplist form, this evaluation process
involves the formulation and solution of the set of coupled differential equa-
tions that describe the variation in the formation and consumption of each
species in the reaction mixture as a function of time.  This set of equations
can be expressed as follows:
R     -  ^ R        ,                   (1)
                      -JI     -s  f
                            j=l   i,j   k=l   i,k
where

-------
          y.    = the concentration of species i,

          t     = time,
          R^    = the rate of formation of species i  in reaction j of the
            i,j   set of J reactions in which species i is  formed,
          R     = the rate of consumption of species  i  in reaction k of
            i,k   the set of K reactions in which species i  is consumed.

The concentrations thus calculated can be compared with those measured experi-
mentally in the reaction mixture.

     Unfortunately, the real world presents experimental, computational,  and
operational obstacles to the pursuit of this simple validation scheme.  First;
for the integrity of the reaction  mixture to be preserved,  the mixture must be
contained in some sort of reaction chamber, which in  turn gives rise to two
side effects:   leaks (intentional, as in sampling, or unintentional) and  wall
reactions.  Second, when the most  efficient computer  codes  are used, the  time
needed to solve the coupled differential equations increases  as the square of
the number of species increases.   Moreover, certain sets of rates lead all too
often to systems of "stiff" equations, for which the  solution times can become
quite large.  Finally, the urge always exists to "improve"  a  reaction mechanism,
no matter how closely it approximates the experimental  data;  the computer code
must allow these changes to be performed with a minimum amount of effort.  In
dealing with these realities, the  researcher is called  upon to display his met-
tle and to tax his ingenuity.  The approaches we used in this study are described
in the following subsection.

1.   Treatment of Chamber Effects
     With few exceptions,  reaction chambers are not completely airtight.   Under
normal  operating conditions,  this does not create a serious  problem, since al-
most all  chambers are maintained at atmospheric pressure,  and since the small
amount of interchange by diffusion can usually be ignored.   However, a problem

-------
does arise when samples are removed from the chamber for analysis.   Since
the species that comprise smog exist in the atmosphere in minute (1  to 1000
ppb) concentrations, sample sizes on the order of a few liters  are  commonly
needed to obtain enough material  for an accurate analysis.   Moreover,  sam-
ples must be withdrawn fairly frequently during the reaction to monitor
species concentrations that are changing rapidly.  As a result, it  is  not
unusual for 10 to 20 percent of the chamber volume to be withdrawn  through
sampling procedures.  The reaction simulation technique must take this
"dilution" of the reaction medium into account.

     In the ideal case, the gas used to replace the samples  being withdrawn
is inert with respect to the reaction (e.g., pure nitrogen  in a smog system),
or it contains only reactive species whose concentrations are so large--
relative to the amounts consumed or produced by the reactions — that they  can
be assumed to be constant throughout the reaction (e.g., oxygen or  water
vapor in "clean" air).  In this case, it is sufficient to apply a "dilution
factor" to the concentrations of all the species (inert diluent) or to those
that do not remain constant (clean air diluent).  If samples of an  approxi-
mately constant size are removed at reasonably uniform time  intervals, the
dilution factor can be considered to be a constant, Q, and  the  equation for
the rate of change of the concentration of species i becomes

                        dy.
                                              Rc    -yjQ    .            (2)
                              j = l    i,j    k=l   ci,k    1

     In some chamber experiments,  however, the incoming medium is  just the
natural atmosphere in the laboratory,  which may contain concentrations of
pollutant species as high as  or higher than those being followed in the
reaction chamber.  Moreover,  it may be desirable in some cases  to  inject
pollutants or pollutant precursors deliberately into the chamber to simulate
the effect of fresh emissions on the reacting  mixture.   As  long as the con-
centration of species i ,  y .  , in the incoming  medium is. known,  the effect of
such inflowing species on the rate equation can be easily expressed:

-------
                    dy-
                                           Rc    - (y,. - yin) Q    .      (3)
                           j=l   i,j   k=l  ci,k     -1    in

     Unfortunately, wall effects cannot be handled as neatly.  Wall  absorption
is best determined by placing the species in question, A, in a "nonreactive"
environment within the chamber and following its decay with time.   One can
then include within the reaction mechanism an equation such as

                            ka
                           A -> Wall         k                            (4)
with an appropriate rate constant.

     Heterogeneous catalysis by the reactor walls is even more troublesome.
However, by determining reaction rates at several different surface-to-volume
ratios for the reactor (e.g., through the use of "artificial"  walls to parti-
tion the chamber), the rate constants for the homogeneous and heterogeneous
reactions can be obtained, and both reactions can be included in the mechanism.

                               kc
          Homogeneous:    A + B •+ C    ,                                 (5)
          Heterogeneous:    A + B -> C                                    (6)

2.   Computational Considerations

     As mentioned earlier, the computer time required to solve a set of differ-
ential equations increases at least as does the square of the number of equations
to be solved (or, in the present case, as does the square of a number of distinct
chemical species that appear in the reaction mechanism).  Thus, any techniques
that can be used to decrease the number of species concentrations that actually
require coupled differential equations for their solution should be applied.
Such techniques include the assumption of constant concentration; the uncoupling
of product-only species; the invocation of the steady-state approximation; and
the aggregation, or "lumping," of. species that. yield similar products.  The last
two techniques are the subjects of further discussion in Sections B and E and
are thus not treated here.

-------
     Certain species  that appear in the reaction mechanism either are present
at truly constant concentrations (e.g.,. the reactor walls)  or have concentra-
tions so high — relative to the amounts  of that species  formed or consumed
during the reaction—that they remain essentially constant with  respect to
time (e.g., oxygen).   Since the change  in concentration for these species is
only negligibly different from zero,  they can be excluded from the differen-
tial equation process.

     A second category of species that  need not be included in the set of
coupled differential  equations is those that appear in  the reaction mechanism
only as products (e.g., C0? or HNO.J-   The rate of formation of these product
species is, to be sure, represented by  the following equation:
                           A
                           dy-
However, the presence of this  species  has  no effect on  the  rate  of formation
or depletion of any other species;  thus,  the differential equation describing
its formation can be uncoupled from the set of all  differential  equations  and
solved independently, at a significant savings in  overall computer time.

3.   Ease of Changing Reactions

     Most computer codes used  in the simulation of reaction kinetics  incor-
porate, in one form or another, the features described  above.  The major
advantage offered by the present program is the ease of preparation and,
particularly, the ease of alteration of the mechanism and its  associated
species concentrations.   The user need know nothing about computer program-
ming or the solution of differential  equations, and very little  about chem-
ical reaction mechanisms, to obtain meaningful results  from the  program.

     On the first line of input, the user specifies the run identification,
the number of reactions in the mechanism to be studied, how many of these
are lumped reactions, the number of each of the various types of species
described in the previous subsection, and an indication of whether the reac-
tion rates should be printed.   The second line continues this specification
of parameters with an indication of the frequency of printout, the time step
sizes, and the dilution factor.

-------
                                                                              10
     The user then submits his reaction mechanism, one reaction per line,
restricted only by one requirement on ordering:  The lumped reactions must
appear last.  Each reaction appears as an ordinary chemical equation, with
a list of reactants, a list of products, and a rate constant.   The products
can have coefficients (either fractions or integers), but the  reactants
cannot—each reactant molecule must be entered separately.   The user can
choose any four-letter mnemonic he wishes for the species names.

     If there are any lumped reactions, the sets of individual  reactions com-
prising each "lump" are then entered.  Their formulation is exactly the  same
as that of the lumped reaction, except that the name of each species that
contributes to the composition of the lump appears in place of the lumped
species as the first reactant.

     The user than provides the list of species and their initial  concentra-
tions—one per line.  The order of their types must be the same as that  given
on the initial parameter line, but no particular order is required within
each species type.

     Should any of the species be present in the gas flowing into  the reac-
tion chamber, their concentrations in the inflowing stream and, if needed,
the time and new values of any change in this concentration are entered  next.
Finally, the user can request concentration-time plots of any  species.   If
desired, these plots can contain experimental points with which those points
calculated by the proposed mechanism can be compared.

     To change a rate constant or chemical reaction, the,user  need merely
alter the corresponding input line.  New reactions can be added by insertion;
old ones can be removed by deletion.  A similarly easy process can be used to.
change an initial species concentration or to add or remove species names from
the list.  Species can be transferred among, species types (e.g.,  differential
to steady-state)  by a single interchange of lines.

     A complete user's guide to the computer program is included in Appendix
A.  This appendix provides detailed information on each of the features  de-
scribed above, descriptions and listings of all of the computer routines, and
sample inputs and outputs.

-------
                                                                              11
B.   DEVELOPMENT OF AN IMPROVED KINETIC MECHANISM FOR
     INCORPORATION IN PHOTOCHEMICAL DISPERSION MODELS
1.    General  Considerations  in the Design
     of a Suitable Mechanism
     The selection of a chemical  mechanism for inclusion in an atmospheric
diffusion model  depends substantially on two factors:

     >  The accuracy of prediction of the chemistry module.
     >  The computing time required to evaluate the mathematical
        equations representing the mechanism.

From the standpoint of developing a chemical  transformation model,  the second
factor is subordinate to the first.  After a reliable  mechanism has been devel-
oped, it can be condensed in several  ways to reduce the computing time necessary
to obtain predictions — for example, by eliminating unimportant reactions, by
combining species that react in the same ways  and at similar rates  into a gen-
eral grouping, and by invoking the steady-state approximation where applicable.

     Depending on the degree to which these compaction measures are applied,
the resultant mechanism can be assigned one of three broad categories:  detailed
mechanisms, lumped general mechanism^, or parametric models.  A detailed mechan-
ism consists entirely of elementary reactions.  Because there are hundreds of
different organics in the atmosphere, such a mechanism requires an  extremely
large number of mathematical equations to represent the chemical  transforma-
tions.  Although a detailed mechanism is ultimately the most accurate (in the
limit if all rate constants and reactions are  known),  it is unsuitable for atmos-
pheric modeling because of the second criterion listed above.  A lumped general
mechanism results when reactants  and reactions of the  same type are combined
into general classes and reactions and those that clearly do not contribute to
the .predictions are eliminated from a detailed mechanism.  A lumped general
mechanism strikes a balance between detail of representation and compactness
of form.  The elimination of reactions that significantly affect predicted con-
centrations would oversimplify the mechanism to the point where it could not

-------
                                                                              12
provide accurate predictions unless corrective measures  were taken.   In par-
ticular, adjustable coefficients would have to be incorporated,  forming a
parametric model.  For such a model, the values of the adjustable parameters
are selected to minimize the discrepancies between experimental  data and cal-
culated values.

     When we first began to develop a photochemical  airshed model two types
of chemical mechanisms were available for use.  The first,  a detailed model
for propylene, was unsuitable because it was too narrow  in  scope.  Its pre-
dictions for the atmosphere would have almost certainly  been unreliable.
The second was a parametric model, the Hecht-Seinfeld 15-reaction mechanism,
for which values of the adjustable parameters had been determined for several
hydrocarbon-NO  systems using smog chamber data.  These  hydrocarbons included
              A
propylene, iso-butylene, toluene, and n-butane; binary mixtures  of propylene
and n-butane, propylene and ethane, and toluene and n-butane; and auto ex-
haust.  Thus, to the extent that the atmosphere could be represented as a
surrogate consisting of these species, the parametric mechanism  could be ex-
pected to provide reasonably accurate predictions over the  range of initial
conditions explicitly used in selecting values of the parameters.  Moreover,
the mechanism was mathematically compact.  The predicted time-varying behav-
ior of the pollutants could be obtained at every time step  through the solu-
tion of only four differential equations and six algebraic  equations.  Given
a choice of these two mechanisms, we selected the parametric model for incor-
poration in the airshed model because it came closest to meeting our two
criteria.

     The compact mechanism is far from ideal, however.  Recent experimental
studies have demonstrated the key roles of OH and H02 reactions  in smog for-
mation, reactions whose importance is understated in the mechanism.   Other
studies have shown that 0 and CO are less important that we thought at the
time we formulated the model.  And one limitation that is particularly dis-
comforting is the narrow range over which values of the  adjustable parameters
are valid.  This last shortcoming would limit the accuracy  of the results
obtained from the atmospheric dispersion model in such applications as the
evaluation of alternative emission control strategies.

-------
     The mechanism most suitable for use in atmospheric models is a lumped
general mechanism.  Under EPA Contract 68-02-0580, we recently undertook
the development of such a general  kinetic mechanism.   In this  mechanism,
we incorporated state-of-the-art knowledge of the reaction processes,  and
we provided for the rapid and straight forward modeling of organic species
not explicitly evaluated using smog chamber data.

     In the new kinetic mechanism, the inorganic reactions common to all
organic-NO  systems are treated in great detail.  We  introduced generality
          A
into the model by lumping similar types of organics and free radicals  into
several new classes.   In particular, olefins, aromatics, paraffins, and
aldehydes constitute four separate classes of organics.  We segregated
organic free radicals into alkoxy, peroxyalkyl, and peroxyacyl subgroupings.
Using propylene-NO ,  n-butane-NO , and propylene-n-butane-NO  smog chamber
                  XX                           X
data over a wide range of HC/NO  ratios, we evaluated the model  and showed
                               X
that its predictions  of the dependence of peak ozone  on the initial concen-
trations of hydrocarbon and oxides of nitrogen qualitatively agree with
experimental observations.  Seinfeld et al. (1973) discussed the rationale
and formulation of this lumped kinetic mechanism, and Hecht et al. (1973)
and Hecht et al. (1974) presented initial and secondary evaluation results
using the mechanism.

     This new mechanism appears to be more accurate than the Hecht-Seinfeld
mechanism that we previously employed in the atmospheric simulation model.
In addition, we can easily extend the new mechanism to new organics that
have not been explicitly evaluated (the values of the adjustable parameters
do not need to be determined).  Unfortunately, the computing time that is
initially required to carry out a simulation with the new mechanism is much
higher than that needed for the Hecht-Seinfeld model.  At the  outset of this
project, representation of the chemistry of a system  consisting of a paraffin,
an olefin, and NO  in air (no aromatics or CO) required 36 reactions and the
                 X
solution of 16 differential equations and 4 algebraic equations to obtain
predictions.  Such mathematical complexity would certainly be  excessive if
the mechanism were imbedded in the airshed model, where the kinetics must be
evaluated at every grid point for every time step. We therefore set out to
reduce the computing  time necessary to obtain predictions from the mechanism.

-------
                                                                              14
     We approached the problem of long computational  time requirements  from
two directions.   Initially,  we sought to condense the mechanism to the  small-
est number of reactions required for accurate predictions.   We  identified
critical  reactions by means  of a sensitivity analysis, and  we subsequently
eliminated insensitive reactions from the model.   We  also found that taking
a flexible posture toward solving the representative  rate expressions resulted
in time savings.   Our experience in working with  the  kinetic mechanism  showed
that computation  time increased approximately linearly with the number  of
reactions, but quadratically with the number of coupled differential  equations.
Thus, we identified and verified species for which the pseudo-steady-state
approximation is  valid; this step permitted the replacement of  three coupled
differential  equations by three coupled algebraic equations.  Next,  we  took
advantage of the  fact that differential  equations describing the concentra-
tions of species  that are formed as a result of chemical  reactions but  do not
themselves enter  into reactions can be solved independently of  reacting species.
We separated these so-called uncoupled species from the coupled species,  elim-
inating three coupled differential  equations, but adding  three  uncoupled  dif-
                    *
ferential equations.   Finally, we  eliminated one species from  the mechanism
by algebraic manipulation, thus reducing the number of coupled  algebraic  equa-
tions by one.

     Since the computing time was the single greatest hindrance to our  incor-
porating the improved kinetic mechanism in the airshed model, we focused  a
great deal of attention on this problem.  As a result, we condensed the mech-
anism in both its physical and mathematical structure to  a  form that is amenable
to diffusion modeling.  We discuss  the details, methodology, and results  of
this program below.

2.   Elimination  of Unimportant Reactions
     in the General Kinetic  Mechanism

     During the period in which we  first formulated and subsequently modified
the lumped kinetic mechanism to achieve satisfactory  predictions, we added and
deleted several  reactions.  However, we made no attempt to  eliminate unimportant
 Appendix A discusses the concept of coupled and uncoupled species,

-------
                                                                              15
reactions.  Under EPA Contract 68-02-0580, we recently completed a sensitivity
analysis of the kinetic mechanism; we used the results of this study to help
us select possible reactions for elimination.  Our goal  in the sensitivity
study was primarily to identify the "critical parameters" in the model, that
is, those whose uncertainties most greatly influence the reliability of pre-
dictions.  In essence, we calculated the rate of change in predictions with
changes in the value of each rate constant, holding all  other rate constants
fixed at their standard values as a measure of sensitivity.   Rate constants
for which the measure has a high value correspond to sensitive reactions.
Low values indicate insensitive reactions that may not have  to be included
in the model  to make accurate predictions.  Before proceeding with a discus-
sion of our results, we describe the procedures and methods  that we used in
the sensitivity analysis.

     The sensitivity study focused on a binary hydrocarbon-NO  system (EPA
                                                             X
'Run 352) in which the initial concentrations were as follows:

                                        Concentration
                      Species               (ppm)
                        N02                 0.07
                        NO                  0.27
                    Propylene               0.265
                     n-Butane               3.29

We chose this particular experiment for several reasons:

     >  Both  high and low reactivity hydrocarbons were present initially.
     >  The initial concentrations of total hydrocarbons and oxides of
        nitrogen were typical of those found in a polluted atmosphere.
     >  The accumulation of ozone reached an asymptotic level during the
        experiment.
     >  We had the run-modeled with reasonable success in our evaluation
        study.

-------
                                                                              16
     We performed the sensitivity analysis in the following manner.   Using
the nominal  (or base) values for all  parameters reported by Hecht et al .
(1973) (see  Tables 14 and 16 in that reference), we obtained base concen-
tration-time profiles for propylene,  n-butane, NO, N0?, and CL by integra-
ting the governing rate equations with each parameter at its nominal value.
We then increased (and subsequently decreased) one of the parameters by  a
fixed percentage, holding all  other parameters at their nominal  values.   We
integrated the equations twice, once for each of the two new settings (+50
percent and  -50 percent) of the selected input parameter.  Repeating this
process for  each rate constant, we carried out, for n parameters, integra-
tions for a  base case and 2n parameter variations.  Finally, for each of
the 2n + 1  integrations, we determined the values of the sensitivity measures
or "decision variables."  We compared the magnitudes of the decision varia-
bles for each variation in a parameter with those computed for the base  case,
and ranked the sensitivities of the parameters by tabulating the magnitudes
of the differences.

a.   Measurement of Sensitivity

     Central to a sensitivity  analysis of a mathematical model is the mean-
ingful quantification of changes in model predictions that result from per-
turbing the  input parameters one at a time.  As the measure of sensitivity
for each parameter,  we chose the absolute area between the concentration-
time profile for the given parameter, with all parameters held at their  base
values, and  the profile generated when the i-th parameter was perturbed  by a
fixed percentage.  We denote these parameters as A^, A|\JQ, Ao3> A0lef (Pro~
pylene), and A^^ (n-butane).   Since Hecht et al . (1973) discussed these
              pard
criteria in  some detail, we review their appropriateness only briefly here.
     The five indices (A[^OO»  ANQ,  AQ~,  ^olef-'  anc' Apara)  constitute continuous
measurements of sensitivity determined  experimentally over the entire period
of simulation for each species.   Mathematically, we represent this relation-
ship as follows:
                        400 miri
/
                                lC.(p.t)  - C.(p + %p,t)|  dt

-------
                                                                              17
where
               =  the absolute area between the concentration-time
                  profile predicted for the i-th species, with all
                  parameters at their base values, and the profile
                  obtained when one parameter is perturbed by a
                  fixed percentage.
           C   =  the concentration of the i-th chemical  species.
           i   =  the species index--NCL, NO, CL, olefin, paraffin.
           p  '=  the parameter that is being perturbed.
           %   =  the percentage perturbation in p divided by 100.

If the perturbation of a given parameter greatly alters the time history of
the i-th chemical  species, indicating high sensitivity to that parameter,  A.|
will have a large value.  But if the concentration-time trace remains essen-
tially unchanged, the predictions of the model  will  be insensitive to varia-
tions in the parameter under evaluation, and |A.|  will be small.

     To facilitate the comparison and ranking of their sensitivities, we varied
all parameters in turn by the same fixed percentage.  Some of the  input param-
eters are very poorly characterized, having associated uncertainties of up to
an order of magnitude, whereas the values of other parameters are  known within
an uncertainty of 10 percent.  The comparative table of rate constants [Table
16 in Hecht et al. (1973)J suggests that a representative "degree  of precision"
among the several  alternative experimental determinations or theoretical esti-
mates available for any particular parameter is on the order of 50 percent.
Therefore, we used that percentage as the magnitude of perturbation for the
sensitivity calculations.  However, because the precision bounds of the rate
constant values for individual reaction rate constants vary greatly, our choice
of the "range of perturbation" was arbitrary.  The significance of the 50 per-
cent figure rests only on its approximate division of the very uncertain from
the less uncertain parameters.  Table 1 ranks the reactions by the amount of
uncertainty.  (We define the uncertainty bound for a rate constant as the range
within which the "true value" of the constant can be presumed to fall with con-
fidence.)

-------
            Table-!

THE REACTIONS RANKED BY AMOUNT
        OF UNCERTAINTY


1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.

.'Reaction
N205 + H20
NO + HN03
HN02 + HN03
NO + N02 + H20
HN02 + HN02
HN02 + hv
OH + N02
OH + NO + M
R02 + NO
RC03 + NO
RCOo + N09
o c.
H02 + H02
H02 + R02
R02 + R02
RO + N02
ALD + hv
RO + NO
OLEF + OH
0 + N02 + M
°3 + N02
NO, + NO
Percent
'Uncertainty
± 100 %
100
100
100 ,
100
100
100
100
100
100
100
100
100
100
80
70
65
45
40
40
40

-------
                  Table  1  (Concluded)
                                            Percent

   Reaction                                Uncertainty
22.  PARA +.OH                               ±40-%



23.  ALD +_ OH                                  40



24.  RO + 02                                   40



25.  PARA +0                                  35



26.  H02 + NO                                  30  -



27.  OLEF + 03                                 30



28.  0 + N02                                   25
29.  N02.+ hv                                  20



30.  03 + NO                            .       20



31.  NO. + N09                                 20
       0     C.


32.  H202 + hv              '                   20



33.  OLEF +0                                  20

                                 «


34.  0 + NO + M                                15



35.  0 + 62 + M             .                  10




36.  NoOr                                       5

-------
                                                                              20
     The small  amount of experimental  data available upon which  to base our
estimates limited the procedure we followed to estimate the uncertainty
bound for each  rate constant.   In  essence, we calculated the percentage de-
viation of the  highest and lowest  expected values  of each rate constant,
having surveyed the literature to  find independent determinations  of these
rate constants.  Thus, the so-called estimate of uncertainty bounds is, in
fact, simply an indication of the  degree of agreement (or more precisely,
the disagreement) among a number of independent determinations of  the same
rate parameter.  In three situations,  this "definition" does not apply:

     >  If only a single determination was made for a given rate
        constant, the uncertainty  bound is an indication of the
        precision of the experiment.
     >  Since photolysis rate constants (e.g., k,)  must be deter-
        mined in situ for a smog chamber experiment, the bounds
        are an  indication of the reliability of the experimental
        method.
     >  The uncertainty estimate for k3y (HCL + HCL) was taken from
        Lloyd (1974), who reviewed the reactions of the HCL radical.

Because of the  imprecision of these estimates, we  assumed that the uncertainty
bounds were symmetric about the nominal value.  Therefore, although we esti-
mated the uncertainties associated with several  parameters to be +100 percent,
the true upper  bound may be considerably higher.

b.   Results of the Sensitivity Analysis

     For the purpose of ranking the parameters by  sensitivity, we  averaged  the
values of the area indices calculated  for plus and minus percentage variations
in the parameters to obtain a single characteristic value.  Although this  pro-
cedure facilitated ranking, some information was lost in the process.  Because
each of the measures of sensitivity is based on the difference between nominal

-------
                                                                              21
and perturbed concentration-time  profiles,  the  magnitude  of  each  difference,

in general,  depends  upon  the  degree  of  perturbation  (e.g.,   10,   25,   50  per-

cent).   Because the  equations governing the kinetics  are  nonlinear, the values
of the  sensitivity measures typically are not  identical for  plus  and  minus

perturbations in a given  parameter.  But, in examining  the values  of  the  "area"

sensitivity  measures for  plus and minus perturbations,  we found them  to agree

sufficiently well  to justify  using their average  values to rank the parameters.


     Since we were interested chiefly in quantifying  the  overall  sensitivity

(or insensitivity) associated with each reaction  in  the model, the use of an
indicator based on the changes in the predictions  of  several  species  was  ap-

propriate.  Thus,  we combined the five  area sensitivity measures  into a single
scalar, which we term the "sensitivity." We defined-the  synthesized  sensitiv-

ity measure  for each rate constant as follows:
                    Sensitivity.
                               J
where
          i      =  1,  2,  3,  4,  and  5  refer  to  the  average  area  sensitivity
                   measures  for NOg,  NO,  03, olefin,  and paraffin,
                   respectively.

          j      =  the number of the  reaction  rate constant  in  the  kinetic
                   mechanism.

          w.     =  the weighting of the  individual  measure in the combined
           1        sensitivity  scalar.   (We weighted  each  of the five  mea-
                   sures  equally because, in developing and  evaluating the
                   general  kinetic  mechanism,  we were interested in predict-
                   ing the  concentration-time  behavior of  each  of these
                   species with equal  accuracy.  However,  we might  have
                   chosen different weights if our goals were-different.   For
                   example,  to  predict oxidant and NOX for evaluating  alter-
                   native emission  control  strategies, we  might have weighted
                   03,.NO,  and  N02  more  heavily.)

-------
                                                                              22
          A. .    =  the average area measurement determined for the i-th
           1J       criterion  and the j-th  reaction.
          A     =  the maximum value of the i-th criterion observed for
              i     any of the reactions.   (Dividing  by the maximum value
                   scales each of the arrays of the  five individual  area
                   measurements between zero and one.)

Table 2 presents the values of this "sensitivity"  index (seventh  column),  along
with the values  of the five averaged individual  area indices.   Table 3 ranks
the reactions  according to the combined sensitivity.   As Table 3  shows,  the fol-
lowing rate constants, presented in the order of decreasing sensitivity, display
the greateast  overall  sensitivity:

                  Rate
                Constant                    Identification
                   k-,                  N02  photolysis
                   kpo                Oxidation  of n-butane by OH
                   k20                Oxidation  of NO by H02
                   k.-.                  Reaction  of 0~  with NO
                    *5                               O
                   k16                Photolysis of HN02

                   k?~                Oxidation  of propylene by 0^

                   kpq                Photolysis of aldehydes

                   kp*                Reaction  of OH  with propylene

     Just as there are critical  parameters in  the model  whose values must be
determined with certainty,  some  parameters are  almost insensitive.  Large vari-
ations in the magnitude of  these parameters result in small changes in model
predictions.  By identifying those reactions that contribute minimally to the
total  predicted response, the sensitivity  analysis provides the basis for elimi-
nating reactions from  the mechanism.   Removal,  of course, is subject to further
limited individual testing  of each reaction over a range of initial conditions
and bounds of uncertainty.

-------
               Table 2

INDIVIDUAL AREA AND SENSITIVITY INDICES
NQ.
- 1
"2
3
4
"5
6
7
~ 9 	
" 9
'10 "
~11
12 '
13
""14"
'15"
16
17~~
18
19
"20"'
'21 '
22
"23"
24
25
~26""
Z7 "
28 -
~29"'
30
31
~32~~
33
34
-35""
36
37
~38~
'39
A (N02)
4.0999994E 00
2.9399997E-01
3.5400000E 00
9.4299972E-02
J.9699997E-01"
9.42.99972E-02
3.4399996E 00
'3.5099993E""Ob~
3.1999998E 00 ""
3.479999SE OQ" '
2.-6999998E""00"-
e.Z599993E 00 :
9.9599999E-01 '•
2-4799995E""00
1.6799994E 00'
5.0799999E 00' "
3.7799997E"'00~
a-559999SE On
0. 0
8.4699993E'"00""
1 .3699999E 00 	
3.2299995E-0]""
'6.9899998E~'00^~
2.2399998E 00
0. 0
'0,0 "" ~ 	
8.2799971E-02""
5.6199999E 00 "
"6.6099997E"00
1.5799999E 00
1.0400000E 00
'9. 1899997E--01
1 .0499992E 00
6. 1299998E-01 ' "
'3.8299996E-01~
1 .7699999E-0.1
4-1499996E. 00
"2;0400000E-0'1 —
3.1699997E~01"""
A (NO)
3.0100000F-01
l'.5199995E-6r
2.3799998E-OI
5.8300000E-02
'4.7399998E-02'
5.S399998E-02
J.5199995E-01
9.3699992E-02""
8.8999987E-02
9.3499959E-02
9. 5 4 9 '9 9 9 2E- 02."
7.1199954F-0.2
5.7499997E-02
'l.'2599993E""00
7.2599995E-01
1.9299994E 00
r.'0499992E~00"
1.1999998E 00
0.0 ''
6.4899999E-OT
8.7499976E-02
I.2799996E-01
"4'.'5400000E'-OT
8.2699996E-01
'o.o
"0". 0 ":
'4.9599998E-02
'1.5999994E 00
'8.7399995E-01
7.5099945E-02
7.9399943E-02
'r.-'7499995E-01
1.1099994E-01
3.4599996E-OI
T.2399995E-OI
1.0799998E-01
2.7099997E-01
"S;"8300000E-0'2
'6.2099997E-02
A(03)
'• 4.B099991E 01
'5.8699995E-01'
4.2899994E 01
1 .5199995E-01
6.3000000E-01"
- 1 .4699996E-01
'"•li2000000E 0]
5.2399998E"00'
"'4.5699997E 00
	 5.179999^6! 00
. 3".'5"699997E'"00'
'4. 1699991E. no
8.6199999E-01
~~4T'6699997E"00'
~"3.0799999E "00
"7.5000000F 00
~"6".8199987E~"00'
3.9299994E 00
o.o
" "2.1699997E""OI
'""1.4699993F' 00
- 5.9899990E-01
•~"4":'1"199999E~00'
. 7.7199993E .00
••o.o.
0 . 0
•"8.6S99996E-02
1.2500000E 01
1V7000000E"01
8.6499996E 00
3-2099991E 00
7.7599993E""00
"'8.6499996E no
'"3-5AOOOOOE no
2'.2699995E" no
4.5599997E-01
1 .0200000E 01
4T3599999E-01
"""9".4499999£-01
A(OLEF)
5.9399996E 00
3.7799996E-01"
4.4599991E 00
3.4999996E-02.
~2.5999997Ei-02"
3.4299999E-02
7.7999997E-01
"2'.'8599995E-01"
" 3.4199998E-01"
2.8399998E-01 '
l'.3060000E-01~
3.02.99997E-01
2.6499998E-02
"l'.4899998E'"00"
1.0999994E 00
" 2.799999?E 00
"T2.'3799992E "00"
': 1 .4499998E. 00
0.0
~-"3."2799997E"""00"
'"2.3099995E-01 '
••"' 3.0100000E-01 '
7.1'099997E~00~
: 5.4199991E 00
"0.0
•"0:0 — 	 "
'" 1.6500000E-02'
' 1 .8799996E-01
"~2.4799995E 00'
1.3499994E 00
, 1.7299998E-01
"'"7. 1499997E-01"
'" 7.3699999E-01
' '6.9699997E-01
— 4'.0099996E-or
1.2899995E-01
1.4799QOSE 00
177199997E-02-
"' '2.5099996E-02'
    ft(PARA)
• 1.1099999E  01
'7.1099997E-01
 9.1799994E  00
 6.1699998'E-02
"3.0299997E-or~
 6.0699999E-02
_1.9899998E  00
 5.9499997EXU '
'4.9599999E-01
 S.8999937E-01
"3.6199999E-01'
 1.2500000E  00
 4-8899996E-01
"2;2900000E  '00'
 2.0199995E  00
 5.15S9998E  00
"5.4899998E' 00 "
 2.0299997E  00
 0.0
"6.7199993E  00"
'2.0499992E  00
 1.6999996E-01
"1 . 1199999E"00'
 S.0999994E  00
 0.0
"0.0	
' 7.9099953E-02
 2.0699997E  01
""4 .3899994E  00'
 7.7999992E  00
 1.1399996E-OJ
~"3.2099998E:-01
 3.3099997E-01
' 1.0299997E  00
"'6.1999995E-01'
 1.6499996£-01
 3.309999-5E  00
-4.149999SE-02
'•7.1799994E-02
                                                      SENSITIVITY.
                                                     6.0233879E-01
                                                     "4.2636652E-02"
                                                     5.0078332E-01
                                                     1.0480S03E-02
                                                     '1 .5842Q65E-0?
                                                     1 .0441024E-02
                                                     1.8804300E.-01
                                                     "1 .2817216E-01"
                                                     1 .1538500E-01
                                                     1 .2708902E-01'
                                                     "9.5649123E-0?"
                                                     9.8632284E-0?
                                                     3.853H06E-OP
                                                     ~2.?009088E-Ol~
                                                     1.7816830E-01
                                                     4.7975492E-01
                                                     "3.4641325E-01 '
                                                     2.6154286E-01
                                                     0.0
                                                     ' 5.14674aSE-Ol "
                                                     7.3833704E-02
                                                     '3.3491254E-02
                                                     '4.400S167E-01'
                                                     3. 7242830E-01
                                                     0.0
                                                     0.0   '  '  • " -'
                                                     9.6847395E-03
                                                     5.5576986E-0]
                                                      }.9439411E-01
                                                      5.2100249E-n2
                                                     "9.5314741r-02
                                                      9.4606638E-02
                                                     '4.860229-4E-n2'
                                                      2.2490099E-OZ
                                                     "1.3556119E-02
                                                      K9249547E-02
                                                                              CO

-------
                                                                   24
                      Table 3



         THE REACTIONS RANKED BY SENSITIVITY





    Reaction                            Sensitivity



 1.  N02 +  hv                              0.60



 2.  PARA i OH                             0.56



 3.  H02 +  NO           '.                   0.51



 4.  03 + NO                               0.50



 5.  HN02 + hv                             0.48



 6.  OLEF + 03                             0.44



 7.  ALD +  hv                              0.43



 8..  OLEF + OH                             0.37



 9.  OH + N02                              0.35



10.  NO + N02 + H20                        0.27



11.  OH + NO                         '      0.26



12.  H02 +  H02                             0.24



13.  ALD +  OH                              0.19



14.  03 + N02                              0..19



15.  HN02 + HN02                           0.18



16.  N03 +  NO                              0.13



17.  N205                                  0.13



18.  N03 +  N02                         .    0.12



19.  NO + HN03                             0.10



20.  RC03 + N02                            0.10



21.  N90, + H90                            0.10
      c. 0   • c.


22.  RCO, + NO                             0.10

-------
                                                                    25
                 Table 3 (Concluded)
     Reaction                            Sensitivity
23.  RO + 02                                 0.09
24.  H2~02 + hv                  .             0.07
25.  R02 + NO                        '        0.05
26.  RO + N02   '       •                      0.05
27.  0 + 02 + M                     ' '        0.04
28..  HN02 + HN03        '                     0.04
29.  OLEF + 0                                0.03
30.  RO + NO                                 0.02
31.  R02 + R02                               0.02
32.  0 + N0                                  0.02

33.  HO, + R09                              0.01
       ^     t-
34.  0 + NO + M              '               0.01
35. -0 + N02 + M                            0.01
36.  PARA + 0                               0.01

-------
                                                                              26
     Reactions that could potentially be  removed from the  mechanism,  based
on the results of the sensitivity analysis,  appear in the  lower portion of
Table 3.   These reactions included the oxygen  atom oxidation of the species
tabulated below:
                          Species
                Paraffins
                NO
                N09 in the presence of
                  L a third body (M)
                N0? second order reaction
                Olefins
Reaction

:  k27
 >k.
Other candidates included the following:
                          Species
                         HN02-HN03
                         RO-NO
                         R02-R02
Reaction
                                              36
                                             ^38
  '39
Finally, we included among the candidates for potential  elimination the reac-
tion between NO and HN03 (k,2).   After carrying out the  sensitivity calcula-
tions, we learned that the experimental  value of this rate constant was
several orders of magnitude less than our earlier estimate; this change
sharply decreased the sensitivity of the reaction.

c.   Elimination of Insensitive Reactions
     We based the tentative conclusions reached thus far largely on the averaged
sensitivity criteria characterizing a single set of initial  reactant concentra-
tions.  If we were to repeat the calculations using only half the initial hydro-
carbon and twice the initial NO  used in the present study,  we would expect to
                               A

-------
                                                                              27
find the order of parameter ranking to be somewhat different than that given
in Table 3.   Thus, we had to scrutinize each reaction carefully prior to its
elimination; our criterion for elimination was that the reaction be "insensi-
tive" (a term defined quantitatively shortly) over the range of initial con-
centrations of interest, as well as over the uncertainty bounds of the
reaction rate constant (Table 1).

     We chose three EPA smog chamber runs as a representative set of initial
concentrations and ratios over which to evaluate the reactions for possible
removal.  As shown in Table 45 two of these runs are binary hydrocarbon sys-
tems; in the other experiment, propylene was the only hydrocarbon present.
These three  runs  span a total  hydrocarbon-to-NO  initial  ratio of 0.7 to 10.5
                          *                    x
and a reactive hydrocarbon-to-NO  initial ratio of 0.2 to 0.8.  Air quality
                                A
data obtained in  Los Angeles indicate that the ratios for polluted air there
are often within  these ranges.  (In the atmosphere, the ratio can vary with
both location and time of day.)

                                   Table 4
                 CHARACTERISTICS OF THE SMOG CHAMBER RUNS
                        Concentrations

FPA
Pun [n-butane]0
329
333 3.40
352 3.29
(ppm)

[propylene]Q [NO]Q
0.24
0.23
0.26
0.29
1.25
0.27

[N02JQ
0.06
0.08
0.07
Total
[HC]/[NOxJ
0.7
2.7
10.5
High
Reactivity
[HC]/[NOx]
0.7
0.2
0.8
     We considered reactions to be "insensitive"  if, upon their removal  indi-
vidually and as a group, the remaining set of reactions was able to predict
the following within 10 percent of the values predicted by the complete
mechanism:
 Defined as  propylene.

-------
                                                                              28
     >  The time to the NCL peak (T)
     >  The height of the N02 peak (H)
     >  The magnitude of the ozone peak (M).

These three scalars, all of which can be easily quantified,  are of interest
because the onset of formation of many secondary products  formed in the atmos-
phere accompanies the peak concentration of NCL and because  the intensity of
smog is often associated with the ozone and N0? concentrations.  Thus,  T, H,
and M constitute three major indicators of smog formation  and severity.  Al-
though the choice of the 10 percent range was arbitrary, this value is  lower
than the uncertainty bounds associated with the experimental  chamber data
used to evaluate and "tune" the model.   Thus, we felt that the choice was
reasonable.

     Consideration of the sensitivity values  associated with  each rate  con-
stant led us to select 10 reactions for possible removal from the mechanism,
Of these, we found that only six could actually be eliminated based on  the
criteria cited above:

                           Species       Reaction
                         0 + NO             k4
                         0 + N00 + M        kc
                               <-             0
                         NO + HN03          k]2
                         HN02 + HN03        k]3
As shown in Table 5, the values of T, H, and M after removal  of these six
reactions were within 10 percent of the values before the reactions were
eliminated.  Several other reactions could have been removed  for one or two
of the EPA runs, but not for all  three.  However, since their elimination
would have limited the applicability of the kinetic mechanism to a narrower
range of initial concentrations and ratios, we did not drop them.

-------
                                                                              29
                                 Table 5
                 VALUES  OF T,  H,  AND M BEFORE AND AFTER
                      REMOVAL  OF  THE SIX REACTIONS   '
                                              Concentration
Height of the
NOp Peak (H)
Before After
0.25 0.25
0.75 0.70
0.25 0.25
Magnitude of the
Ozone Peak (M)
Before After
0.39 0.40
0.40 0.41
0.50 0.52
                Time to the
                NO?  Peak (T)
        EP/\       (minutes)
        Run    Before    After
        329       87       86
        333      285      281
        352       65       70
     The conclusions we reached during this  study were based on the lumped
general  mechanism.   If this  mechanism proves to be fundamentally inadequate,
the sensitivity calculations should be repeated withithe corrected mechanism,
and reactions that we eliminated should be examined again to judge their sen-
sitivity in the environment  of the corrected mechanism.

3.   Further Modifications To Reduce Computing Requirements

     Although the elimination of unnecessary reactions saves computing time,
the condensation discussed thus far focused  primarily on giving prominence to
the important reactions in the mechanism.   Significantly greater reductions in
computing time can  be obtained by varying  the mathematical  representation of
the chemical mechanism.  From a purist's point of view,  a series of differen-
tial  rate equations most accurately represents changes in the concentrations
of reactants with time.  (Ideally, one would solve these equations analyti-
cally.   We used numerical  methods to solve the equations on  the computer, but
these techniques were evaluated using test systems of equations for which
analytical solutions were available.)  Over  the years, scientists have used
the following approximations and simplifications to facilitate the solution of
complex  kinetic systems:

-------
                                                                              30
     >  Recognizing  the  fundamental  mathematical  difference  between  the
        differential  equations  of  species  that  are  produced  but  do not
        enter into  reactions  and those  of  species that  do  react.  The
        differential  equations  for reactants  are  often  mathematically
        coupled  and  must therefore be solved  simultaneously.   If these
        coupled  species  have  vastly different characteristic  times of
        reaction.,  the equations become  "stiff"  numerically and must  be
        solved using very small time steps  to preserve  accuracy.  In con-
        trast, the  differential equations  for species that do not react
        are not  coupled  and can be solved  accurately one at  a time using
        a method as  simple as Simpson's rule.
     >  Applying the steady-state  approximation.  If the concentration
        of a species equilibrates  rapidly  (relative to  many  other species
        in the system),  one can assume  that the summation  of  the  rate terms
        for formation and consumption of the  species is identically  zero.
        This assumption  reduces the differential  equation  to  an  algebraic
        equation.
     >  Combining  second-order  reactions into higher order reactions.   In
        some special  cases, two or more reactions can be combined into  a
        single reaction, with the  elimination of  an intermediate as  well.

The following subsections summarize the results of  applying  each  of  these  tech-
niques  to the lumped kinetic  mechanism.

a.   Treatment of Uncoupled Species

     In a system containing propylene,  n-butane,  NO , and  air, four  species
                                                   X
form that do not react subsequently: nitric  acid,  peroxyacylnitrates,  organic
nitrites, and organic nitrates.  Because these  products do not enter into
reactions with other species  present in the system, we  can uncouple  and  solve
the differential  equations for  each of  the  four species independently.

-------
                                                                              31
b.    Invocation of the Steady-State Approximation

     In earlier work,  we demonstrated the validity of the  steady-state  approxi-
mation for 0, OH, RO,  and N03 (Hecht et al . ,  1973).   Recently,  we  justified the
application of the approximation to obtain  predictions of  the  concentrations of
H0?, N20r, R02, and RCCL.  To demonstrate the validity of  the  approximation for
any given species, we  compared the concentration-time profile  for  the species
predicted by an algebraic description with  that predicted  by a  differential
expression.  In so doing, we found that the profiles  generated  using  the  two
mathematical representations agreed to within 0.1  percent  for  these species.
Thus, we eliminated four additional  coupled differential equations, which were
replaced by four coupled algebraic equations.

c.    Combination of Reactions into a Single Higher Order Reaction

     The species N^Or  exists in equilibrium with NO,,  and NO,,:
                         N03
The only important reaction of N205 other than Reaction  II  is  hydrolysis  to form
nitric acid, a stable product in the mechanism:


                        N2°5 + H2°  +11  2HN03

If we assume that N205 is  in a steady-state (we  have established the validity
of this assumption)  we can combine these reactions into  the single third-order
reaction:

                                      TV
                     N03 + N02 + H20  iv 2HN03

-------
                                                                              32
having the rate constant
                           kiv
The combination of these three reactions  eliminates  NpO^  as  a  species,  thereby
saving one algebraic equation  and removing  a  net  total  of two  reactions  from
the mechanism.

4.   The Present Status  of the Mechanism

     As a result of the  procedures  described  thus far,  we added  nine  reactions
to the mechanism.   Thus, a total  of 31  reactions  are necessary to  represent
the chemistry of a system of paraffins, olefins,  NO  , CO  and air.   In addition,
                                                  A
to facilitate usage of the mechanism in the airshed  model, we  included  two
additional reactions involving 0  and OH reactions with  aromatics.   It is to  be
understood, however, that this is an interim  treatment  of the  chemistry  involv-
ing aromatics and  is subject to revision  at such  time as  a more  suitable mech-
anism is developed.  Table 6 presents the revised mechanism.   Of the  25  species
included in the mechanism, 10  are represented by  coupled  differential  equations,
7 by algebraic equations, and  4 are constant, as  shown  in Table  7.   Although the
computing time associated with individual sets of initial conditions  varies  be-
cause of changes in the  stiffness of the  system of equations,  we found that
incorporating the changes presented here  reduced  the- required  computing time by
approximately 50 percent over that required previously  (Hecht  et al.  1973).
This saving is significant enough to justify  the  replacement of the simplified
15-step mechanism by the more  accurate lumped kinetic mechanism as the kinetics
module in the airshed dispersion model.

-------
                     Table  6

A LUMPED KINETIC MECHANISM  FOR PHOTOCHEMICAL SMOG
                                                                 33
       N0
                      + 0

                      + M
                                       N02~NO-03 Cycle
       0 + N0
                  -N0
       N03 + NO~^2N02

      Wo + HoO-J-^2HNO,
                                       Chemistry of NO,
NO + N02 + 2H20-

          2HN00-
      HN02 +
                      + NO
                                        Chemistry of HNO,
      OH +
       OH +
               12
 OH + CO
                         H0
                                         Important  Reactions  of
                                         OH with  Inorganic  Species
      H02 + NO-11*-OH + N02
                                        Oxidation  of NO  by  H02
     H0
                                         Hydroperoxy  Radical
                                         Termination
               16
                                        Photolysis of

-------
Table 6 (Concluded)
HC1 + 0-*-*
ur 4- n
nu-i T u^j
1 O
HC1 + OH-12-*
HC2 + O-22-*
HC2 
II 2
0
RO + 02-^
RO + N02-^-
RO + NO— 1J-
-ROO + aRCOO + (l-a)H02 I
0
^-RCOO + RO + HC,
1 1
0
-ROO + HC3 '
-ROO + OH
\
^ROO + H20 /
^gROO + (2-s)H02 I
^3RCOO + (1-3)'H02 + H20
0
^-ROO + OH
*-ROO + H20
5-RO + N02
^•ROO + N02 + C02
T RCOON02
0
-H02 + HC3
*-RON02
                        Organic Oxidation
                        Reactions
                           HC, = Olefins
                           HC2 = Paraffins
                           HC3 = Aldehydes
                        ^   HC, = Aromatics
                         Reactions of Organic
                         Free Radicals with NO,
                         N02  , and 02

-------
                                                                              35
                                 Table 7

         TYPE OF MATHEMATICAL REPRESENTATION REQUIRED TO PREDICT
           CONCENTRATIONS OF SPECIES IN THE GENERAL MECHANISM
   Coupled
Differential
  Equations
  Uncoupled
Differential
  Equations

    HN00
Steady-State
 Algebraic
 Equations

   0
                                                            Constant
                                                               M
       NO
       HN0
    PAN

    RNO,

    RNO,
                                       N0
                                             OH
                                                         H20
                                             RO
 CO

 Olefins
                                             RC0
        Paraffins

        Aldehydes

        Aromatics
C.   DEVELOPMENT OF A KINETIC MECHANISM DESCRIBING
     S02 REACTIONS AND SULFURIC ACID FORMATION
     During the past decade,  air pollution investigators  have focused a substan-

tial  amount of scientific attention  on SO,,, the precursor of sulfuric acid and

s ill fate, because of its  effects  on visibility and health.  They observed that

the oxidation of gaseous S02  occurs  both  through reactions with gas phase oxi-

dants and through reactions with liquid aerosol  droplets.  They demonstrated

that the addition of S0? to a reactor in  which atmospheric concentrations of

organics and NO  in air  are being irradiated (i.e., a smog simulation experiment)
               -X

-------
                                                                              36
results in a substantial  decrease in visibility due to the formation of a
sulfuric acid aerosol.  And they established that SCL is oxidized in fog.
In this section, we review current knowledge and speculation concerning
the oxidation of SCL through reactions that occur in the gas phase and in
solution.  Since Bufalini (1971) has extensively reviewed the oxidation of
SCL in polluted air, our discussion focuses primarily on more recent re-
sults.  We conclude this section with a discussion of our efforts to model
a set of dynamic organic-NO -SCL smog chamber experiments and a summariza-
                           A   L.
tion of our future plans to simulate the chemistry of SCL.

1.   The State of the Art of Gas Phase SCL Kinetics
     Until recently, air pollution SCL research focused primarily on the
qualitative and semi-quantitative characterization of the interaction of
SCL with components of smog.   Scientists have been particularly interested
'in evaluating the effect of SCL on oxidant levels and visibility in simu-
lated smog (irradiated mixtures of organics, NO , SCL, and air); they have
                                               X    L.
used environmental chambers extensively for this purpose.  In these experi-
ments, they observed that the concentration of SCL slowly diminishes with
time.  However, most of the early (prior to 1970) experiments were not con-
trolled carefully enough to allow 'an accurate-.estimate to be made of the
rate of S0? oxidation due to  gas phase chemical reactions.  Variations in
relative humidity, the reactivity of chamber surfaces, and the accuracy of
the analytical instrumentation all served to introduce imprecision into the
data.  And, by their very nature, smog chamber experiments provide minimal
insight into the actual individual reactions by which S0? is oxidized in
smog.  Observations are limited to macroscopic changes in the concentrations
of the major reactants and products with time.  The results of recent chamber
experiments and detailed kinetic studies of elementary reactions have pro-
vided sufficient insight so that we can now postulate a provisional mechanism
for the oxidation of S02 by homogeneous gas phase reactions.  We discuss this
10-step mechanism briefly below.

-------
                                                                              37
     Experimental  studies have indicated that peroxy radicals, diradicals,
and hydroxyl  radicals are the most potent gas phase oxidizing agents with
respect to S(L in  photochemical smog.  Davis et al. (1973) obtained a pre-
liminary measurement of the rate constant for the reaction

                                     Sl
                          H02 + S02  -*1  OH + S03

of 0.45 ppm  min  .   The observed rate is sufficiently high to suggest that
the HO?~SO? reaction is important at about the time that N0? reaches its
maximum values and 0., begins to accumulate.  Studies of S0? in smog Simula-
                    •3                                     L—
tion experiments have shown that this is the time at which the oxidation
rate of S0? is greatest.  H0? is, of course, generally regarded as the prin-
cipal oxidant of NO:
                           H02 + NO ->- OH + N02
     Because of the functional similarity of peroxyalkyl and peroxyacyl
radicals to H02, it does not seem unreasonable to presume that these three
species would undergo the same chemical reactions with a given reductant.
Both R0? and RCOo apparently oxidize NO through a reaction similar to the
HOp-NO reaction:

                           R02 + NO -> RO + N02
                          RCOO + NO -> R£0 + N02
                           6           0

Although the rate constants for these reactions are not known yet, the reac-
tions are thought to proceed more rapdily than the H02-N0 reaction.  We feel
that, because of the analogies between the structure and behavior of H02,
R09, and RCO-, the last two species oxidize S09 at a rate somewhat faster
                                                               -1-1
than that of H09.  We therefore estimate that k<.  = kc '= 1 ppm  min   :

-------
                                                                              38
                                 S2
                                 +   RO
                                 S3
                     RCOO + S09  _/  RCO + SO,
                      II       ^-       II      O
                      0               0

     Cox and Penkett (1972) observed that SCL forms with reasonable rapidity
                                            •j
when a system containing 0.,, olefin, and SCL react, and they postulated that
diradicals, products of the 0^-olefin reaction, are the species that oxidize
S0£:

                                 S/L
                     R2COO + S02 +  R2CO + S03

Since diradicals apparently form in smog only as a result of CL-olefin reac-
tions, this reaction, depending on its rate, may be less important that
Reactions S, through S~ in polluted air, where normally less than 20 percent
of the organics are olefinic.   O'Neal and Blumstein (1973) recently reconsid-
ered the mechanism of the 0,-,-olefin reaction, and they feel that the interme-
diate complex of the reaction  may decompose to form free radicals, including
H.  A hydrogen atom formed in  this manner could combine with 09 to form H09,
which is known to oxidize SO,,  (Reaction S,).  Thus, in the Cox and Penkett
experiments, H02, rather than  a di radical, may have been the specie generated
by the Oo-olefin reaction that oxidized S09.  Consequently, Reaction S. is
very speculative.

     Recent measurements of the OH-S02 rate constant have suggested that the
reaction
                                    S,    0
                                     b    ,
                           OH  + S09 ^  HO S
                                  L.       |
                                          0
may be an important loss mechanism for S09 in photochemical smog.  Cox (1974)
                           -1    -1
obtained a value of 850 ppm  min   under atmospheric conditions, and Castleman
et al.  (1974)  found the value  to be 600 ppm" min" .

-------
     One can only speculate as  to subsequent reactions of HOSCL in smog [see.
for example, Smith and Urone (1974)  and references therein].   We offer one
possible reaction scheme here,  which is largely an analogy to reactions of
organic free radicals.

     We assume that CL adds to  the HOSCL radical

                                         S
                              HOS02 + 02 +6 HO S02
                                               0
and that this peroxy radical  can oxidize nitric oxide
                             Q        S    ?
                                 + NO J HOSO + NO,
                               C           I       t-
The HOSOo might abstract a hydrogen atom from an organic molecule or from an
HO, radical, forming H^O, directly:

                            0       $    °
                          HOSO + RH +8 HOSOH + R-
                            0            0
                            0        S    °
                          HOSO + H09 +8 HOSOH + 09
                            i       <-      i       <~
                            0             0

Or the HOSO,, might undergo a unimolecular decomposition reaction to form OH
and S03:

                             0  s
                           HOSO +  HO + SO,
                             i             <5
                             0

-------
                                                                              40
Sulfur trioxide is, of course,  the anhydride of sulfuric acid:
                             S03
     Although we can set forth other reactions for the HSO  radicals  describing
                                                          X
their behavior in the presence of NCL  and-other reactive  species,  we  cannot  sub-
stantiate such reactions (including S7 through S-.Q)  with  the results  of experi-
ments that have been carried out to date.

     Although we did not include several  reactions in  the core  mechanism (S,
through S-.Q) for the oxidation of SCL, some comments about them are in order.
The 0-S02 reaction, for example,

                            0 + S02 +  M -> S03 + M

has a reasonably high rate constant but is, nevertheless, slow  because of the
extremely low concentration of oxygen  atoms in smog.  The direct photolysis  of
S02 in otherwise clean air results in  the slow disappearance of S02,  but the
rate of S02 loss is not comparable to  the rates observed  in polluted  air.
Wilson and Levy (1969) showed that N02 reacts very slowly with  SO,,.  Calvert
(1975) determined upper limits for the rates of reaction  of NO^ and N?0r with
          r-in            fill                 O      £ D
S02 of 10   ppnf inin"  and 6 x 10~  ppnf  mirf , respectively.  Consequently,
both of the reactions are of negligible importance in  photochemical smog.   In
addition, Calvert found, in agreement  with others, that the 0~-S09 reaction
                                                -8    -1    -1
is very slow, having a rate constant of about 10   ppm  min  .   In summary,
each of this last group of reactions results in the slow  oxidation of S02 to
S0~.  Although we could have included  in  the core S02  mechanism, the results
of kinetic studies of these reactions  suggest that their combined contribution
to the total predicted loss rate of S02 is minor.

     Because kineticists have studied  in  detail only two  of the ten elementary
reactions included in the mechanism for the gas phase  oxidation of S02 the
mechanism has an extremely high level  of uncertainty.   EPA is presently fund-
ing investigations of some of these reactions; therefore, more  accurate values

-------
                                                                              41
of the corresponding rate constants may be forthcoming in the near future.
Despite the uncertainty, we attempted to test this mechanism using smog
chamber data.   (Section C-3 describes these efforts.)   However,  we found
that the chamber data were inadequately characterized  in many important
respects and,  consequently, were unusable.

2.   The State of the Art Regarding the Oxidation
     of SQ0 in Solution
          2_

     A large percentage of the volume of aerosol  particles consists of water.
Gas phase borne SCL can dissolve in these particles, especially  in the envi-
ronment of a stack plume, where the S0? concentration  is often high.   Once
S02 is dissolved, both direct and catalyzed reactions  apparently lead to the
oxidation of S09 to sulfate.  However, it is not  now possible to assess the
relative importance of these competitive pathways under conditions of photo-
chemical smog  formation.  Certainly, the contribution  of these two types of
reactions to the total S0? oxidation rate in solution  depends on such factors
as aerosol  size, oxidant concentration, catalyst  concentration,  species of
oxidant present, catalyst species present, and other chemical species in the
droplet that might enter into reactions with the  oxidants or the catalysts.
In this section, we identify possible important direct and catalyzed reactions
in solution and attempt to explain the mechanisms of these reactions.

a.   Reactions of S00 in Solution with
     Oxidants Produced in the Gas Phase
     Investigators have studied the reactions of S02 in solution with three
products of photochemical  smog:  N02, 03> and HpSO^.
                          N09 + S09	»-NO + SO
                            L     *• H0 [t]
                               S09 ——-^ 09 + S00
                                            L     3
                         H9SO. + SO.,	*- H9SO., +  S00
                          24     3          2   3     3

-------
                                                                              42
The first of these reactions (Wilson et al. , 1972) and the last (Gerhard and
Johnstone, 1955) proceed very slowly.  Ozone and SCL, however, react rapidly
in the presence of liquid water, and the reaction probably occurs in solution
(Wilson and Levy, 1969).  The rate of this  aqueous reaction contrasts sharply
with that of the gas phase 0.,-SCL reaction,  which is extremely slow.  Thus,
the reaction between SCL and CL could be significant in aerosol particles,
and measurement of the rate constant of the  reaction in a simulated atmos-
pheric environment is important.

b.   Direct and Catalyzed Reactions of SOg
     with Metal Ions in Solution

     As reported in the literature, SCL is  slowly oxidized when dissolved in
water (probably through a direct reaction with dissolved oxygen); however,
                                      +2    +2    +3    +2        +2
the presence of metal ions, such as Mn  , Fe  ,  Fe  5 Ni  , and Cu  , in the
.solution accelerates the oxidation rate of  SCL substantially (Urone and
Schroeder, 1969; Bufalini, 1971).  The metal  ions can interact chemically with
SCL in either or both of two ways:  through  direct reaction with SCL or through
catalysis of the (dissolved) air oxidation  of SCL.  We now turn to a discussion
of each of these classes of reactions.

     Direct Oxidation-Reduction Reactions Between Metal Ions and S00.  An exam-
                                                                   j/
ination of half-cell potentials provides a  straightforward means of evaluating
whether a given reaction is expected to occur on the basis of purely thermody-
namic considerations.  In the context of this discussion, we are particularly
interested in learning whether oxidation-reduction couples (i.e., reactions)
between S00 and metal ions result in the oxidation of S09 to SCL (or.SCL) and
          e.                                             <-*     3       lr
the reduction of metal  ions to some lower oxidation state .

     We first observe that according to predictions, CL, CL, and FLCL should
all oxidize SCL.  Noting that the SCL-SCL half-reaction is:
*
 We used reduction potentials for these calculations; thus,  for a reaction
 couple to be favored, the combined potential  must be positive.

-------
                                                                               43
     S02° x H20 -v SO"2 + 4H+ + (x - 2) HgO + 2e~    ,    E° = -0.17 V
we see that S09 is oxidized as a result of any of the following half-reactions
                    2H+ + 2e~ -> H0     ,        E° = 0.682 V
                    2H+ + 2e~ -»-.0  + H0    ,    E° = 2.07 V
                    2H+ + 2e~ -> 2\\      ,        E° = 1.776 V
The coupled potentials.are, therefore, positive by 0.51 V, 1.90 V, and 1.61 V,
respectively.

     Of the five metal cations known to "oxidize" S02, only two would be pre-
dicted to enter into direct reaction with ?09 on the basis of thermodynamic
                         +3       +2
considerations alone:  Fe   and Cu  .   Their respective half-cell potentials
are
                      Cu+2 + 2e~ •* Cu°    ,         E° - 0.34 V

                      Fe+3 + e" -»• Fe+2    ,         E° = 0.77 V

                                   +2    +2        +2
Direct reactions between S09 and Mn  , Fe  , and Ni   are extremely unfavored.
Their  respective half-cell potentials are:
                      Mn+2 + 2e~ -> Mn    ,          E° = -2.375 V

                      Fe+2 + 2e" -> Fe               E° = -0.41 V

                      Ni+2 + 2e" + Ni    ,          E° = -0.23 V

                                                              +?    +?
These data indicate that for the direct oxidation of S09 by Mn   , Fe   , and
  +?
Ni   to occur, one would have to apply 2.54 V, 0.58 V, and 0.40 V,  respectively,
of energy to the reacting system.

-------
                                                                              44
     Theoretical  results  such as  these should,  of course, be subjected to
experimental  scrutiny.   In fact,  experimenters  have observed the direct
reactions between SCL and 0?, CL, and H?0?,  in  a water solution that are
predicted to  take place on the basis  of thermodynamic principles.   The two
          +3        +2
cations Fe   and Cu   are known to accelerate the rate of oxidation of S0?.
However, it has not yet been shown (to our knowledge) that the mechanism
                         i Q       J_ O                                 _L O
of oxidation  of S0? by Fe   and Cu   is direct.   The isolation of Fe   and
                                                             +3        +2
Cu as products of reactions in an aqueous  solution of SCL, Fe  , and Cu
would, for example, constitute acceptable  evidence for the direct oxidation
mechanism.  (It is important to remember the limitations of these electro-
chemical cell calculations.  Although half-cell  potentials provide a means
of predicting the direction of a  chemical  reaction, they do not in any way
                                                          +2    +2
indicate the  rate at which the reaction will proceed.)  Mn  , Fe  , and
  +2
Ni   do not enter into a  direct reaction with SCL unless energy is supplied
to the system; thus, their roles  in the oxidation process must be catalytic
or indirect.
     Catalytic Oxidation of SO,,.   Catalytic oxidation may well  be the prin-
cipal process for SCL conversion  under conditions  of high humidity and high
particulate concentration,  such as those that exist in plumes  from power
plants.   Gartrell et al. (1963) reported, for example, that the rate of SCL
oxidation in a smoke plume  was quite low for relative humidities less than
70 percent, but it increased markedly for higher humidities.   In one case,
they measured a rate of  SCL conversion of 55 percent in 108 minutes.  Al-
though such a rate is too high to be accounted for by a photochemical mech-
anism [a  conclusion based on early studies of the  photochemical oxidation
of S02 by Gerhard and Johnstone (1955)], it is similar to that expected of
oxidation in solution in the presence of a catalyst.   Since the metal sul-
fates (and chlorides) emitted in  a plume from a coal-burning process are
potential catalysts for  the liquid phase oxidation of SCL, a reasonable ex-
planation for this process  is that these particles act as condensation nuclei,
producing droplets of metal  salt  solution, which then act as loci for the SO,,
conversion.

-------
                                                                              45
     The atmospheric catalytic oxidation of SCL involves both water and
            ••*                                <-
dissolved CL,   .and it requires the presence of a catalyst:

                                        catalyst
Catalysts for this reaction include several  metal  salts, such as sulfates
and chlorides of manganese and iron, which usually exist in air as suspended
particulate matter.   At high humidities, these particles act as condensation
nuclei or undergo hydration to become solution droplets.  The oxidation pro-
cess then proceeds by absorption of both SCL and CL by the liquid aerosol,
with a subsequent chemical reaction in the liquid phase.

     Early experiments conducted by Johnstone and Coughanowr (1958) and
Johnstone and Moll (I960), in which they measured SCL oxidation in droplets
of MnSCL, confirmed the basic catalytic mechanism.  In addition, studies per-
formed by Junge and Ryan (1958)  of the oxidation of -SCL in' bulk- catalyst solu-
tions yielded valuable information on the effects of solution acidity on the
rate of SCL oxidation.

     Recently, Cheng et al . (1971) reported laboratory results involving the
catalytic oxidation of SCL in aerosol drops containing metal salts.  They
developed an aerosol-stabilizing technique in which aerosol  particles were
deposited on inert supporting Teflon beads in a fluidized bed.  This deposi-
tion process altered neither the physical shape nor the chemical properties
of the aerosol.  After packing the Teflon beads with the deposited aerosol
particles into a flow reactor, in which the catalytic oxidation of SCL oc-
curred, the experimenters passed a mixture of SCL and humid air through the
reactor.   The SCL concentrations at the reactor entrance ranged from 3 to 18
ppm.  To  monitor the progress of the oxidation, Cheng et al . continuously
measured  the SCL concentration at the reactor exit.  They identified reaction
products  by analyzing the reactor contents at the completion of an experiment.
_
 The rate of the direct reaction of SCL with CL,

                           2S02 + 02 + 2S03

 is too slow at room temperature to be of importance in the atmospheric
 oxidation of SO,,.

-------
                                                                              46
     The SCL conversion progressed in three stages.  During the initial per-
iod, all of the influent SCL was converted; none appeared at the reactor
exit.  A transitional period followed, in which the SCL conversion rate de-
creased from the initial maximum value to a steady value.  From then on, a
steady-state conversion of SCL took place.  The three-stage process can be
related to the change in solubility of SCL in a water solution as the solu-
tion becomes more acidic.  The initially rapid conversion of SCL apparently
results from the high rate of dissolution of gaseous SCL into liquid catalyst
drops.  The increase in sulfuric acid in the drops soon affects the initial
stage of rapid conversion.  Because HLSCL in a dilute concentration undergoes
                                   +             +
complete dissociation to HSCh and H , the added H  concentration diminishes
the solubility of SCL.   Finally, as the solution acid concentration exceeds a
                    *-    -j-
certain level, the high H  concentration prevents further dissociation of
hLSCL, and the solubility of SCL becomes constant.  In this final stage, the
rate of conversion of SCL to sulfate e'quals the rate at which SCL is absorbed
in the drops.

     Although NaCl , CuSCL, MnCl?s and MnSCL all exhibited the same general
behavior, each salt differed in effectiveness as a catalyst for the oxida-
tion of SCL.  Table 8 shows the steady-state conversions found by Cheng et
al. (1971).  in the case of CuClg, Cheng et al . found that, rather than act-
ing as a catalyst, CuCl? reacted directly with SCL according to the following
                       t-                         C-
reaction:
                 SCL + 2CuCl2 + 2H20 i±2CuCl  + HgSO, + 2HC1

     Although the conversion of SCL proceeded even at very low relative humid-
ities (less than 40 percent), it did so slowly.   Above about 70 percent relative
humidity, which is the level at which the transition from solid crystals sur-
rounded by a layer of water to actual solution drops takes place, the rate of
conversion increased dramatically.

     The individual  steps in the liquid-phase catalytic oxidation of S02 are
as follows:

-------
                                                                              47
     >   The  gas-phase  diffusion  of  S0?  to  the  drops,
     >   The  diffusion  of  SCL  from a drop's surface to the interior,
     >   The  catalytic  reaction  in the  interior.

Under steady-state  conditions,  the  slowest of  these  three steps  limits  the
overall  rate of SCL conversion.   If the gas phase  diffusion  of SCL  to  the
drops is the controlling  step,  then the rate of  SCL  conversion should  depend
on the  gas velocity in the system.   If  the Tiquid-phase  diffusion of SCL  con-
trols the conversion rate, then  the rate can be  expected to  be independent  of
the type of  catalyst.   In varying the  gas  flow rate  through  their reactor,,
Cheng et al. found  that the overall rate of SCL  conversion was unaffected.
Since,  as the results  in  Table  8 show,  these rates clearly depend on the  type
of catalyst, the rate-controlling step  is  the  chemical  reaction  itself.   Foster
(1969)  reached similar conclusions.
                                 Table 8
            THE EFFECT OF DIFFERENT CATALYSTS ON S02 OXIDATION
Catalyst
NaCl
CuSO*
MnCl2
MnS04
Mean Resi-
Weight dence Time
(nig) (min)
0.36
0.15
0.255
0.51
1.7
1.7
0.52
0.52
Influent S02
Concentrations Fraction
(ppm) Conversion
14.4
14.4
3.3
3.3
0.069
0.068
0.052
0.365
Effective- *
ness Factor
1.0
2.4
3.5
12.2
    *
     The catalytic effectiveness  of the  various  materials  was  compared with
     that of NaCl.   Thus,  the  effectiveness  factor is  the  product of the
     ratio of the  weight of  the  catalyst in  the  reactor, the ratio of the
     reactor mean  residence  time, and  the ratio  of the reaction conversion
     of SOp in the reactor.  The  effectiveness  factor  for  MnSO^, for example,
     is:

-------
     For steady-state conversion in the atmosphere, Cheng et al.  derived the
following first-order rate expression from their data for MnSCL:
                          RSQ  = 0.67 x 10"2[S02J

where R$Q9 is the micrograms of SCL converted per minute per milligram of
MnSCL, [SCL] is the gas phase concentration of SCL in micrograms per cubic
meter, and the constant factor is for drops containing 500 ppm of MnSO,.
The factor can be altered for other catalysts using Table 8.  We can compute
the rate of conversion of S0? for conditions typical  of natural fog in an
urban atmosphere:

     >  (S02) = 0.1 ppm.
     >  The average diameter of the fog droplets is 15 \i.
     >  Half the fog droplets contain a catalyst capable of oxidizing
        S0? to HpSO..  The catalyst concentration within these droplets
        is equivalent to  500 ppm MnSO,.
     >  The fog concentration is 0.2 gram of H^O per cubic centimeter of
        air.

Under these assumptions,  the equivalent catalyst concentration is 50 micro-
grams of MnSO, per cubic  meter of air, and the rate of conversion of SOp  is
2 percent per hour.  Typical concentrations of catalyst metals are tabulated
below:
                                             Concentration
                           Catalyst            (yg m~3)

                              Mn                  '10
                              Cu                  10
                              Zn                  58
                              Fe                  74
                              Pb                  17
Thus, the conditions of the sample calculation are reasonable for actual air.

-------
                                                                              49
     The detailed mechanism of the catalized oxidation of SCL is not yet
known; however, the first step in the process may involve the association
of a reactant with the catalyst.   If the catalyst is a transition metal
cation, the reactant apparently enters into a coordination complex with
the cation; thus, the reactant occupies a position in the ligand field of
the metal.  Matteson et al. (1969) observed that catalyst potency toward
the oxidation of SCL tends to decrease as the number of possible sites at
which SCL can complex on the metal ion decreases.  Thus, the configuration
of the ligand field (e.g., square planar, octahedral) of a given metal ion
strongly influences the catalytic behavior of the ion.

     If the first step in catalysis is, indeed, the coordination of SCL
with the cation, the rate of displacement of other ligands in the ligand
field by SCL must be examined.  Some species form much stronger coordina-
tion bonds with transition metal  ions than others do.  For example, carbon
monoxide poisoning of the blood results because the binding energy of CO
to the iron in hemoglobin is much greater uhan that of CL.  Consequently,
CL cannot displace the CO from the iron, and the body rapidly depletes the
blood of 0?.  S0? can, in principle, coordinate with transition metals,
since it contains unshared electrons — a general characteristic of ligands:
                            S        ,S
                             \  *-*  /  \\
(Other ligands include, for example, FLO, NO, and CO.)   But,  if,  as a result
of this mechanism involving transition metal  cations, S0? is  to be catalyti-
cally oxidized in aerosols, it must be able to displace other ligands from
the catalyst.   Because of the high concentration of water and the presumably
low relative concentrations of S0? and catalysts in aerosols, the tendency
for S0? to displace water from the ligand field must be especially great.
Thus, an experimental  investigation of the rate of hLO displacement by SOp in
the principal  catalysts for S02 oxidation is  clearly needed.

-------
                                                                              50
     One explanation of the catalytic oxidation of S0?  in solution is  the
series of four equilibria proposed by Matteson  et al.  (1969):
                          S0  + Mn+2 £ Mn •  SO2
                          2Mn •  S092 t [(Mn  •  S092)  •  09]
                                  ^             ^      <-
                  [(Mn *  S092)  •  09J  £ 2Mn  •  SO*2
                           L         6
7
-»-
8
                     Mn •  SO*2 + HO + Mn+2 + HSO"  + H+
Matteson et al.  made three crucial  assumptions  in this  mechanism:

                                       +2
     >  S0? coordinates rapidly with Mn   (Step 1)

        The association of Mn • S09  complexes  is likely (Step 3)
     >  Oxygen  transfer to the [(Mn •  S0? )9 •  09]  complex occurs  (Step 5).

Although Matteson et al.  did not address these  issues  in formulating their
mechanism, the  series of reactions  provides  a construct for further experi-
mental and theoretical  inquiries.

     It is not  possible now to ascertain the extent to  which the oxidation of
S09 in solution  competes  with the gas  phase  reactions.   Very little data per-
taining to the  kinetics of the reactions between SO,, and dissolved salts exist
that can be incorporated  in a predictive model.  Understanding the role of S09

-------
                                                                              51
in the atmosphere and, indeed, the formulation of effective S0? control  stra-
tegies will critically depend on the fundamental  investigation of the types
of reactions discussed in this section.   Without quantitative data upon  which
to build a model, predictions are of little significance.

3.   Efforts To Test the Gas Phase Reaction Mechanism for S00
                                                            L.

     Shortly after the inception of the  project,  we received the results of a
series of smog chamber experiments from  EPA to use to test the 10-step mechan-
ism described in Section C-l as a possible explanation for the oxidation of
SO^ in the gas phase.  The experiments were carried out in a dynamic flow
reactor, and propylene, NO  , S09, and air were used as reactants.   To simulate
                          X    c.
the system, we added the SCL reactions (Reactions S,  through S,Q)  to a general
mechanism for smog (Hecht et al . , 1974).   We had previously performed exten-
sive tests of the organic-NO -air reactions using propylene-NO -air data
                            A                                 A
obtained in the same smog chamber operated in a static mode.

     Unfortunately, we found that the dynamic SCL experiments were unsuitable
for modeling for two reasons.  First, the concentration of SCL in  the inlet
tube fluctuated substantially during an  irradiation,  but the inlet concentra-
tions were not measured often enough to  permit an accurate inflow  profile of
S0? to be generated.  Second, the oxidation reactions of S0? are quite slow
relative to the majority of other chemical transformations of interest in this
particular chemical system (e.g., the oxidation of NO and organics, and  the
formation of CL).  The net effect of these two characteristics of  the system
was that the fluctuations in the inlet tube S0? concentrations masked any loss
of S0  due to chemical reactions.
     The mechanism evaluation procedure, therefore, became more a test of the
adequacy with which the mixing and flow characteristics of the chamber were
modeled than of the accuracy of the mechanism.   In view of the substantial  un-
certainties in the inflow data, even very good  agreement between the predic-
tions and the data would not be sufficient to demonstrate the validity of the
mechanism.   Consequently, we suspended our efforts to test the;, SCL mechanism

-------
                                                                              52
until  more carefully controlled smog chamber data become available.   A new
experimental  study involving organics,  NO ,  and S09 is now in progress; we
                                         A        C-
summarize that program in  the following section.

4.   Future Examinations of S0? Chemistry

     Under the direction of Dr. Arthur  Levy, investigators at Battelle
Memorial  Laboratory are presently conducting a series  of organic-NO  -SCL-air
                                                                  X   L-
experiments using propylene (nine runs) and  toluene (six runs)  as the reac-
tive organic.   Under another EPA contract, we expect to employ  these data to
test the  SCL mechanism proposed in Section C-l.   The use of these data offers
several  advantages:

     >  The experiments are being conducted  under static conditions.
        Consequently, we will not have  to' contend with fluctuations
        in inflow reactant concentrations as an additional  variable
        in evaluating the  model.
     >  The chamber is still in operation (the chamber used for the
        experiments mentioned in Section C-l has  been  disassembled).
        Thus,  any chamber  effects that  were  not yet measured can still
        be determined, if  needed, for the model  testing exercises.
     >  Dr. Levy's group at Battelle has considerable  experience and
        expertise in studying SCL in smog chambers.  Therefore, the
        new SCL data will  almost certainly be the best that are
        currently obtainable.

     The  evaluation of a mechanism describing the oxidation of  SCL in solution
or in aerosols is more difficult.  To our knowledge, no systematic experimental
study of  this  process suitable for model testing has yet been carried out. Un-
til  the  oxidation rate of  SCL in systems containing aerosol particles has been
determined as  a function of particle size (volume, surface area), composition
and concentration of reactants in the particle, pH of  the particle,  and concen-
tration  of SCL in the gas  phase, it will be  difficult  to propose with any con-
fidence  a physical model for the oxidation of SCL in solution.   As a temporary

-------
                                                                              53
measure, it may be possible to develop a parametric model  in which the oxida-
tion of SCL in particles is described by the first-order reaction
                               so2
The parameter k  can then be estimated from the following:

     >  Observations of the S0? oxidation rate in droplets  under well-
        controlled conditions, such as those used in the experiments
        of Cheng et al. (1971).
     >  Knowledge of the composition of atmospheric aerosols.

Although a parametric mechanism is necessarily simplistic,  combined with the
gas phase mechanism, it may be sufficient to predict the atmospheric conver-
sion of SOp to sulfate within the uncertainty bounds of aerometric measure-
ments.  We expect to analyze the methods for selecting values  of k  during
Contract 68-02-0580.
D.   SPECIAL CONSIDERATIONS REGARDING THE TREATMENT OF TEMPERATURE,
     WATER, AND HYDROGEN PEROXIDE IN THE AIRSHED MODEL
     In the process of reviewing previous airshed modeling exercises, as well
as considering some of the possible difficulties that might arise in the use
of the latest version of the SAI model, we identified the following three ques-
tions that seemed to need further  clarification:

     >  To what extent should temperature effects on reaction rate
        constants be considered in the model?
     ?•  How important are the spatial  and temporal variations in
        water concentration?
     >  Will  the model predictions be sensitive to the initial con-
        centration distribution of hydrogen peroxide?

-------
                                                                              54
In an attempt to answer these  questions,  we  carried  out  various  sensitivity
studies using the kinetic  mechanism,  and  we  reviewed available measurements
for some of these parameters  in  one  of the most  severe and  persistent photo-
chemical air pollution  regions — the  South Coast  air  basin of  California.

     It is well  known  that reaction  rate  constants are a function  of temper-
ature.  This effect is  commonly  expressed using  the  Arrhenius relationship:

                               k(T)  = A

where

          k  =  the rate constant,
          A  =  a constant (sometimes referred to as the frequency factor),
          E  -  the so-called  activation  energy  for  the  reaction,
           a
          R  =  the gas constant,
          T  =  the absolute  temperature.

Given k at some  temperature T~ and  the activation energy, the value of k  at
any other temperature  can  be  estimated from
                            k(T)  =  k(Tn)  exp|-^-(f-f-)|    .           (8)
                                                       'o
Thus, we do not need  to  determine  b.   In  the  computer programs,  we  input
and the values  of E  and k(Tn)  for each  chemical  reaction.   Then k  can  be
                   a         u
calculated at any other  temperature T  using Eq.  (8).
     Although  the algorithm  outlined  above  is  not  difficult to  incorporate in
the model,  there  is  some  question  of  the  extent  to which  spatial  and temporal
variations  in  temperature must  be  considered.  For example, complete specifi-
cation of the  temperature as  a  function of  x,  y, z, and  time would require
significant amounts  of  additional  computer  storage, not  to  mention the extra

-------
                                                                              55
effort required of the user to assemble sufficient data to estimate the com-
plete temperature field.   Thus, we undertook a study to examine the sensitivity
of the kinetic mechanism to variations in temperature that might be found in an
urban airshed.  These results can be used as a guide for determining under what
conditions spatial and temporal features of the temperature field must be con-
sidered in the model.

     Similar questions arise concerning the distribution of water in the gas
phase over an urban area, especially a region like the South Coast air basin,
in which there are coastal  areas as well as inland valleys.  We note that
though spatial variations of relative humidity are significant in this airshed,
it is important to examine the variations in water concentration because this
is the parameter entering the kinetic rate expressions.  Thus, to determine
the extent to which provisions for treating spatial  and temporal variations in
water concentration should be included in the model, we examined the sensitiv-
ity of the mechanism to variations in water concentration.

     Finally, incorporation of the 31-step mechanism (excluding SCL chemistry)
in the model will require the user to specify initial and boundary concentra-
tions of HNCL and FLCL, two pollutants that are rarely measured routinely in
most urban areas.  To obtain a rough estimate of the concentrations of these
pollutants, we can assume that each is in chemical equilibrium; thus, from the
kinetic mechanism, we can write


             -k,n + {k2  + 8kq(2kJNO][NO?][H O]2 + k,?[OH][NO]H 1/2
          "] „   1 \J   \ I W     «/ \  O       L—   L~       1C-        I }
             k(-[HOj2
    [HOJ =
                16
If a simulation is to start somewhat before dawn, use of the above relationships
would be tantamount to assuming that chemical  equilibrium had been approached
during the preceding nighttime period.   Although this assumption may be reason-
able for HNCL, we note that the I^CL photolysis rate constant, k,g, would be

-------
                                                                              56
essentially zero at night.   In  fact,  from the  mechanism we see that there is
no "sink"  for FLO? other than  the photolysis  reaction.   Thus,  the use of the
equilibrium assumption  for  HLO,,,  especially at night,  does not seem desirable.
To examine this  issue further,  we carried out  simulation runs  using the mech-
anism to ascertain its  sensitivity to the initial  hL02  conditions.   In the
following  sections, we  discuss  the results obtained from these sensitivity
studies involving temperature,  water, and HLO  .

1.   The Predicted Effects  of  Changes in  Temperature
     and Water Concentration on Smog  Kinetics

     To determine what  effect  changes in  temperature or water  concentration
have on the concentration predictions, we carried  out  simulations of a smog
chamber experiment using the new  kinetic  scheme incorporated in the airshed
model.  The base values used were those of EPA Run 333:

     >  [N0]c =  1.25 ppm,
     >  [N02JQ = 0.08 ppm,
     >  [C3H6]0  = 0.23  ppm,
     >  [n-C4H10]0 = 3.41 ppm,
     >  [H20]0 = 16,000 ppm,
     >  T  - 25°C.

For each simulation run, we changed only  one  parameter  from the base values.

     We performed the simulations for two different temperatures, 15°C and
35°C, with all other factors kept the same.  We ca-lculated the rate constants
at the new temperatures from the  base values  of the rate constants (25°C) and
from measured or estimated  reaction activation energies, as shown in Table 9
(Garvin and Hampson, 1974;  Demerjian  et al.,  1974; Johnston et al., 1970).
Because the majority of the reactions in  the  mechanism  are thermal  and because
they have  small  positive activation energies,  raising  the temperature acceler-
ated the conversion of NO to N0?  and  decreased the time to the onset of 0~

-------
                                                                                     57
                                    Table 9



             ACTIVATION  ENERGIES OF  REACTIONS  IN THE GENERAL MECHANISM
EA
Reaction ___ kcal mole"
N02 + hv-i-^NO + 0 0
0 + 02 + M-^03 + M -1
03 + NO-WN02 + 02 2.4
0 + NO_-iL*-NO + 02 0.6
03 + N02— 5-^N03 + 02 4.9
N03 + NO-^~2N02 1.4
N03 + N02 + H20-z-^-2HN03 -1.9*
NO + N02 + 2H20-^-*-2HN02 + H20 0
2HN00-2-»-NO + N00 + H00 9
2 2 L
HN02 + hv-^2*-OH + NO 0
OH + N02-^-HN03 -2.2
OH + NO-^^-HNOp -2.2
OH +-CO + (Oj-^-VcO., + N09 0.15
c. c. c.
H02 + NO-ii^-OH + N02 2
HO + HO -lp-^H 0+0 0

Reference
Estimate
Garvin and Hampson (1974)
Garvin and Hampson (1974)
Garvin and Hampson (1974)
Garvin and Hampson (1974)
Johnston et al. (1974)
Davis (1974)
Demerjian et al . (1974)
Demerjian et al . (1974)
Estimate
Garvin and Hampson (1974)
Garvin and Hampson (1974)
Davis (1974)
Estimate
Estimate
                                                             Estimate
* The value of E.  listed  for  this  composite  reaction we determined from the

  values  of E^ for the three  equivalent reactions:




      	Reaction                     EA                        Reference
       NO,  +  N09	^N,0,-               -2                   Demerjian et al.  (1974)
         0      c.        (_  D


           N2Q5	*-N02 +  N03          19.4                 Garvin and Hampson  (19.74)



      N205  +  H20	^2HN03               0                   Estimate

-------
                               Table 9 (concluded)
                                                                                   58

HC1 + 0-ii.

HC, + 0- — -^H
1 3

19
HC1 + OH — i
HC2 + 0 ~^-H
HC2 + OH -iin
HC3 + hv -22-i
HC3 + OH -ilr
24
HC. + 0 _iL
4
HC4 + OH -iL:
ROD + N0_£l^
RCOO + NO + (02)-£2L
0
RCOO + N02-J£-
0
29
RO + 02 	 	
RO + N02-2!L
RO + NO -1L
Reaction
-ROO + aRCOO + (1-
II
ii
Q
^RCOO + RO + HC3
0

^-ROO + HC3
,-ROO + OH
^ROO + H20
^gROO + (2- )H02
i^BRCOO + (l-e)H00
8
t-ROO + OH

.ROD + H20
>-RO + N02
>-ROO + N02 + C02

^RCOON02
Q

^H09 + HC.
L. O
^RON02
^RONO
                                            kcal mole
                                                     -1
5 Estimated E.  for propylene
t Estimated E.  for n-butane
                                                0.1.
                                                3.8
                                                    §      I
                                                5t
                                                It
                                                0
                                                0
                                                 6
                                                 0
       Reference
Garvin and Hampson  (1974)

Garvin and Hampson  (1974)

Estimate
Estimate
Estimate
Estimate
Estimate
Estimate
Estimate

Estimate

Garvin and Hampson (1974)
Estimate
Estimate.

-------
                                                                              59
accumulation,  as  expected.   Conversely,  lowering  the  temperature  noticeably
slowed the smog formation  process.   Figure  1  presents  concentration-time
profiles for NO,  NCL,  CL,  and  propylene  for each  of these  two  runs.

     We carried out similar runs  at two  extreme  conditions of  relative
humidity--0 and 100 percent—at the base temperature  (25°C).   These  percent
ages correspond to 0 and 32,000 ppm of hLO, respectively.   Figure 2  shows  a
comparison of the predicted concentration-time profiles  for these two cases
with the profile for the base  case.  Increasing  the water  concentration
accelerated the conversion  of  NO  to N0?, whereas  a  complete elimination of
water dramatically slowed  down the  overall  smog  kinetics.   Both of these
effects are attributable to changes in the  production  rate and equilibrium
level of nitrous  acid, governed by  the reactions
                         NO
     Because it is virtually impossible—even  with  pumping  and  baking—to
obtain a water concentration of 0 ppm in  existing  smog  chambers,  we  carried
out one final  run at 3.2 ppm of I-LO.   The concentration-time  profile obtained
under these conditions  differed from  those of  the  completely  dry  run by  less
than 2 percent after six hours  of simulation time.

     In urban  areas, ambient temperatures and  water concentrations change
considerably during the day and from  one  day to  the next.   Thus,  the results
of these simulation runs suggest that it  may be  necessary to  account for vari-
ations in temperature and water concentration  when  modeling urban photochemical
smog.   Toward  this end, smog chamber  experiments conducted  at various constant
levels of temperature and water concentration  would be  most useful in ascer-
taining the effects of  variations of  these two parameters on  smog kinetics.

-------
                                          EPA  RUN 333 CONCENTRATIONS  (ppm)
E
Q.
CX
o
o
o
      0
                                              N0 = 1.25

                                              N02 = 0.08

                                                   = 0.23
                       	  35°C


                       	!5°C
       200


Time (minutes)
300
                    FIGURE 1.  CONCENTRATION-TIME PROFILES FOR NO,  N02,  03,

                                AND PROPYLENE AT 15°C AND 35°C
                                                                                                           en
                                                                                                           o

-------
E
QL
0.
C
o
c.
o
o
c.
o
o
                                      ': EPA RUN 333 CONCENTRATIONS (ppm)
                              100
                               200

                     Time (minutes)
300
           FIGURE 2,
PREDICTED CONCENTRATION-TIME PROFILES FOR NO, N02,  03,  AND  PROPYLENE

       AT 0,  50,  AND  100 PERCENT RELATIVE HUMIDITY

-------
                                                                              62
2.    Specification of the Initial  Concentration of H000
            — —  .	^—^

     With the implementation of the new kinetic scheme in the airshed model,
we must now specify the emission rate and the initial and boundary concentra-
tions of a new reactant., FLCL.   To ascertain the accuracy with which these pa-
rameters must be determined, we carried out kinetic simulations of EPA Chamber
Run 333 (under the initial  conditions listed in Section D-l)  at three differ-
ent initial hL02 concentrations:  0, 0.01,  and 0.1  ppm.

     The concentration-time profiles obtained for the case in which [HLO^L =
0.01 ppm did not differ appreciably from those for the base case,  in which
[O2J  = 0 ppm.  The small initial  HLO^ concentration resulted in a five-
minute reduction in the time to the NO- peak (305 versus 310  minutes) and a
small increase in 0, at 360 minutes  (0.32 versus 0.30 ppm).

     In constrast, the effect on the predictions of the presence of 0.1  ppm
of H?0  initially was far more  visible.  The conversion of NO to NO,-, was
accelerated considerably., and the NO- peaked at 264 minutes.   As a result of
the substantial  reduction in the time to the N0? peak, 0~ accumulated to  0.46
ppm before the simulation was terminated at 360 minutes.

     For similar simulations of another smog chamber experiment (EPA Run  349),
the initial conditions were as  follows:

     >  [NO]Q =0.31  ppm,
     >  [N02]Q =0.03 ppm,
     >  [propyleneL = 0.44 ppm,
     >  [n-butane]n = 3.25  ppm,
     >  [H20]Q = 16,000 ppm,
     >  T = 25°C.

In these simulations, a maximum in the 0, concentration did occur, and the
results indicate that the asymptotic ozone level is not affected appreciably
(less than 2 percent) by the initial presence of as much as 0.1 ppm of hLCL.

-------
                                                                              63
However, the H^CL did serve to reduce the time that elapsed before the maximum
was reached.  For example,  the predicted 0- maximum occurred at 194 minutes
for EPA Run 349 when the initial  charge contained 0.1  ppm of H?0?, compared
with 225 minutes when HLCL  was absent initially.

     On the basis of these  simulations, we feel  that an effort should be made
to construct an emissions inventory for H?0? only if the sources of such emis-
sions would lead to an ambient hydrogen peroxide  concentration of more than
0.01 ppm.  Should I-LO,-, sources contribute less than this amount, the error in-
curred by neglecting these  sources would be very  small, especially prior to
the formation of the N0? peak and at the 0^ asymptote.

     With regard to the specification of initial  and boundary concentrations
in the airshed model, the sensitivity runs indicate that care should be exer-
cised in specifying HpO? concentrations when they are  on the order of 0.1 ppm
or larger.   Data presented  by Bufalini et al.  (1972) suggest that FLOp in the
South Coast air basin may reach levels as high as 0.18 ppm during a very smoggy
day.  However, early morning and  late afternoon  levels were reported to be
about 0.01  to 0.02 ppm, thus indicating that overnight carry-over effects may
not be too significant.  We hasten to add that these observations are based on
a very limited number of ambient  air measurements.   Additional measurements of
the diurnal behavior of H^O,, in an urban airshed  would be useful.

3.   Spatial and Temporal Variations in Temperature and Water Concentration
     in the South Coast Air Basin

     Having shown in Section D-l  that the kinetic mechanism is somewhat sensi-
tive to changes in temperature and water concentration, we carried out a
limited effort to examine the extent of these  variations in an -actual airshed.
We chose the South Coast air basin for this study for  two reasons.  First,
photochemical  smog is particularly severe in this region.  Second, we expected
that the spatial and temporal  variations in temperature and water concentration
found here  would be as large as those found in most other airsheds where the
model  might be applied.

-------
                                                                              64
     During  the  summer,  an  onshore  flow of moist marine air generally keeps
coastal  areas  relatively cool  [temperatures in  the 70s  to 80s  (°F)].   By the
time the air has traveled to  the  inland valleys, however, significant heating
has taken place, and  the temperature often exceeds 100°F.  In  addition,  rela-
tive humidities  near  the coast are  usually higher than  those measured inland.
Of course, since water concentration is the parameter of interest,  the effect
of temperature on relative  humidity must be considered.

     Tables  10 and 11  present hourly ground-level  temperature  and relative
humidity data for three  smoggy days in  June 1974.   The  station location  asso-
ciated with  each code number  is as  follows:

                         Number             Station Name
                          13W            Lennox
                          21W            Long Beach
                          41W            Burbank
                          61W            Ontario
                          75W            Downtown Los  Angeles

Figure 3 shows the location of each  station.  Lennox  and Long Beach are repre-
sentative of coastal  Iocations5 whereas  Downtown Los  Angeles, Burbank, and
Ontario are representative of  inland communities.

     To examine variations in  temperature and relative humidity with height
above the terrain, we reviewed some  of the measurements recently reported  by
Blunienthal  et al.  (1974).   They measured the three-dimensional  distribution of
pollutants  and meteorological  parameters throughout the South Coast air basin
using a fully instrumented fixed-wing aircraft.  We chose to  examine a two-day
period--26-27 July 1973--for which numerous aircraft  spirals  were made, both
during the  day and at night.

-------
                                Table 10
GROUND-LEVEL AIR TEMPERATURES IN THE LOS ANGELES  BASIN  ON  28-30 JUNE 1974
• JOS NUMBER =GAMTA8'.S AIR POMUTION CONTROL
PRQCRAM =GAKTABLS
PATF: 07/19/74
• STA
0 1 2 3 4 5 ' A
13W 64
2 in 65
4 1 vi . 60
A 1 W 67
75W _ 67
STA
•0123456
13W 63
21V! .64
4 1 w 65
61W 65
7SW ' 67
.STA
0123456
. 1 3 W • -62
21W 63
41W • 62
61W 57
75W 66
TFMPFR.ATMHF

7
70
70
74
75
70
7
65
65
CO
69
69

8
75
75
79
03
74
H
69
68
74
77
72

9
77
91
85
88
79
9
71
•70
79-
81
74
DISTRICT - COUNTY OF
/ AT HHIIIX / IN

10
76
84
9?
"3
06
10
69
72
04
HA
76
HM,
11
77

97
9fl
93
„„,
12
78
90
99
1 
-------
                                                     Table 11
                  GROUND-LEVEL RELATIVE HUMIDITIES IN THE LOS ANGELES  BASIN ON 28-30 JUNE 1974
JOS
                                       AIR  POLlWri'OH CDKTROL DISTRICT - CQOWr Qf LOS
PROGRAM =GAMTA8LS
DATE: 07/19/74
STA
13V
21W
41 W
61W
75W
STA
13W
21W
41w
61W
75W
STA

13W
?.1W
A 1 w
61W
75W

0123456
87
65
37
38
65
0123456
90
75
43
47
H?

7
76
59
40
30
61
7
90
73
46
47
HO
RFI ATIVE


fl
60
50
31
27
56
H
78
65
43
36
73


9
58
36
28
23
51
9
73
63
' 3H
35
69
HUMIDITY /AT HOUR


10
5H
43
17
20
39
10
76
59
36
2.5
65


11
5?
12
13
35
i—
in-; P<;T
12 13
AH 44
32 31
12 19
io._ 10 _
32 48
Hliim PST
11 1 ? 13
6H
57
40
17
54
66 66
55 52
4? - 40
16 21
60 57'

14
44
31
19
10
46
14
68
50
30
54
f TN

15
44
28
13
..!)..
43
\ 5
6(1
52
31
25
56
PERCENT

16
A?
26
14
_n
40
16
73
52
42
28
60

17
48
79
14
13
39
17
81
57
38
34
65

18
50
31
19
38
1s
84
55
46
36
68

19 20 21
63
40
17
.38
19 ?0 21
87
65 .
56
50
80

22 23 AVE
55.3
38.5
20.9
1 n . o
45.1
22 23 AVF
76.3
59.3
40. fl
3?.l
6S.9
HOUR PST
0123456
90
73
ai
100
83
7
90
73
78
100
82
0
87
70
70
93 •
-31
9'
78
70
63
76
76
10
70
61
60
54
70
11
70
57
58
44
62
12 13
64 66
57 59
52 51
36 32
62 63
14
66
57
54
32
62
15
66
57
51
34
6B
16
6H
63
51
4)
67
17
76
65
53
46
70
18
81
65
61
. 54
77
19 20 21
84
68
68
66
80
22 23 AVfi
75.4
63.9
6O. fl
57.7
71,6
VA 76,
A/28/74
N H I M
14 47
13 26
14 12
12 . 10
14 32
6/29/74*
M HIM
14 66
1 4 50
14 30
13 16
14 54
6/30/74*
N M I H
14 ft 4
14 57
14 51
14 32
14 62
                                                                                                                                     CTl

-------
                                     SAN GABRIEL MOUNTAINS
                        41W ©
                           BURBANK,
                                75W ®
                                   LOS ANGELES
                                  DOWNTOWN
SANTA MONICA MOUNTAINS
                                                                         61W ©
                                                                        ONTARIO
                                        21W®
                                           LONG BEACH
          FIGURE 3.  LOCATIONS OF TEMPERATURE AND RELATIVE HUMIDITY
                             MONITORING SITES

-------
     From the data presented  in  Table  10,  we  note  that  the  maximum  differ-
ence in temperature at any hour  during the day  and the  variation  in average
air temperatures  across  the basin  for  each of the  three days  are  as follows:
                                                                              68
          Day
        June 28
        June 29
        June 30
                           Maximum Spatial
                        Temperature Difference
°c
15
14
10
°F
27
25
18
    Average Spatial
Temperature Difference
°c
10
9
5
°F
18.5
16.5
8.5
Thus, spatial  variations in temperature of as  much as  15°C may exist in  the
Los Angeles area during the middle of the day.   However,  on the average,  the
variations in temperature are somewhat smaller.

     To show temperature variations aloft, we  plotted  in  Figure 4 temperature
profiles above Rialto, California, at five times on 26-27 July 1973.  The 13:07
sounding on July 26 exhibits a temperature difference  of  about 9°C.   If  adia-
batic conditions had persisted, we would have  expected the temperature gradient
                i
to be -0,01°C tn.   Thus, over a 1000m interval,  the temperature difference
would be 10°C, which is approximately the amount observed at Rialto  at 13:07.
As illustrated below, when an elevated inversion layer is present, the tempera-
ture differences in this situation may be smaller than those that would  exist
under adiabatic conditions:
                     O)
                     re
                           Temperature
                                           Inversion  layer

-------
                                                                            69
5000 -
4000 -
3000
2000
1000
                                                           9:39 j—
                                                          13:07 j—
                                                          17:20 (-_.
                                                          23:17
                               ^Temperature-- C
           FIGURE 4.  DISTRIBUTION OF THE TEMPERATURE ALOFT
                   ABOVE RIALTO ON 26-27 JULY 1973

-------
                                                                              70
For a modeling region extending to,  say,  1000m in height above the terrain,
vertical  temperature differences may be as  large as  horizontal  variations.

     In considering the distribution of water in the basin,  we must first
convert relative humidity measurements  to water concentration in  ppm.   Using
the definition of relative humidity, we can calculate the concentration of
water, [HLO], in ppm from the following formula:
                                           x 10

where

          RH  =  relative humidity (in percent),
           P  =  vapor pressure of water (in mm Hg)  at temperature  T.

     Figure 5 illustrates the temporal  variation  of  water concentration  at the
five ground stations on 28 June 1974.   The two coastal  locations  tend  to exhibit
similar behavior, as do the two inland locations.  Concentrations at the Down-
town Los Angeles site seem to be more  characteristic of those found near the
coast than those observed farther inland.   In general, the spatial  variation
in water concentration is about 7000 to 11,000 ppm.

     Examining the temperature and humidity profiles observed at  Rial to  on 26-27
July 1973, we calculated vertical  profiles of water  concentration for  five times
during this two-day period.   These profiles are illustrated in Figure  6.  The
maximum variation in concentration measured on these days was about 8000 ppm,
as shown in the 17:20 profile for July 26.

     In the analyses described above,  we found that  spatial  variations in tem-
perature and water concentration in the Los Angeles  basin can be  as large as
15°C and 11,000 ppm, respectively.   Of course, since only a very  limited number
of days were examined, it is highly probable that  even greater variations fre-
quently occur.  Considering the sensitivity results  presented in  Section D-l
and the variations in temperature and  water concentration cited above, it is

-------
0600
1000
                             1200
1400
                              1600
                                                           1800
                                Time—hour
                              (28 June 1974)
          800
 1000
                             1200
1400
1600
1800
                               Time—hour
                              (30 June 1974)
  FIGURE  5.   TEMPORAL  VARIATIONS  IN WATER  CONCENTRATION
        AT FIVE LOCATIONS IN THE  LOS ANGELES BASIN

-------
                                                                            72
5000 r
              5,000
10,000      15,000      20,000

  Water Concentration--ppm
25,000
       FIGURE  6.   DISTRIBUTION  OF THE WATER  CONCENTRATION ALOFT
                   ABOVE  RIALTO ON  26-27 JULY  1973

-------
                                                                              73
 difficult to conclude that these variations can be completely ignored.  There-
 fore, we recommend that future studies be carried out using the airshed model
 itself to test various alternative strategies for treating temperature and
 water.  Such strategies might include treating temperature or water concentra-
 tions as functions of

      >  Time only
      >  z and time
      >  x, y, and time
      >  x, y, z, and time.

 Toward this end we have included provisions in the computer codes to allow the
 user to input temperature and relative humidity fields that vary in both space
 and time.

 E.   TREATMENT OF ORGANICS IN THE AIRSHED MODEL
                                       V
     Use of the kinetic mechanism discussed in Section B-l requires that the
 organic species be grouped into four classes:  paraffins, olefins, aromatics,
 and aldehydes.  To treat a mixture of numerous organics, such as those found
 in the atmosphere, "average" rate constants must be estimated for 0, OH, and
 0~ attack, as appropriate, for each of the four organic groups.  In general,
 specification of a single set of average rate constants that are invariant in
 space and time is possible only if the individual  members of each particular
 group are of similar reactivity (neglecting spatial and temporal temperature
 effects).  Table 12 presents rate constants for 0, OH, and 0, attack on vari-
 ous hydrocarbons.  Because of the abundance of methane in the atmosphere and
 the .wide disparity in reactivities of various paraffins, we conducted a study
 to ascertain the best treatment of this hydrocarbon group in the airshed model.

     We considered four strategies for grouping paraffins:

     (1)   One reactive group including all paraffins.
     (2)   Two reactive groups—C,  through C~ low reactive; C.,
          C,-, ...  high reactive.
     (3)   Two groups--C,  through  C,, nonreact'ive; C., Cg, ... reactive.
     (4)   Two groups—methane nonreactive; C?, C^, ... reactive.

Strategy  3 has been employed in previous  applications  of the airshed model.

-------
                                                         Table 12
                            RATE CONSTANTS FOR 0, OH. AND 03 ATTACK ON VARIOUS HYDROCARBONS

                                                                                OH
    Hyrdrocarbon
•Paraffin's
   Methane
   Ethane-
.  Propane
   Butane
   Isobutane
   n-pentane
   Isopentane
   2,2-dimethylbutane
   Cyclopentane
   2,3-dirnethylbutane
  •2-methylpentane
   3-methylpentane     •' .
   n-hexane
   Methylcyclopentane
   2,4-dimethylpentane
   2-methylhexane
   3-methylhexane
 •  2,2,4-trimethylpentane
'  'n-heptane
   Methylcyclohexane
   2,4-dimethylhexane.
Rate Constant
1.8 x 10"2
1.37
1.23 x 10
3.2 x 10
8.8
8.5 x 10
1.9 x 102
3.0 x 102
2.9 x 102
1.5 x 102
2.2 x 102
2.2 x 102
1.36 x 102
1.3 x 102
3.3 x 102
2.5 x 102
2.5 x 102
2.5 x 102
1.91 x 102
1.6 x 102
3.7 x 102
Reference
Herron and Hufe (1969
Herron and Huie (1969)
Heicklen (1967)
Herron and Huie (1969)
Wright (1965)
Herron and Huie (1969)
Herron and Huie (1969).
Herron and Huie (1969)
Herron and Huie (1969)
Heicklen (1967)
Estimate
Estimate
Herron and Huie (1969)
Estimate
Estimate
Estimate
Estimate
Herron and Huie (.1969)
Herron and Huie (1969)
Estimate
Estimate
Rate Constant
1.6 x 10
4.5.x 102
1.8 x 103
5.72 x 103 ••
5.12 x 103. ••
5.81 x 103
6.76 x-103 -;.
2.80 x 103 /
1.11 x 104
8.2 x 103
8.41 x 103
8.41 x 103
7.16 x 103
6.87 x 103
1.13 x 104
1.06 x 104
1.06 x 104
7.34 x 103
8.81 x 103
8.5 x 103
1.30 x 1 04
• Reference
.Greiner (1967)
Greiner (1967)
.Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner.'(1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiher (1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Greiner '(1967)
Greiner (1967)
Greiner (1967)
Greiner (1967)
Rate Constant
Reference

-------
                             Table 12  (Concluded)
                                                       OH
Hydrocarbon
2 , 5-d jmethyl hexane
2 ,3 ,4-trimethyl pentane
n-octane
n-nonane
n-decane
Rate Constant • -Reference
3
1
2
2
2
.7
.8
.5
.0
.6
X
X
X
X
X
102
102
102
102 .
102
Estimate
Herron and Huie
Herron and Huie
Estimate.
Estimate

(1969)
(1969)


Rate Constant
1.30
1.58
1.28
1.21
1.38
X
X
X
X
X
104
104
104
104
104
' Reference Rate Constant • Reference
Greiner
Greiner
Greiner
Greiner
Greiner
(1967) -.. " - •
(1967)
(1967)
(1967)
(1967)
Olefins
  Ethylene
  Propylene
  Butenes
  1-pentene
  Trans-2-pentene
  Cis-2-pentene
  2-methyl-2"butene
  Cyclopentene
  1-hexene
  Cis-2-hexene
  1-heptene

Aldehydes
  Formaldehyde
  Acetaldehyde
  Propionaldehyde
7.72 x 10
4.41 x 103
4.41 x 103
1.69 x 10
6.03 x 104
2.35 x 104
5.00 x 103
4.41 x 10
4.41 x 102
Cvetanovic (1963)
Cvetanovic (1963)
Cvetano'v.ic (1963)
Cvetanovic (1963)
Cvetanovic (1963)
Cvetanovic (1963)
Cvetanovic (1963)
Estimate
Estimate.
2.13 x 10°
2.13 x 104
5.12 x 104
5.33 x 104
1.13 x 106
1.13 x 106
1.49 x 105
1.92 x 104
1.92 x 104
3.84 x TO4
Morris and
Morris and
Morris and
Morris and
Morris and
Morris- and
Morris and
Mikl (1971)
Niki (1971)
Niki (1971)
Niki; (1971)
Niki (1971),
Niki (1971)
Niki (1971)
Morris and Niki (1971)
Morris and Niki (1971)
Morris and Niki (1971)
3.8 x 10
1.6 x 10'
1.3 x 10'
1.3 x 10
S'.O x 10
4.1 x 10
-3
-2
-2
-2
-2
-2
                                           1.5 x 10
                                           4.1 x 10
                                           1.21 x 10
                                  -2
                                  -2
                                                                                              -2
Wei (1963)
Wei (1963)
Wei (1963)
Wei (1963)
Wei (1963)
Wei (1963)
                                  Wei (1963)
                                  Wei (1963)
                                  Cadle  (1952)
Aroma tics
Toluene ' 1.1 X 102
M-xylene
p-xylene
0
4.4 x 10^

Estimate

Estimate
                                                          Estimate

-------
                                                                             76
     To test these schemes,  we  performed  smog  chamber  simulations  for  a  mixture
of paraffins, NO , and  CO  having  proportions typical of  those  found  in the  Los
                X
Angeles atmosphere in 1969.   For  comparison, we  carried  out  a  baseline simula-
tion in which each paraffin  was treated as  an  individual  reactive  species in
the mechanism.   Thus, we compared the  predictions  for  Strategies 1 through  4
with those for the baseline  case  to  determine  the  errors  introduced  by each
lumping scheme.

     Initial conditions for  the simulation  runs  were derived from  air  quality
measurements taken at Commerce, California, on 30  September  1969 by  Scott
Research Laboratories.   In particular, we used the following concentrations,
which were measured at  8 a.m.  on  that  day:
                         Species
                         CO
                         NO
                         crcs
                         V
Concentration
    (ppm)
    10.0
     0,4
     0.1
  1.6 x 10
     4.213
                                                  4
     0.476
     3.944
     0.269
The predicted values of NO,  NCL,  and  0~  after  12  hours  of irradiation  were  as
follows:
                                    Predicted  Concentration
                    Strategy
                    Baseline
                        1
                        2
                        3
                        4
NO
0.05
0.04
0.05
0.09
0.05
N02
0.30
0.30
0.30
0.28
0.30
°3
0.05
0.07
0.05
0.03
0.05

-------
                                                                              77
These results indicate that Strategies 2 and 4 led to the best agreement
with the baseline case.   Since Strategy 2 uses two reactive species, whereas
Strategy 4 involves only one, we plan to treat the paraffin class  according
to Strategy 4 to minimize computing costs.

     Thus, five organic  classes are considered in the airshed model:  non-
reactive hydrocarbons (methane and acetylene), nonmethane paraffins, olefins,
aromatics, and aldehydes.  We recommend that future studies be carried out to
ascertain whether the olefins should be treated as a single lumped species or
as several lumped species.  In addition, it may be possible to combine the
aromatics with the nonmethane paraffins, since both groups have similar reac-
tivities and may produce similar products (according to the mechanism given
in Section B-l).
F.   INTRODUCTION OF THE IMPROVED KINETIC MECHANISM
     INTO THE AIRSHED MODEL
     In Sections B and C, we delineate efforts  aimed  at developing improved
mechanisms for describing the chemical interactions  of hydrocarbons,  NO  ,  0-,,
                                                                       X   3
and S0?.  With regard to the HC-NO -0., system,  the generalized mechanism dis-
      C.                           X  O
cussed in Section B represents a significant improvement over the  15-step
mechanism previously employed in the airshed model.   Thus,  we have incorpo-
rated the expanded mechanism into the model.  In addition,  we have implemented
in the model  the S0? mechanism described in Section  C, even though the  mechan-
ism has yet to be validated using smog chamber  data.   In the present  section,
we discuss our efforts to use the improved kinetic mechanism in an actual  air-
shed simulation.
     Installation of the new mechanism in  the airshed model  required that
numerous changes be made in the computer codes.   Particular  difficulties  arose
because the number of species that must be followed in the airshed model  in-
creased from 6 to 12 (NO,  N0?,  CL, H?02, HN02> nonmethane paraffins, olefins,
aromatics,  aldehydes, S0?, CO and  unreactive  hydrocarbons).   Moreover,  the
programs were to be exercised on the  CDC 7600 computer,  which has  only  a  lim-
ited amount of small core  memory,  at  Lawrence Berkeley Laboratory.  Thus, we

-------
                                                                              78
restructured the programs somewhat to make efficient use of available small
core memory, as well  as the more abundant amounts of extended core memory.
After the coding changes were made, we checked the programs by running sev-
eral test cases.

     To gain some experience in using the new mechanism in airshed simula-
tions, we decided to  exercise the model  using meteorological  and emissions
inputs derived in previous model evaluation efforts.  We felt that using the
same meteorological  and emissions inputs, to the extent possible would pro-
vide a means for ascertaining how sensitive the model  predictions were to the
change in the kinetic mechanism itself.   Because of our previous experience
in simulating the Los Angeles basin on 29 September 1969, we  chose that day
for our initial model application effort.

     Before the simulations could be carried out, we first had to compute new
splits for hydrocarbon emissions and initial and boundary concentrations.
Previously, available hydrocarbon emissions and air quality data were divided
into two groups — reactive and unreactive hydrocarbons.   To use the new mech-
anism, we revised the categories to reflect the new definition of the five
organic classes--nonmethane paraffins, olefins, aromatics, aldehydes, and
nonreactive hydrocarbons (methane and acetylene).

     Organics are emitted from a variety of sources in  the Los Angeles basin,
including motor vehicles, refineries, and numerous other stationary sources.
Although the organic  composition of automobile emissions has  been documented
by several  investigators, very little information is available for use in es-
tablishing guidelines for estimating the compose!tion of the stationary source
emissions.   For the  purposes of this study, we assumed  that the composition of
stationary source emissions is the same  as that for automobiles.  Although  we
recognize that this  is not necessarily a good assumption, our main objective
was simply to make "reasonable" estimates of the emission splits to exercise
the model.   A more refined inventory can be derived using the results of a
recent study of organic emission control strategies carried out by Trijonis
and Arledge (1975).   Unfortunately, their results were  not available in time
for inclusion in this study.

-------
                                                                              79
     Using organic composition data derived from tests of 10 automobiles
reported by the Bureau of Mines (1973), we estimated the following mass
emission splits:
                                                  Mass Split
                     ^	Group	         (percent)
                   Nonmethane paraffins               29
                   Olefins                            30
                   Aromatics                  •        23
                            *
                   Aldehydes                           5
                   Nonreactive hydrocarbons           18
Thus, we added previous  estimates  of reactive and nonreactive hydrocarbon-
emissions to estimate the spatial  and temporal  distribution of total  hydro-
carbon emissions.   Then, we multiplied the emission splits  cited above by
the total hydrocarbon emissions in each grid cell to estimate the distri-
bution of emissions for  each of the five classes.

     We calculated initial and boundary concentrations using our previous
estimates of reactive and nonreactive hydrocarbon concentrations in conjunc-
tion with gas chromatographic analyses of ambient air in the basin for
29 September 1969 reported by Scott Research Laboratories (1970).  We derived
the following relationships:

        [Olefins]  -  0.211[CR],
        [Paraffins]  =  0.414[CR]  + 0.057[CNRJ,
        [Aromatics]  =  0.376[CR]  + 0.003[CNR],
        [Aldehydes]  =  0.04 ppm,
        [Nonreactive hydrocarbons]  =  0.94[CNRJ,

where [CnJ and [C.m] are the original estimates of reactive and nonreactive
        K        INK
hydrocarbon concentrations, respectively.
 Aldehyde emissions are estimated to be about 5 percent of the total hydrocarbon
 emissions.   Since aldehydes were not included in the original SAI inventory for
 Los Angeles, the total percentage adds up to 105 percent.  Thus., we increased
 the total  organic emissions by 5 percent to reflect the additional  aldehyde
 emissions.

-------
                                                                              80
     Figures 7 through 12 illustrate some of the predictions obtained from
the SAI  model  using the 31-step kinetic mechanism and the emissions and air
quality  inputs described above.  These figures also show the predictions
from the analogous simulation in which the 15-step mechanism was employed.
In general, the most obvious  characteristic of these results is that the 0~
production seems to have been accelerated, leading to higher predicted 0,
levels.   However, in many instances the NCL predictions are in better agree-
ment with the  measurements,  especially during the late morning and early
afternoon.

     It  is difficult to make  any assessment now of the enhanced reliability
of the model resulting from  the incorporation of the new 31-step mechanism.
However, considering the nature of the available model inputs used in this
study, the results are encouraging.  We recommend that a greater effort be
expended in future work to assemble an appropriate organic emissions inven-
tory. Furthermore, the enhanced production of CL observed in the results
presented here may be caused  in part by inaccuracies in the treatment of
aldehyde photolysis or NO removal  in the mechanism.   We assumed that alde-
hyde photolysis is proportional to that for NCL.  However, shifts in the UV
spectrum throughout the day may invalidate this assumption.   Finally, NO may
be removed too rapidly in the mechanism, thus, allowing OQ levels to build
up prematurely.  These issues can  be resolved only by subjecting the model
to a comprehensive evaluation.   We recommend that such an undertaking be
considered in  the near future.

-------
                                                                             81
D_
Q_
•w
CJ
O
o
    VJO
    30
    20
    10
      LA HABRA
       9/29/69
   (U  MEASURED   NO
 	PREDICTED  NO  (15)
  A  MEASURED   N02
	PREDICTED  N02 (15)
  ©  PREDICTED  NO  (31)
   A  PREDICTED  N0'2 (31)
                                   9     10     11
                                 TIME  CPST)
     12
                                                            13
1U
    60
    50
•>-
cu
Q-
o
•z.
o
o
    30
    20
    10 -
        -6-
              J2_
.  <:  LA HABRA
       9/29/69
_  D  MEASURED   0,
	  PREDICTED  03  (15)
  ©  PREDICTED  0,  (31)
                                   9     10    11
                                 TIME  CPST)
     12
                                                            13
    FIGURE  7.  PREDICTED AND MEASURED CONCENTRATIONS FOR LA HABRA
           USING THE 15- AND 31-STEP KINETIC MECHANISMS

-------
                                                                              82
    60
    50
CL-
OU
CJ
    30
    20
   ANAHEIM
   '  9/29/69
[D  MEASURED
                                       NO
                         - PREDICTED  NO  (15)
                         A  MEASURED   N02
                         • — PREDICTED  N02 (15)
                         ©  PREDICTED  NO  (31)
                          A  PREDICTED  N0'2 (31)
5      6
   60
   50
                                   9     10     II
                                 TIME  CPST)
                           12
       13
                           ANAHEIM
                              9/29/69
                        IH MEASURED   0-,
                        	 PREDICTED  03  (15)
                        © PREDICTED  03  (31)
1C
O_
Q-
O
   110
   30
   20
   10
                                                       I	I	t
8      9      10.     11     12     13
     TIME CPST)
                                                                    1U
        FIGURE  8.  PREDICTED AND  MEASURED CONCENTRATIONS  FOR ANAHEIM
               USING THE 15- AND  31-STEP KINETIC MECHANISMS

-------
                                                                               83
                                                       POMONA
                                                        ' 9/29/69
                                                    D MEASURED
               NO
                                                    	PREDICTED   NO   (15)
                                                    A  MEASURED    M02
                                                    — PREDICTED   N02  (15)
                                                    ©  PREDICTED   NO   (31)
                                                    A PREDICTED   N0'2  (31)
31
0-
0_
o
z
o
o
   10 -
                                  9      10     11
                                 TIME (PST)
         13
   POMONA
      9/29/69
Q MEASURED    0,
   PREDICTED   0~
   PREDICTED   0.,
                                                                    (15)
                                                                    (31)
                                  9     10     11
                                TIME  (PST)
         FIGURE 9.   PREDICTED AND MEASURED CONCENTRATIONS FOR-POMONA
                 USING THE 15- AND 31-STEP KINETIC MECHANISMS

-------
                                                                          84
    50
    UO
'£  30
 o
    20
 -   io
                                                       PASADENA
                                                         9/29/69
                                                       MEASURED
               NO
 	PREDICTED  NO  (15)
  A  MEASURED   XQ
 	PREDICTED  N02 (15)
  ©  PREDICTED  NO  (31)
  A  PREDICTED  N02 (31)
                                   9     10    11
                                 TIME  CPST)
   12
13
    60
    50
 o_
 Q_
 CJ
 ~z.
 ED
 CJ
    30
    20
    10 -
    PASADENA
       9/29/69
 tD  MEASURED   0,
	  PREDICTED  03  (15)
 ©  PREDICTED  0,  (31)
                                   9      10     11
                                 TIME CPST)
   12
13
      FIGURE 10.   PREDICTED AND MEASURED CONCENTRATIONS  FOR  PASADENA
               USING THE 15- AND 31-STEP KINETIC MECHANISMS

-------
                                                                            85
                                                       DOWNTOWN LA
                                                         9/29/69
                                                    13  MEASURED   NO
                                                  	PREDICTED  NO  (15)
                                                    A  MEASURED   N02
                                                 	PREDICTED  N02 (15)
                                                    ®  PREDICTED  NO  (31)
                                                    A PREDICTED  N0'2 (31)
   15
                                  9     10     11
                                TIME  (PST)  .
12
                                13
                          DOWNTOWN LA
                             9/29/69
                       B  MEASURED   0,
                      —  PREDICTED  0,
                                                     PREDICTED  0.
               (15)
               (31)
                                        ED
n:
o_
   10
8     9      10     11
     TIME (PST)
                                                            13
             1U
             FIGURE IT.   PREDICTED AND MEASURED  CONCENTRATIONS FOR
          DOWNTOWN LOS ANGELES USING THE'15- AND 31-STEP MECHANISMS

-------
                                                                           86
    50
    •10
IE  30
o
    20
    iO
                        WEST LA


                         9/29/69

                    CI MEASURED


                    	PREDICTED
                                                     A MEASURED
                NO


                NO


                NO,
                                                                       (15)
                                  N02 (15)
	PREDICTED


   ©  PREDICTED  NO' (31)


   A  PREDICTED
                                                                   N0'2 (31)
         5.     6      7
   20
2:
x:
o_
a.
u
2:
o
o
    10
  9     30    II

TIME (PST)
                                        ©
          13
                       WEST LA


                         9/29/69

                    13  MEASURED   0,


                   —  PREDICTED  00
                                                                      (15)
                                                        PREDICTED  0   (31)
                                   9      10    11

                                 TIME CPST)
                     12
                1U
           FIGURE 12.  PREDICTED  AND MEASURED CONCENTRATIONS FOR

           WEST LOS ANGELES USING THE 15- AND 31-STEP MECHANISMS

-------
                                                                        87
         III    METEOROLOGY-RELATED DEVELOPMENT ACTIVITIES
                              Steven  D. Reynolds
                                Mark  A. Yocke
                                  Jody Ames
     In our previous model  development and  application  efforts, we made
several assumptions about the treatment of  meteorological  parameters.  Among
these, the most notable are the following:

     >  Wind shear effects  can be neglected.
     >  A diffusivity algorithm that is solely  a  function  of wind
        speed and height can be used.
     >  The base of an elevated inversion layer is  a  suitable  choice
        for the top of the-modeling region.

However, these assumptions  clearly introduce  inaccuracies.  First, the wind
flow field is fully three-dimensional  and should  be treated accordingly.
Second, the magnitude of the turbulent diffusivity  depends on  atmospheric
stability and surface roughness,  as well as on  wind speed  and  height.  Third,
significant quantities of pollutants trapped  in an  elevated inversion layer
may be injected into the mixed layer as the stable  layer is eroded by surface
heating effects.  Moreover, ground-based inversions frequently occur at night.
Thus, further consideration needs to be given to  the  definition of the model-
ing region and the treatment of inversion layers  in the model.  In the follow-
ing sections, we discuss our efforts to improve the treatment  of wind fields,
diffusivities, and inversions in  the airshed  model.

A.   MODEL SENSITIVITY TO THE INCLUSION OF  WIND SHEAR

     The results of model sensitivity  studies reported  in  Volume I indicate
the importance of accurately specifying the wind  speed  and direction through-
out the region of interest.   In this section, we  discuss additional sensitiv-
ity studies that were carried out to assess the importance of  characterizing
wind shear effects.   The results  of this effort will  be useful for establish-
ing (1) the need to extend  our existing meteorological  algorithms to treat

-------
                                                                        88
wind shear and (2) the extent to which vertical  wind soundings should be
taken over urban areas.

     Accurate specification of winds aloft is usually-hampered in a grid
model by a dearth of appropriate measurements.   Since the full three-
dimensional structure of the wind field must be  input to the model, par-
ticular attention must be given to this aspect  of model  usage.  The fol-
lowing are four possible means for establishing  the complete wind field:

     (1)  Assumption of a "flat" velocity profile, where the estimated
          ground-level wind speeds and directions are assumed to be
          invariant with height (i.e., wind speeds and  directions are
          a function only of x, y, and t).
     (2)  Calculation of the winds aloft by scaling the  ground-level
          winds according to the findings of previous wind shear
          studies (i.e., assumption of a form for the wind shear,
          such as a power law profile).
     (3)  Interpolation for the wind speeds and  directions using
          actual wind soundings aloft.
     (4)  Prediction of the wind flow field using a numerical  simu-
          lation model.

The first alternative, which is the simplest, is useful  for establishing the
basic characteristics of the flow field.  Previous SAI  simulations  have used
this approach.  For more refined estimates  of the wind  field when no  measure-
ments aloft are available, the second technique  can be  used.   The last two
alternatives afford the best means of specifying winds  aloft,  provided that--
for Alternative 3--the measurement network  is sufficiently dense and  that—for
Alternative 4--the model  has been validated.  At present, Alternatives 2 and 3
appear to represent the best means for accurately specifying winds  aloft.

     An important step in procuring a data  base  for describing the  upper level
winds is being made in the RAPS program for St.  Louis.   One aspect  of this com-
prehensive data gathering study will  be the regular monitoring of winds aloft

-------
                                                                        89
at two to four sites  in  this  metropolitan  area.   Using the  surface wind data
in conjunction with  the  upper wind  measurements,  one should be  able to esti-
mate with reasonable  accuracy the  structure  of the  wind field over this urban
area.

     To gain some insight into the  importance of wind shear effects on the
predictions obtained  from the photochemical  airshed model,  we carried  out a
series of comparative simulations  for the  Los Angeles basin,  using both "flat"
and power law wind velocity profiles.   Since vertical wind  soundings were not
available for Los Angeles,  we used  only the  surface-based measurements to gen-
erate both wind fields.   In the following  subsections, we further  describe the
treatment of wind shear  and discuss the results  of  the simulations.

1.   Hind Velocity Profile

     Variations in horizontal wind  with height have been the  subject of inten-
sive study in meteorology for years.   Assuming neutral stability  conditions,
von Karman derived a logarithmic relationship for the mean  wind velocity in
the sublayer (surface layer)  of the atmospheric boundary layer  from theoreti-
cal considerations (Plate,  1971):


                                *' "    \ O/
where

        U  =  wind speed at height  I,
        u* =  the friction velocity,
        Zfi =  the roughness parameter,
        K  =  the von Karman  constant.

Subsequently, this relationship was verified through experiment.   For  diabatic
conditions, Laikhtman (1944)  and Deacon (1949) proposed that the  expression

-------
                                                                        90
                                - g)
be used, where 3 is a function of atmospheric stability.
                                                                         (10)
     Within the remainder of the atmospheric boundary layer (i.e., above the
sublayer), wind profiles are usually characterized by an empirical power law.
Blasius was the first to describe the mean velocity distribution by the fol-
lowing general' relationship:
where LL is the wind velocity vector at a reference height ZR.   The exponent
M is a function of ground surface roughness and atmospheric stability.
DeMarrais (1959), Davenport (1965), Shellard (1965), and Jones  et al.  (1971)
performed experiments to derive quantitative relationships for  M.  On  the
basis of their findings, they estimated that M is likely to be  within  the fol-
lowing range:

                               0.4 > M > 0.2

     Because of its applicability over the entire boundary layer, we selected
the mean velocity power law relationship [Eq.  (11)] as the most suitable avail-
able description of the wind speed shear.  We  chose 0.2 as a representative
value of M for an urban area, such as Los Angeles.

2.   Implementation of the Wind Velocity Profile

     The numerical  integration scheme used in  the grid model requires  that the
average wind velocity be specified at each grid cell interface.  The integra-
tion of Eq.  (11) along the vertical  axis from  the lower cell boundary to the
upper, followed by division by the cell depth, yields the expression for the
mean horizontal wind velocity over a horizontal cell interface:

-------
                                                                        91

                 u = 	^R	•  /ZM+1  _ ZM+1\                   ,12)
                 "   (V VZR(M + 1)  \t   ~  b  /

where Z,  is the elevation at the top of the cell and Z,  is the elevation at
the bottom of the cell.   Equation (12)  can be used to obtain both the x and
y components of the mean velocity for each horizontal grid cell interface
within the modeling region.   Assuming that turbulent atmosphere flow is in-
compressible, the vertical advective velocity,  ws can be computed from the
continuity relationship:


                             3x   3y   3z

3.   C o mp u t erjCodjjnj_

     To incorporate the wind shear algorithms given by Eqs. (12) and (13),
we made appropriate coding changes in the computer programs embodying the
airshed mode"!.  The result of these alterations was a slight increase in
both machine storage requirements and CPU time.

4.   Description of the Experiment

     After we altered the computer codes, we designed an experiment to ex-
amine the sensitivity of the airshed model to wind shear effects.  To insure
that the deviations in predicted concentrations are caused only by dissimi-
larities in the prescribed wind fields, we made test runs using both the
original code (in-which a flat profile was assumed) and the newly revised
code with M set equal to zero.   From Eq. (12), if M = 0, we obtain the same
flat wind profile as was used in the original formulation of the model.
Meteorological and emissions data for Los Angeles on 29 September 1969 served
as input data for predicting concentrations of RHC, URHC, NO, N02, 03, and CO,
using both the modified and unmodified programs for the hours 0500 through
1500 PST.  The two programs produced identical  predictions, thus indicating
that all coding alterations had been implemented properly.

-------
                                                                        92
     Finally,  we ran the modified program using a value for M of 0.2,  and
we compared the output of this  program with  the previous unmodified compu-
tations (assuming a flat velocity profile).   Figures  13 through  20 show
plots of the average and maximum deviations  in ground-level  concentrations
for each species as a function  of time.  Here, the concentration deviations
are defined as the predicted concentrations  with wind shear minus the
corresponding concentrations predicted when  wind shear is neglected.

5.   Discussion of the Results

     An examination of the simulation results reveals the significance of
incorporating the power law wind profile in  the grid  model.   Because the
velocities are systematically altered through the application of the wind
profile algorithm, the computed wind velocities at the inversion base  and
ground-level heights were increased, relative to the  straight profile  •
values (M =0), by as much as 70 and 20 percent, respectively.   The wind
velocities averaged over the entire mixing depth were consistently much
larger than the uniform profile values.  As  one would expect, therefore,
the results of the sensitivity  experiment, which was  performed with a  25
percent increase in all wind velocities (see Chapter  IV of Volume I),  are
strikingly similar to those shown here.

     A characteristic of both the wind speed and wind shear sensitivity
studies is that, when the concentration maps are compared with those gener-
ated for the base case, a perceptible translation of  concentration isopleths
toward the northeast, the prevailing wind direction,  is observed.  In  addi-
tion, the majority of maximum concentrations are located in Grid Columns 20
through 25; this result was expected because the translation of  concentration
isopleths is greatest when the  path of travel is longest.  The fact that
average overall deviations for  all  species are negative also supports  the
hypothesis that the net effect  of the inclusion of wind shear is similar to
that resulting from a simple increase in wind speeds.

-------
                                                                                  93
      1.2
      1.0
      0.8
      0.6
              O  NO

              D  .NO,
      0.4
cx
cx
 I
 I
O)
O

CD
O)
to
$_
Ol
 0.2
-0.2
     -0.4
     -0.6
     -0.8
     -1.0
     -1.2
                                   _L
_L
         500       700       900       1100      1300      1500


                              Pacific Standard Time--hour
                FIGURE 13.  THE EFFECT-EXPRESSED AS AVERAGE DEVIATION—OF
                     VARIATIONS IN VERTICAL WIND SHEAR ON NO AND  N00

-------
                                                                               94
      30
                  O  NO

                  D  NO
      20
c
O
O)
O
O)
en
c:
QJ
o
S-
II)
D.

Q)
CD
Itf
S_
QJ
>
cf.
      10
                                       I
        500      700       900       1100       1300      1500


                              Pacific Standard Time—hour
              FIGURE 14.   THE EFFECT—EXPRESSED AS PERCENTAGE DEVIATION—OF
                      VARIATIONS IN VERTICAL WIND SHEAR ON NO AND M09

-------
                                                                               95
      15.0
                    O  NO

                    D  NO,
      10.0
 Q.
 D-
 I
 I
QJ
O

E
Z!


X
re
s:
       5.0
          500       700       900       1100      1300      1500


                               Pacific Standard Time—hour
                FIGURE 15.  THE EFFECT--EXPRESSED AS MAXIMUM DEVIATION—OF
                       VARIATIONS IN VERTICAL WIND SHEAR ON NO AND N00

-------
                                                                              96
-p
ro
OJ
O


-------
 CO
o
-G
D-
Q-
 I
 I
     3.0
      2.0
      1.0
                  O CO

                  D °3
i.
Q-

 I
C
O
 
-------
                                                                             98
     30
                 O  CO

                 D  °3
 c
 O
+J
fO
d)
01
rO
QJ
O
5-
0)
D-

O)
CD
n3
S-
QJ
20
     10
                                      _L
       500       700        900       "MOO      1300      1500

                             Pacific Standard Time—hour
            FIGURE  18.   THE  EFFECT—EXPRESSED AS PERCENTAGE DEVIATION—OF
              '   VARIATIONS  IN  VERTICAL WIND SHEAR ON CO AND 00

-------
                                                                                99
 CO
o
.c
o.
ex
 I
 I
i.
Q_
 I
 I
03



>

QJ
£
X
to
s;
       500       700       900        1100       1300       1500



                             Pacific  Standard  Time—hour
             FIGURE 19.  THE  EFFECT—EXPRESSED AS  MAXIMUM DEVIATION—OF

                   VARIATIONS  IN  VERTICAL  WIND SMEAR ON CO AND 00

-------
                                                                                    100
+J
ID
O)
Ol
td
4->
C
OJ
O
l-
(L)
D_

E
'a


X
10
     200
     180
     160
     140
     120
100
 80
      60
      40
      20
                  O CO

                  D °3
                            JL
        500      700        900      1100       .1300       1500

                              Pacific Standard Tinre--hour
         FIGURE 20.   THE EFFECT-EXPRESSED AS PERCENTAGE MAXIMUM  DEVIATION-OP
                     VARIATIONS IN VERTICAL WIND SHEAR ON CO AND  0

-------
                                                                         101
     The results of this study clearly indicate that wind shear phenomena
should be included in the airshed model.  Of course, it would be most help-
ful in the construction of velocity profiles to have wind data taken aloft
in cities where the model is to be applied.

B.   TREATMENT OF WIND SHEAR IN THE AIRSHED MODEL

     On the basis of the sensitivity results presented in the previous
section, we included provisions in the computer programs for treating a
fully three-dimensional wind field.  To facilitate usage of either theo-
retical wind shear relationships or actual wind data aloft, we structured
the air quality simulation program to accept the three-dimensional  wind
field inputs directly from the meteorological  data  file.  The user assembles
the wind field inputs by employing the Automated Meteorological  Data
Preparation Program.  Thus, all wind shear algorithms and interpolation
routines are embedded in the meteorological data program.  By structuring
the airshed simulation package in this way, we.have enabled changes in the
treatment of wind shear to be accomplished without modifying the photochem-
ical dispersion model code.

     In many urban areas, sufficient soundings of the winds aloft are sel-
dom available for use in constructing the complete flow field.   Therefore,
we initiated efforts to derive a set of theoretical  wind shear relationships
using results obtained from Deardorff's planetary boundary layer model.
These relationships are presented and discussed in Chapter II of Volume III.
For use in those urban areas where numerous pibal  or other suitable data are
available, we recommend that an algorithm be developed and installed in the
Automated Meteorological  Data Preparation Program for the construction of
wind fields aloft using the available wind soundings.

C.   EXAMINATION OF AN ALGORITHM FOR DERIVING MASS-CONSISTENT WIND FIELDS

     One of the assumptions commonly invoked in airshed modeling is that the
air flow  in  the planetary boundary layer is incompressible.   Under these con-
ditions,  the  velocity components  satisfy the following continuity relationships

-------
                                                                         102
                         3Ji+ 3V + 3W = Q                                 /]
                         ax   ay   3z   u    '                            U

where u, v, and w are the x, y, and z components, respectively, of the wind
velocity vector.  In the SAI model, the z coordinate is normalized by the
depth of the modeling region, in which case Eq. (14) becomes
                      8UAH   3vAH_   SW _ A
                         ~             = U
where
        P  =  [z - h(x,y)J/[Ht(x,y,t) - h(x,y)J,
        H.  =  the elevation of the top of the modeling region
        h  =  the terrain elevation,
        AH =  Ht(x,y,t)  - h(x,y),
        W  =  w -•u[(3h/8x)  + p(8AH/8x)]  - v [(ah/3y)
     In typical  airshed model  applications, estimates of u(x,y,z,t) and
v(x,y,z,t) are obtained from the available data on wind speed and direction,
both at ground level  and aloft.   Once the horizontal components are specified,
Eq. (15) can be solved for W.   Writing this equation in finite difference
form, we obtain
w. • lxl  = w. .  ,  l  - -^  (UAH)..!   . ,  - (UAH)
 i i Ix -U-L---    n "I  [/ _ £>-•   A Y I      T "ri1*'  T K
                                                           1->,,J,k]
where the integer triple (i,j,k)  designates the center of a grid cell.  Equa-
tion (15) is  solved subject to the constraint of

                               W  = 0                                     (17)

-------
                                                                        103
at p  =  0,  which  simply  states  that  either the wind speed is  zero at the
ground  or  the  wind  is  flowing  parallel  to the terrain.   By first estimating
the horizontal wind components  and  subsequently solving Eq.  (15) for W, we
insure  that  the  net flux  of  air into  each grid cell  is  zero.

     One difficulty associated  with the windifield methodology described
above is that  nonzero  vertical  velocities may be calculated  at the top of
the modeling region.  Thus,  pollutants  may be advected  out of the modeling
region, even though a  stable capping  inversion layer is present.  This situ-
ation is somewhat  contrary to  the usual  belief that an  elevated inversion
layer suppresses vertical transport,  although buoyant air parcels may pene-
trate the  stable layer  to some  extent.   It is important to note that the
calculated vertical  motions  are,  in part, the result of inaccuracies in the
predicted  horizontal wind components, especially aloft, where few measure-
ments are  generally available.

     In previous efforts, we examined means  for removing convergence and
divergence areas in the flow field  aloft (see Roth et al., 1971).   However,
these attempts to  force the  vertical  velocities to obey some  specified con-
straint, such  as a  zero velocity  at the inversion base, failed to produce
acceptable wind  fields.   In  many  instances,  the algorithms generated hori-
zontal  wind  speeds  aloft  in  excess  of 40 mph.  Under the present contract,
we revisited this  issue of constructing mass-consistent wind  fields in light
of the  findings  of  recent studies in  this area reported in the literature.

1.   The Governing  Equations

     The problem that we address, here is as  follows: Given  a set of initial
estimates  of u and  v over the modeling  region, how should these wind speeds
be adjusted  to yield vertical wind  velocities that not  only  satisfy Eq. (14),
but also obey  some  imposed constraint.   The  methodology described below is
similar to that  given by  Fankhauser (1974).

-------
                                                                        104
     Let UQ  and  VQ designate  the initial  estimates  of the horizontal  wind
components,  which  have  been obtained  through,  say,  the application of inter-
polation procedures.  The values of u and v to be employed in the grid model
are obtained by  defining  a function $ in  the following manner:

                     UAH =  (UAH)Q +  |i    ,                             (18)
                      VAH  =  (VAH)Q +       .                              (19)

Note that we attempt here  to adjust only un and vn,  not AH.   Substituting Eqs.
(18) and (19)  into Eq.  (15), we obtain
We define the terms on the right-hand side of Eq.  (20)  as  follows:

                         D  = -  —

                              8(uAH)0   8(vAH)Q
                          0     3x        8y.

Thus, Eq. (20)  becomes

                          A   A
                          d 


-------
                                                                         105
     Operationally, Eq. (23) is written in finite difference form ancMs
 solved on successive layers of grid cells in the x-y plane.  Thus, in
 finite difference notation, Eq. (23) becomes
                    [V2*]      = D. . k - (D )         ,                 (25)
                         1  j |<     i >3 >*•     U 1 ,J ,K

        2
where [v 
-------
                                                                        106
more satisfactory relationship for D - D~.   Note that the perturbations to
the flow field near the surface are significantly smaller if one uses
Eq. (27) than they are if Eq.  (26) is used,  as demonstrated in the next
section.

     In specifying boundary conditions,  one  has two possible choices:  the
Dirichlet (= 0)  or the Neumann (8/9n = 0)  boundary condition.   Physically,
the former treatment leaves the u  component  of the velocity unaltered along
boundaries parallel  to the x-axis  and the v  component unaltered  on boundaries
parallel to the y-axis.  In the latter case, just the opposite is true.
Fankhauser (1974) employed the Dirichlet condition in his study, and Liu et
al. (1974) report that in simulations using  a similar type of model, the
results were not  significantly influenced by the choice of one formulation
over the other.

2.   Tests of the Model
     To test the model  described in the previous  section,  we carried out a
study to determine the  magnitude of the alterations  that would be predicted
for the typical  wind fields  previously used as  input to the SAI airshed model.
Thus, we rendered the wind fields used in the 29  September 1969 model  evalua-
tion study for Los Angeles (see Reynolds et al.,  1973)  mass consistent; more-
over, we constrained the vertical velocity W to be zero at the base of the
inversion layer.  We used a  25 x 25 x 5 grid layout, where Ax = Ay = 2 miles
and  Ap= 0.2.   Since wind shear was neglected in  the Los Angeles study, we
considered UQ  and VQ to be functions only of x, y, and  time.   Tables 13 through
15 illustrate  the nominal  wind speeds and directions and mixing depths for
6 a.m.  and 3 p.m.  on 29 September 1969.   These  maps  served as the inputs to
the mass-consistent wind algorithm.

     In performing the  calculations with the model,  we  wished to assess the
sensitivity of the predictions to (1) the manner  in  which  D - DQ is approxi-
mated [i.e.,  the use of Eq.  (26) or (27) and (2)  the choice of either
Dirichlet or Neumann boundary conditions.  Furthermore, we examined the nature

-------
                                                                           107
                            Table  13

HOURLY -AVERAGED WIND SPEED AND  DIRECTION IN THE LOS ANGELES BASIN
              ON 29 SEPTEMBER 1969 AT 6:00 a.m. PST

                          (a) Hind  Speed
                                       I*  15   16  IT  18  19  2« Zl 22 23  24  23
81
24
20
Z2
Zl
20
14
la
ir
16
15
14
13
12
11
10
«
e
T
6
I!
4
8
9
1
23
24
23
22
21
20
19
in
ir
16
19
14
13
12
II
10
9
o '
T
«
8
4
a
1.0
1 ,u
I.e


1 .0
1 .6
2.0
8.0
0.6
2.5
2.6
2.0
2.6
2.6
2.6
2.0
2.0
90
01
60
73
76
77
77
70
79
79
00
01
02
00
04
84
03
C7
08
98
9»
90
90
»
i.e
i.e

1.0

1 .V
I .0
z.e
3.0
3.0
2.5
2.0
2.0
2.0
2.0
2.6
3.0
2.0
Z
103
93
05
T>
60
01
01
02
03
03
04
04
04
03
03
03
06
00
69
91
91
91
91
41
1.0
I.D

2.0

1 ,u
I .D
2.0
3.0
2.6
2.6
2.0
2.0
2.0
2.0
2.8
2.0
0
loa
97
04
03
(16
06
O7
07
no
00
or
07
00
06
07
00
90
92
92
92
93
a.e
2.6

3.0

1 .0
1 . D
2.0
3.0

2.0
2.0
2.0
2.0
2.0
2.6
2.0
4
1 13
106
96
94
93
92
91
91
92
91
90
09
09
07
OB
09
91
93
93
93
93
93
0.0
Z.D

3.0

i.e
1.0
2.0
3.0

2.0
2.0
2.0
2.0
2.0
2.0
2.0
B
1 10
113
110
103
102
99
V7
96
94
92
90
00
00
09
90
92
94
94
94
94
»4
3.0

3.0

1. u
1.6
2.0
3.0

2.0
2.0
2.0
2.0
2.0
2.0
2.0
6
123
121
110
112
109
163
102
100
97
94
91
09
09
90
92
94
96
96
96
96
96
0. V

3.0

1 .u
1 . o
Z. W
3.6

2.0
2.0
2.6
2.0
2.0
2.0
2.0
T
120



117
1 14
110
106
104
106
90
90
90
92
94
16
90
90
90
90
90
S.u

2.0

2.0
2.6
3.0
4.6

2.6
2.0
2.6
2.0
2.0
2.0
2.0
a
136



124
120
116
11 1
107
163
99
91
90
93
96
90
106
100
100
too
109
ice
2.3

2.0

2.0
2,3
a.u
4.3

3.0
2.5
2.0
2.0
2.0
2.0
2.6
2.5

2.0

a.O
a.o
4.0


4.e
2.5
2.0
2.0
2.0
2.6
2.0
2.6

2.0

3.0
3.G
4.U
4.0

5.0
a.o
2.3
2.0
2,0
2.0
2.0
(b) Wind
9 19 II
141 146 HI



191
126
122
1 17
no
106
102
96
96
90
99
100
102
102
102
102
102
102



130
133
129
12D
113
100
107
193
103
103
103
194
104
104
104
10-4
104
104



146
100
131
I2G
llfl
113
1 I I
ion
ion
107
107
IOC
1 06
106
106
IOG
106
166
2.0

2.0

3,6
4.0
*.W
8.0

o.t»
3.0
2.5
2.5
2.0
2.0
a.e
2.6

2.6


4. «
3.0
0.6

0,0
S.fc
3.0
'2.C
2.5
2. fl
2.0
2.0
2.0

a.e


9.O
O. W
P.O

b.U
3.0
3.0
2.6
2.0
2.0
K.O
^. V
2.0

2.6


4.6
0.0
6.0

o. u
3.0
3.0
2.6
2.0
2.0
2.0
2.0
Direction
.12 is 14 in
149 ICO IGO 130




142
137
139
1 17
116
IIS
113
III
110
109
ion
100
100
loa
100
100
147


141
139
129
123
117
1 17
117


115
113
111
119
110
110
110
no
110
146


13C
133
130
120
1 10
1 18
lie


no
110
117
tie
1 1 1
118
no
no
tic
146
100
133
130
120
126
1 10
1 17
110


123
124
124
•125
1 12
110
no
109
1M
2.0

2.5


6.0
0.0
6.0

5.0
3.0
3.0
2.6
2.9
2.0
2.0
16
150
146
135
130
127
123
120
117
1 16
116


126
120
120
I2S
116
(12
lid
109
169
2.8

3.0


6.0
5.0
4.0

6.0
3.0
2.6
2.0
2.0 '
2.0
2.6
17
101
143
139
123
120
119
110
113
II 1
114
122

121
430
130
127
119
111
110
111
111
2.3

3.6


4.5
4.5
4.0

6.6
3.0
2.6
2.0
2.0
2.0
2.0
2.0
10
161
140
123
123
120
1 10
110
100
107
103
103

113
130
130
124
120
III
100
111
III
2.6

3.3


4. D
4.0
3.0

4.0
2.5
2.0
2,0
2.0
2.0
2.0
19
161
136
120
117
1 10
106
109
105
107
106
102

111
126
120
113
I 16
109
111
111
fi.w 3.6 3.5 J.D 4.0 4.0

3.5 4.6 4.0 4.0 4.0 4.U


*.u •i.u *.U 3.0 y.u 3.0
4.0 y.O 3.0 2.5 2.5 2.S
3.0 *.5 2.5 2.5 2.5 2.5

3.0 2.6 *.« 2.6 2.6 2.0
2.5 2.0 a.u 2.6 2.0 ^.»
2.0 2.6 2.O 2.0 2.0 Z.U
2.0 2.6 H.W 2.0 2.0 K.»
2.0 2.0 2.0 2.0 2.0 2. -
2.0 2.0 2.0 -.0 2.6 2. ft
2.6 2-ft 2.0 2.0 2.0 2.0
20 21 22 23 24 23
160 160 130 156 156 150
136 131 120 126 120 120
115 116 I07 105 163 10-)
1 10 95 93 92 91 90
100 95 93 92 91 9**
100 95 93 92 91 90
90 95 93 92 91 90
100 93 93 92 91 91
101 96 92 90 90 90
101 97 93 90 96 99
109 95 93 93 93 93

100 101 99 9G 92 92
100 96 96 92 90 9(»
115 110 100 93 91 96
104 90 93 90 90 96
169 103 96 92 98 90
104 104 ICO 92 96 90
165 IOO 92 90 90 90
135 109 92 99 9* 90

-------
                                                                       108
                           Table 14

HOURLY AVERAGED  WIND SPEED AND DIRECTION IN THE LOS ANGELES BASIN
              OH 29 SEPTEMBER 1969 AT 3:00 p.m. PST

                         (a) Wind Speed



22
20

17
15


11
19
9
B
7
D

2


24
£3
21





14


'•



6

1
1
1


6.0
B.O

0.0
6.0
7.0


0,5
B.9
8.0
0.9
0.0
0.0

B.O


131
126
110





7»


71



74

79



6.0


6.0
9.0
7.0


8.5
0.0
0.0
0.0
0.0
B.O

n.e


123
126
110





69


71



74

74
V4



6.0
6.6

6.0
7.0
8,e


0.5
0.0
o.e
o.e
o.o
o.o

o.o


lie
112
no





66


72



74

74
74



6.0
6.5

9.e
7.0
o.e


o.e
B.O
o.e
o.o
s.o
0.0

7.3


IB7
110
no





69


72



79

73
73



6.0
5.5

9.0
7.0
9.0


o.e
B.O
o.e
e.e
o.o
o.o

7.6


90
99
103





65


70



75

73
73



6.0
3.5

C.O
7.0
9.0


0.0
0.0
8.0
0.0
0.0
0.0

7.0


94
93
93





66


60



73

73
73
7


6.0
O.i5

7.0
9.0


0.0
0.0
0.0
0.0
0.0
B.O

7.0


90
«e
90





66


67



73

73
73
e


6.0
8,0

6.e
o.e
8.6


9.0
o.e
8.9
0.0
o.e
o.o

6.0


83
83
79





66


66



V3

73
78
»


9.6 6
9.0 v

7.0 7
B.O 0
9.0 9


10.0 11
9.O 10
9.0 JO
0.9 9
0.9 D
o.e 8

6.6 0
(b]

70
03
73





66


64



va

73
73
10 11 12 13 14


.0 B.O 0.6 G.O 6.0 6
. w o.o (j.w 7,0 D.O 6

.0 0.0 ?.t> 9.0 10.0 12
.0 y.u 10.0 10.0 12.0 11
.0 10,0 11.0 12,0 I (.0 10


.0 12.0 14.0 13.0 11. w 10
.0 12.0 13. 0 13. V 11, e 10
,0 11.0 12.0 13.0 ll.o 10
.0 9.0 10. 0 11 .0 10.0 9
.5 0,6 9.0 10.0 *.6 0
.0 0.0 0.0 D.O 7.6 f

.0 C.O 6,0 e.e S.O 0
Wind Direction

70 66 60 37 S3
65 60 47 00 36
63 DO 40 41 40


.


65 64 62 62 61

•
64 62 64 70 66


'
73 73 73 73 73

73 73 V3 73 73
73 73 73 T3 73
It)


.0
.0

.0
.0
.0


.0
.0
.0
.0
.0
.0

. W


36
36
43





61


60



73

73
73
1C


7.0
9.0

12.0
10.0


10. 0
9.0
9.0
0.0
7.0
6.B

0.0


36
36
44





60


72



73

73
73
17 13 19


0.0 B.O 9.0
10.0 10.0 11. w

11.0 10,0 -y.U
10.0 9.0 0.0
9.0 0.0 7.0


9.0 9.0 0.0
9.0 n.O 7.0
0.0 7.0 6,0
7.0 6.0 B.6
6.0 6,6 C.O
6.0 C.O C.O

0,1) D.O 5.0


36 36 36
07 37 37
40 39 37





CO 63 63


72 72 72



73 73 74

73 73 73
73 ra 70
20


v.O
I I.u

0.0
7.W
7.0


7.0
6.0
6.0
5.5
0.0
6.0

6.0


36
37
37





65


72



74

77
77
21 22 23


9.0 9.0 9.0
10. 0 9.0 u.o

O.U 7.0 6.0
7.0 6.0 «.0
6.0 5.U 6.0


6.0 6.0 t.u
5.0 4.0 *.U
4.0 4.0 i.u
4.3 4.6 4.0
6.0 4.E 4.3
B.O 5.0 6.0

5.0 5.0 G.»


39 34 31
06 33 30
37 36 35





65 70 72


73 74 75



77 77 79

79 02 83
79 fl2 03
24 21


9.0 B.O
7.0 7.0

b .U it. «
&.B 6.0
» . 0 5. y


4.0 t.v
4.0 4.0
4.0 4.0
4.tJ 4.P
4.B 4.0
4.5 4.5

o.o S.O


30 30
30 30
35 43





73 73


75 75



79 HO

03 BJ
fi3 O't

-------
                                                                    109
                     Table  15
       MIXING DEPTHS  IN THE LOS ANGELES BASIN
ON 29 SEPTEMBER 1969  AT 6:00 a.m. AND 3:00 p.m. PST
                   (a) 6:00 a.m.
S3
24
33
23
21
19
IB
IT
ie
in
14
13
12
11
ie
»
D
r
0
«
8
a
1


29
24
23
22
21
20
19
II)
17
16
19
14
19
12
II
1.0
<
8
y
t
I
4
8
2
1
1
100
100
ioe
led
2oe
809
400
DOO
DOO
Doe
BOO
BOO
DOO
DOO
DOO
COO
DOO
DOO
000
COO
000
DOO

1
2599
2509
2509
2409
2000
1300
1200
909
600
650
050
650
530
630
BGO
050
E50
BOO
CCO
D70
060
030
CflO
BOft
ioe 100
100 joe
100 100
160 109
209 £08
000 300
400 400
060 COO
000 600
000 600
009 DOO
660 COO
COO 000
COO 000
COO 600
600 600
600 600
009 BOO
660 COO
60& GOO
500 000
069 600

2 a
2309 259C
2590 2500
2590 2300
2400 24 CIO
2000 2000
1COO 1600
1200 1200
909 909
600 600
059 600
630 550
650 650
000 050
009- COO
630 BOO
650 650
050 050
630 550
030 R00
630 060
650 BOO
630 659
BB0 KliO
660 630
140
100
100
100
200
300
400
400
600
600
669
009
CCO
G09
GOO
coe
006
GOO
000
009
009
060

4
2500
2ceo
2500
2100
2000
1500
1200
900
600
059
630
GOO
500
550
530
030
050
000
650
630
000
039
090
650
104 104
100 160
109 169
J00 166
200 200
300 273
373 323
430 410
496 460
BOO 470
500 400
500 500
BOO 500
G(JO 600
000 600
600 000
GOO 600
500 000
BOO 000
600 500
000 090
000 009

6 6
2306 2300
2500 2300
2500 2300
2400 2200
2000 2000
1500 1500
1200 1200
909 909
600 600
050 676
650 000
550 650
550 050
006 G50
000 BGO
650 530
650 550
559 050
036 530
030 000
OGO 060
030 630
CCO BF»
6170 6BO
106 166
100 100
109 100
109 100
200 200
25 & 230
000 300
350 323
410 330
423 075
450 409
499 450
470 460
490 450
GOO 400
400 356
400 356
000 600
060 GOO
000 GOO
609 BOO
COO 000

7 0
2400 2300
2400 2300
2200 2200
2100 2100
2000 2000
1500 1500
1200 1200
900 909
600 600
600 609
070 600
609 600
OHO 073
650 070
BRO 576
550 670
659 650
G50 500
659 000
000 030
000 069
050 630
000 S50
BOO 600
fr !&
IBO 200
100 ISO
100 [00
100 100
206 200
2S0 225
273 250
300 275
323 300
350 325
370 309
400 378
430 469
400 400
4.39 400
300 400
350 3DO
600 BOO
500 600
300 000
COO 600
600 COO
Cb)
9 10
2200 2000
2200 2000
2209 2000
2200 2000
2200 2000
1000 1700
1500 MOO
1000 1006
O00 900
600 673
630 650
600 630
600 630
600 630
070 610
670 609
050 G73
BOO 073
600 609
050 609
000 000
650 OQO
050 000
500 DOS
a 12
20O 2O«
200 200
160 160
100 100
ISO 140
200 176
223 206
250 223
270 230
300 £70
323 300
3GO 325
375 330
409. 400
423 430
450 450
400 493
DOO 660
600 500
DOO 500
000 GOO
BOO 000
3:00
II 12
1009 IBC0
1900 1700
1900 1000
1900 1900
1000 JOOO
1500 IS00
1300 1300
1 100 1 100
9GO *930
700 723
600 700
660 690
600 700
600 700
6GO 670
625 640
600 600
676 670
039 BSO
GOO 030
OD9 6BO
BOO 839
000 030
CBQ 669
13 14
200 206
206 200
IGO 200
100 100
120 109
100 123-
175 150
200 J73
223 200
250 223
270 236
300 360
350 350
400 400
450 450
4110 473
493 490
COO 600
600 609
500 COO
DOO 600
609 GOft
p..m-
13 14
1200 1209
16 WO/ 16W&
1(100 1(103
1900 2000
1000 2000
1G00 1000
1300 1009
11 09 1209
9HO 930
750 773
729 7+9
709 000
700 700
700 600
709 730
609 700
609 625
G70 075
006 000
050 660
630 600
050 000
600 050
6«0 600
ID
200
200
209
109
100
109
125
150
170
206
230
200
340
400
450
470
4Q5
GOO
600
GPO
609
000

16
1209
iS00
1000
2100
2300
20UO
1700
1409
1000
025
noo
900
000
03d
773
700
623
670
690
069
COD
059
BOO
609
16 ir
20ft 200
200 200
200 200
100 100
100 100
too toe
100 100
125 100
160 123
173 150
223 . 200
270 225
330 250
375 300
400 350
450 400
475 450
660 495
CCO 308
500 305
609 GOO
666 600

I* 17
1Z69 1209
1609 1300
1000 moo
2200 2200
2300 2300
2300 2500
2000 2306
1700 2200
1100 2000
1200 1500
1200 1400
1100 1309
1060 1100
900 1000
ooo nno
700 750
600 650
600 600
GffO 670
000 009
009 5H0
009 IF30
068 060
690 666
18 19
200 20O
200 200
200 209
100 100
166 100
100 106
100 100
100 100
100 100
125 123
IGO 109
200 175
259 200
300 ?,GO
330 300
370 359
400 400
490 4OO
496 495
FOO 500
500 500
COO 009

10 19
1200 1209
1600 1500
1000 1000
2200 2200
2300 2300
2GOO 2300
2500 2500
2300 2GOO
2200 2300
1000 2000
1600 1000
1300 1000
1300 1300
1100 1109
930 1009
900 900
000 (100
700 7GO
600 600
073 670
BOO BOO
BOO BOO
609 050
050 600
20 21
299 200
200 200
200 200
100 100
160 100
109 109
100 100
100 100
100 100
123 123
100 150
173 175
200 200
225 225
273 250
323 000
073 350
450 425
400 460
500 4(10
009 600
ceo 600

20 21
1200 1200
1000 1500
1000 1000
2200 2200
2300 2300
2300 2fiOO
2500 2300
2300 2200
2200 2209
2000 20(10
1000 1000
1500 1500
1300 1300
1190 1100
1000 1000
900 900
000 000
730 750
600 600
073 070
BOD 030
600 600
tfSO GDO
669 606
22 23
200 000
200 200
200 200
173 200
100 100
100 109
100 109
123 149
125 (00
150 200
173 200
200 2dO
223 225
250 250
230 250
300 000
323 323
400 375
423 400
•150 -125
400 400
BOO 400

22 23
1200 1209
1500 1000
1000 1000
2200 2200
2300 2300
2500 2500
2300 2300
2200 2200
2200 2200
:;ooo 2000
1000 1OOO
1600 1600
1400 MOO
1200 1200
1100 MOO
1000 1000
050 (130
750 700
600 609
073 073
BOO 600
023 OOO
623 006
030 020
24 21
200 2
200 20J
200 200
200 209
100 I5d
I 69 1 ')"
120 15U
166 15.)
ISO 15-1
200 200
200 200
200 200
223 25..
230 20 •
273 20*.
300 25 ft
325 32-
373 35l>
400 40H
400 40-1
359 :i5 .
350 309

24 23
1200 1200
1569 1500
10(10 1009
2200 220'1
2000 2:U>'t
2500 2509
2300 230n
2200 2200
2209 2200
2000 20OO
1000 i no t*
1600 1600
1400 140H
1200 1200
1 U>0 I I Oil
1000 1000
03 0 05 0
750 750
600 600
075 B75
630 65»
BOO 009
609 450
coe 
-------
                                                                        no
of the changes  in  the  flow  field  at  two  levels  on  the  grid:   near the  surface
and at the  top -of  the  modeling  region  (i.e.,  k  =  1  and k  =  5)  .   The  results
of these simulations are  presented in  Tables  16 through  23.   Table 24  summar-
izes the nature  of the inputs and the  treatment of  the parameters  in the wind
algorithms  corresponding  to each  table.   The  predicted changes  in  wind speed
and direction given in these tables  are  defined as  follows:
                  A6  =  57.2958
where AS and  A6  are  the  reported  changes  in wind  speed and direction,  respec-.
tively.

3.    Discussion  of the Results
     Reviewing Tables  16 through  23,  we note  that wind speed  and  direction  are
altered by no more than  about 11  mph  and  93°,  respectively. To place some persp-eo
tive on these results, we must consider the magnitude of  the  errors  associated
with the input wind  fields themselves.  Typically, the  uncertainties in the
wind fields employed in  airshed simulations  are on the  order of 2 mph  in
speed and 60° in direction.   For the most part, the predicted perturbations
in wind direction are smaller than 60°.  However, significant alterations
in the wind speed are predicted for the 3 p.m. wind inputs.   These predic-
tions are the result of generally higher wind speeds and a greater degree
of convergence and divergence in the interpolated wind  field.
 Since  UQ and vn are considered  to be  independent  of z,  the  perturbations
 calculated using Eq.  (26) are the same  in  each  layer  of grid  cells.   Thus,
 the change in speed and direction is  reported only for  the  bottom  layer of
 cells  (i.e., k = 1).

-------
                                Table 16
        PREDICTED CHANGES IN WIND SPEED AND DIRECTION  FOR CASE
                             (a) Wind Speed
                                             IS  10  17  10  19  2«  21  22  23 24  23






















2


S3
24
23
22
21
28
1«
IB
17
16
14
13
12
11
!•
»
e
r
«
e
4
s
a
i






















-.0 -.1 -.1 -.1 -.2 -.2

i a e < g <
-2 -2 4 10 19 23
0 1 B 10 23 32
2 2 0 10 22. 31
-1 -1 -2 8 IB 26
-13 -14 -II -8 1 15
-21 -22 -21 -11 6 19.
-IB -20 -21 -14 1 12
-12 -12 -11 -10 -7 -2

-« -6 -1 -1 -1 -B
1 0 0-0-0-0
600011
o e • e i o
-•-000 1 1
000 1 0-0
o o e 0-0-2
-o -o -• -e -i -a
-i -i -i -i -i -e
-2 -2 -a -a -i -o
-a -2 -2 -i -i -o
-1 -1 -1 -1 -0 -0
-i -o -o -e -» -o
_» -e -• -» -o o






,3 .6
















-.8 -.3






















-.3






















-.4 -.4
(b) Wind

21 26
86 30
29 80

24 24
27 23
22 20
3 6

-fl -4
-1 -1
-0 0
-I »
-0 0
-1 -2
-4 -6
-4 -6
-e -3
0 -1
e e
e o
ft 0
6 1
1 1
9
14
24
20

24
01
10
6
0
-2
-0
1
2
1
-3
-3
-2
-3
-a
0
0
1
i
a
19 11
10 «
17 10
26 22

27 2B
29 26
ia 20
B 10
I 0
-3 -2
0 -0
o e
6 -1
1 2
0 9
0 11
4 10
-0 6
-2 1
-1 0
0 2
a 8
2 e
8 8






















-.4 -.3













































-.3
Direction
12 18
B 7
7 B
16 14

29 32
36 U3
21 22
7 7
-I -3
-2 -2
o e
-0 -1
-3 -4
1 1
9 7
10 7
11 B
9 II
7 14
6 13
6 9
E 7
4 11 '
4 4
14
e
9
13

09
38
27
10
-2
-7
-0
-2
-B.
-8
G
B
4
B
17
IB
12
0
6
f
10
13
13
14

09
30
21
B
-7
-11
-1
-2
-B
-2
1
6
3
7
10
' 16
13
9
6
4






















-.2 -.1

16 17
16 20
16 21
IB 17

30 42
34 34
17 15
2 2
-12 -12
-13 -10
-2 -0
-0 -2
-6 -B
-6 -10
-2 -I
3 1
1 -4
B -0
B 1
11 B
10 9
B 7
I 4
0 2






















-.1

ID
22
24
10

40
31
9
-e
-i i
-17
-15
-0
-3
-16
-4
6
-2
-1
0
•0
6
7
0
2






















-.1

19
22
24
[0

32
23
6
-1
-0
-13
-17
-14
-6
-22
-7
3
-B
-0
2
-1
4
6
2
|






















-.2

29
23
26
16

23
16
8
-0
-4
-10
-15
-19
-14
-13
-B
-3
7
-7
2
-2
1
6
2
2
























21
23
26
17

19
19
7
2
-1
-4
-II
-10
-16
-11
-fl
-
-------
                                                                         112
                              Table 17
      PREDICTED CHANGES IN WIND SPEED AMD DIRECTION FOR CASE 2*
                           (a) Wind Speed

24





17











3
2
1


M
24
23
22
21
2*
19
ID
17
16
19
14
18

!•
»
a
r
6
0
4
»
a
i






,„ ,,j _.p _.o -.0 .e .3 -.3 -.0 -.2












-.0 -.0 -.e -.9 -.1 -,i -.1 -.1 -.1 -.1
-,e -.* -.e -.0 -.1 -.1 -.1 -.1 -.1 -.1
-.B -.e -.» -,o -.1 -.1 -.1 -,i -.1 -.1
(b)
ia04B67BS10
-Z -3 -» 0 B T 0 6 a 2

1 1-08497908

-3 -3-8-1-1 2 7 6 7 18
-7 -7 -0-225 11 11 16 13

-B -4-3-3-1-8-8 I 1 2
-1 -1 -1 -1 -1 -2 -1 1 68
-e -e -» -e » -i -a -e -i -i
1 8-8-8 -1 -2 -3 -2 -8 -2




8 8 8 8-8-8-1-2-8 8
-0-8 8 8 -8 -8 -I -1 -8 1
-8 -8 -8 -8 -8 -0 -8 -6 -8 -8
-1 -1 -1 -1 -8 -» -8 -0 -1 -8



-8 -« -e -8 -8 8 8 8 « 0


-.1












-.3






-.1

.8












-.2





-.1
-.1
Wind
11
B
8

11
11
0

4
8
-0
8
-8
-6

4
«
a
i
8
8
8
1
1
I














-.2





-.1
-.1









a




-.2






-.1




















-, I
~. 1
16 17


















-.1 -.0
-.1 -.e
-.1 -.1
10 19


















-.0 -.e
-.0 -.0
-.0 -.0
26






-. I













». l
21 22



















-.1 -.1
-. i -.1
23 24 . 21
-., -., -.,-





.3 .0 -.e












-.ft .« .0
-.e .B ,o
Direction
12 13
ft
1

10
11
id

i
-i
-i
-0
e
0

3
2
3
2
2
I
1
1
1
1
2
2

1 1
12
11

8
-1
-6
2

-8

S
2
2
4
4
4
2
2
1
1
14
8
2


13
14
0
2
-0
-4
I

-e

2
3
0
1
6
4
3
2
1

IS
4
4


11
12
2
«
-2
-4
-0

-a

e
3
e
2
i
6
4
2
2
1
Ifi 17

9 0
2 4

11 12
9 9
1 2
-0 1
-S -1
-3 -4
-1 -4

I «

-1 0
Z 1
0 -3
2 -6
3 -0
4 1
S 4
Z 2
a i
i e
is 19

a 7
4 4

11 0
e 4
0 1
6 -0
-2 -2
-4 -4
-5 -4

-I -3

-e -i
6 2
-2 -5
-e -o
-0 2
-0 -2
9 1
3 3
6 -S
9 6
20

«
3
3
7
3
4
-e
-e
-3
-3
-6
-6

-2
-2
7
-6
3
-2
e
3
-0
e
21 22

7 0
4 B
B 3
4 3
. 8 7
2 S
8 1
-8 1
-1 1
-2 -2
-4 -1
-7 -4

-3 -3
-4 -4
3 -8
-7 -4
3 1
-1 -8
-6 -4
-2 I
2 -8
8 -8
23 24 23
2 e -y
* I -0
0*4
433
4.4 5
a ie s
5 6 5
331
1 2 "•
e -e -h
-i -i -i
-1 -2 -1
-2 -2 -2

-2 -3 -4
-4 -3 -4
-0 -I -J
-3 -1 -1
0 -1 -I
-1 -I -1
-2 -1 -1
-1 -1 -1
-1 -1 -1
-ft -0 -A
* See Table 24 for a  description of  the experimental conditions.

-------
                                                                          113
                               Table .18
        PREDICTED CHANGES  IN  WIND  SPEED AND DIRECTION FOR CASE 3*
                             (a) Wind Speed

u
14
ss
22
91
19
10
16
ID
14
19
ia
1!
10
t
0
I
6
i
4
g
a
i

-.» -.« -.B -l.» -l.T -
-.« -.2 -.9 -1.8 -1.6 -
-,t -.a -i.o -1.6 -i.e -
.w -.0 -1.2 -l.Q -1.3 -
.a -.8 -.6 -l.li -1.0 -
.3 .a -.0 -.3 -.4
.2 .1 -.« -.1 -, I
'.0 -.8 -.1 -.1 -.1
-.0 -.1 -.-I -.2 -.2
-.0 -.1 -.2 -.3 -.3
-.0 -.2 -.2 -.3 -.2
-,« -.1 -.2 -.8 -.2
-.0 -.1 -.2 -.2 -.2
-.0 -.1 -.2 -.2 -.3
-.» -.1 -.2 -.8 -.3
-.« -.1 -.2 -.3 -.4
-.0 -.1 -.2 -.3 -.4
-.0 -.1 -.2 -.3 -.4
-.0 -.1 -.3 -.s -.8
-.0 -.1 -.n -.3 -.3
-.« -.1 -,2 -.2 -.3

-.8 -.1 -.3 -.2 -.a

•1.9 -
-I .9
•l.X
•1.8
•1.0

. 1


-.0
. 1
-.1
-.2
-.2
-.3
-.5

-.*




-.4

•1.1 -.8 -
-.9 -.6 -
-.0 -.4
-.9 -.3 1
-.8 .6 i

.4 -.5 -

-.0 -. j -1
.* -.<, -1
.0 -.6 -
.0 -.1 -
-. 1 .1 -
-. 1 -.1-1
-.3 -.3 -
-.4 -.4 -

-.8 -.5 -

-.t -.0 -


-.4 -.0 -

.4 -.4 -.u -.2 -
.2 -.2 -.S .0 -

.1 1.6 1.4 1.4 1
.. ,., ... ...
.2 -.0 -.3 -.a -
.0 -.7 -I. i -I .2 -I

.3 -1.3 -1.0 -1.4 -2
. 6 - 1 . U -.9-1.4-2
.7 -1.1 -1.6 -2.0 -2
,7 -1.2 -1,8 -2.4 -2
.8 -1.3 -2.2 -1.0 -1
. 1 -1.2 -2.0 -.0 -
.9 -.9 -I .2 -1.0 -1
.6 -,7 -I. 1 -1.0 -I

,8 -.7 -1.0 -J.O -1
.5 -.7 -.9 -.9 -

.6 -.f> -.7 -.7 -

,C -.6 -.6 .-.0 -

.2 -.2 -
.0 -.1 -

. t .2 -

.3 -.9 -1
.B -2.7 -3

. 1 -2. 1 -2
.2 - 1.0 -1
.0 -1.9 -1
, 1 -J.2 -1
.2 -1. 1 -1
.6 -1.3 -1
.6 -1.0 -
,4-1.1 -
.2 -l.u -
.3 - . (t -
.7 -.D -
.6 -.D -
.6 -.B -

.6 -.3 -

.fl -.8 -
.2 -.3 -

. i -.2 -
.0 .3 -
.3 -2.0 -3
,3 -3.0 -4

,2 -2.3 -1
.9 -2.0 -1
.2 -1.7 -1
.6 -1,6 -1
.2 -1 .2 -1
.1 -.0 -
.7 -.4 -
.8 -.6 -
.6 -.3
.2 .«

.3 -.2 -
.4 -.2 -

.a -.4 -
17 10 19 29
.3 -.8 -,» -.4
.* -.3 -.2 -.0

.3 -.9 -.9 -I.I
.1 -.0-1.1 -.9 •
.3 -3.3 -2,6 -1,6 •
. 1 -3.3 -2. 1 -1.6

.7 -.7 -.6 .0

. t -.3 -.2 l.B



.2 .1 .6 .4
.0 .1 ,3 .3
.3 .4 .2 .0
.2 .2 .1 -.0

.6 .0 -.0 -.1
.1 -,l -.1 -.3
:z -.2 -.2 -.4
.a -.2 -.2 -.4
21
-.4
-.4

-.6
-l.o
-1.3
-.0
.3
••>


1.2
1. 1
.9
,7
.2
-.0
-.0

-.2
-.3
-.4
-.4
23
-.7
-.6

-.4
-.9
-.9
-.6
-.0
.t




.3
.4
. 1
-.e
.e

-.2
-.2
-.3
-.4
23
-.4

. 1
-.4


-. 1

-. i
-.1
-•<

.*
. I
.6
-.e
.0
. i


-.0
-•>

24 23

.-» . J
.3 .»
.1 .3

.4 .4
.2 . f
-.0 -.11
-.1 -.«
-.• .»



.0 . 1
-.1 .<-
-.1 ..t
.0 .0
. I
.« ...
.0 .v


.1 . «
(b} Wind Direction

23
24
23
22
21
20
19
18
17
16
14
13
12
11
l»
9
t
T
«
D
4
8
a
i

-2 -1 13 83 48
0 2 11 SG 61
3 4 1 35 09
-8 -3 -D 16 47
r22 -24 -24 -0 18

-20 -32 -35 -36 -1
-19 -19 -19 -17 -13
-6 -6 -7 -8 -9
-0 -0 -1 -2 -a
2 1 0-0-0
0 b 0 I 1
80812
-00112
00111
0000-1
-1 -1 -0 -1 -2
-2 -2 -2 -a -2
-4 -4 -4 -4 -3
-0 -3 -S -8 -2
-i -2 -2 -2 -1
-1 -1 -1 -1 -0

-« -« -0 0 0

49
61
(1
H6
35

19
-4
-18
-0
-0
1
e
i
-0
-4
-S
-0
-e
-e
-0
-e

i
7 e
47 39
60 04

D2 44
44 39
42 39
30 38
6 11
-a -8
-13 -B
-I -2
-0 0
-1 1
-0 0
-2 -'»
-0 -12
-0 -10
-0 -0
1 -1
0 1
0 0
0 1
1 2
a a
9 10 11 12
37 19 16 11
43 80 10 13

30 36 39 36

39 41 39 46
31 31 33 97



-1 6 « »
211-1
3 1 -1 -0
e i 3 i
-fr 1 17 17
-C 7 22 22
-4 0 23 23
-7 -ft 14 22
-4 -fl 2 19
6 -3 I 14
I 1 6 12
a s T 11
8079
4 B T 0
13 14
12 17
14 16
22 22
, 37 07

CO D6
42 61

-6 -7

0 -2
-4 -B
-6 -8
2 -*
13 9
16 14
17 I I
23 17
S* 32
R7 29
19 22
14 16
11 12
9 9
19 16
23 81
23 80

32 34

59 30
KB B2
30 20

-24 -35
-4 -6
-B -3
-9 -II
-4 -10
2 -3
10 +
7 2
13 B
19 13
27 19
23 10
16 13
13 10
e «
17 16 19 20


33 36 37 33
OB 37 30 27

64 64 03 36
43 30 17 13
4 -3 -3 -1

-43 -33 -23 -17
-19 -20 -27 -24
-0 -16 -24 -29
-14 -9 -IB -22
-16 -24 -32 -19
-4 -0 -12 -13
168-4
-5 -3 -B 7
0 -I -I -9
3621
9 1-1-3
IB 10 6 2
12 11 10 8
T 0 4 4
4088
21


33
28
36
34
14
4

-7
-17
-26
-23
-16
-13
-9
1
-12
e
-3
-2
-1
B
9
22


41
20
29
32
21
10

2
-10
-20
-21
-14
-13
-13
-7
-12
-3
-B
-10
-2
-2
-•
23
20

33
2B
28
33
26
16

3
-11
-16
-20
-17
-14
-1C
-9
-I 1
-6
-7
-0
-7
-0
-I
£4 25
4 -J
10 4
22 13
22 10
25 2-T
07 3 1
29 23
IB 15
10 4
-1 -2
-11 -10
-14 -14
-10 -IB
-20 -2.
-19 -m
-13 -n
-O -1
-7 -6
-7 -0
-7 -7
-7 -7
-7 -7
-6 -3
-2
*See Table 24 for a  description  of  the  experimental conditions.

-------
                                                                            114
                                Table 19
        PREDICTED CHANGES IN HIND SPEED AND DIRECTION FOR CASE 4*
                             (a) Hind Speed

zg
El
£3
22
21
20
19
18
17
16
19
14
13
12
11
IB
9
B
7
6
*
4
3
2
1


23
24
23
22
21
20
19
10
IT
16
13
13
12
II
10
»
0
T
«
B
4
B



-.4 -.5 -.4 -.8 -.4 -
-.4 -.2 -.2 -.1 -.2 -
-.1 .1 .2 .2 .2
.9 .3 .0 .U .9
.7 .6 1.0 1.4 1.6 1

.1 1.2 1.6 2.0 2.0 J
1.0 1.0 2.3 2. B 3.6 4
1.2 S.I 1.1 2.0 3.4 4
.4 1.2 1.3 2.6 2. B 3,
".9 .0 1.0 2.1 3.0 3.
-,S .» 1.3 2.» 2.0 3.

-.3 .U .U 1.9 2.6 y.
-.3 .D l.f) 1,6 2.3 3.




.1 .3 .7 i.i 1,6 I.
.1 .•> ,1 1.0 1.3 1.

.J .3 i« .y ,t I.



-1 I 1 1 «
-6 -7 -2 -0 1
-9 -12 -0 -3 0
-4 -3 -9 -4 1
1 0-2-3 0
23 21 20 10 IB
36 33 33 33 31
39 37 36 34 32
45 43 39 30 30
40 44 37 36 39
3fl 33 32 34 33
26 24 21 23 23
22 23 22 19 10
17 17 17 16 13
13 13 12 11 12
9 10 9 11 9
7 7 7 T 0
44443
23321
« 9 0 » -«

-I -i -i -e -•
-e -o o i i
* T 8 

.» .0 .is . •> .3 .0 .» .^ .1 3.0 4.4 4,0 3.G 2.0 1.1 .t 1.6 2.9 2.3 i.z .4 .z .4 .a .1 .4 .* .4 6 l.f Z.2 2.0 2.2 1.0 1.3 .7 1.1 1,1 1.3 1.1 1.0 .8 .6 .4 .3 .2 .1 .4 4 1.2 l.v 1.6 2.1 1.7 1.3 1.0 .B .8 .0 .7 .b ,D .3 .1 .1 .1 .1 .1 (b) Wind Direction « 70 9 10 It 12 13 14 10 16 17 10 19 20 21 22 23 24 23 0 » 6 2 0 7 9 14 20 9 2 -2 -4 -3 -S -0 1 33 1 2 8 0 B 9 12 17 32 23 10 0 1 1 1 1 3 0 10 6 c 3 1 13 19 20 30 42 47 30 30 24 10 14 II 11 14 17 23 22 23 6 11 21 23 33 43 46 49 41 30 26 16 IB 13 16 19 23 27 20 21 16 17 21 29 34 40 46 40 S3 29 23 17 16 13 10 HI 22 23 23 22 31 32 36 36 39 43 30 34 27 26 22 10 11 12 17 20 21 22 16 17 32 32 35 12 44 37 32 23 22 20 17 11 9 13 13 10 10 10 8 9 3D 43 42 33 31 23 10 10 16 12 9 4 4 3 S 2 1 0 1 3 30 26 23 21 ia 13 0 7 4 1 -2 -0 -11 -0 -6 -3 -1 -3 -1 --• 10 17 16 13 12 B 4 3 1 -3 -6 -10 -14 -10 -It -13 -14 -14 -12 -12 13 13 |3 U 9 6 22-1 -4 -0 -10 -13 -14 -13 -14 -13 -13 -12 -12 , j 0 e | _] -4 -9 -9 -U -12 -13 -10 -16 -16 -16 -13 -15 -13 -in 3 1 -e -3 -7 -10 -12 -13 -12 -12 -13 -14 -1C -14 -12 -12 -13 -14 -13 -16 _» -j -3 -<; -7 -, -n -13 -|2 -11 -11 -11 -12 -12 -11 -11 -11 -12 -13 -14 _„ -i _2 -4 -5 _7 -9 -10 -» -9 -9 -10 -10 -10 -10 -11 -10 -11 -12 -13 -» -1 -2 -3 -4 -0 -6 -7 -6 -5 -6 -7 -0 -9 -19 -U -12 -12 -12 -13 -e -1 -1 -2 -a -3 -4 -4 -2 -2 -3 -4 -» -6 -7 -O -9 -9 -9 -» i a a o » a a i I i i i • • -» -« -i -i -i -i *See Table 24 for a description of the experimental conditions-.


-------
                                                                              115
                                Table  20
         PREDICTED CHANGES  IN WIND SPEED AK'D DIRECTION FOR CASE 5*
                             (a) Wind Spaed

20
24
23
22
21
20
19
10
17
16
10
14
13
12
II
10
9
8
7
«
6
4
9
a
1


25
24
23
22
21
20
19
ia
17
16
10

12
11



6



1


-.2 -.1 -,3 -.1 -.2 -.1 -.1 -.< .2 -.1 .6 .2 .2 .0 -.3 -.4 -.B -.1

-.e .1 .0 .1 .0 .1 .1 -.1 -.6 .*. .9 -.2 -, i -.7 -.6 -.9-1.1 -.7
.1 -.1 ,1 -Z .2 .» .1 .a .a .1 .1 ,| -.0 -,9 -.0-1.0-1.3 -,9 -
.2 .» .8 .4 .5 .8 .4 .6 .4 .2 .5 .1 -.1 -.8 -.3 -.B -1.2 -.7 -
-.8 .1 .2 .3 .4 .5 .6 .B .8 .8 .4 .3 .2 -.1 -,7 -.6 -1.6 -1.4



-.2 ,Z -.1 .0 .4 .V 1.9 1.9 1.2 1.6 1.4 1.4 1.0 1.7 1.0 I.I .2 -.1
-.3 .1 -.1 .4 .7 .9 1.1 1,4 1.9 1.4 l.l .0 1.0 J.0 I.-. 1.1 .3 .0
-.0 -.1 ,2 .-i .r 1.0 J.3 i.O 1.2 1.0 .B .* 1.4 i.« i.a i.l .0 .0


-.2 .a .a .3 ,7 l.e 1.2 1.6 1.3 1.3 .& .a -.2 .7 .b . •* .a .1




.9 .1 .2 .4 .9 .6 .7 .0 .0 .0 .6 .0 .1 .3 .0 .4 .6 .3
,e .1 .2 .3 .4 .6 . B .T . » .7 .« .4 .1 .4 .4 .B .4 .3
.0 .1 ,-t .9 . u .0 .« .3 i.v . v ,o .« .a .3 .4 .a .9 . «


(b) Wind Direction

-4 1 1 ft 0 0 6 6 6 1 3 8 410 2 0-9-1
-4-3-0 6 0 1 I I 8 £ 4 4 14 0 1.2-0 0
-S -6 -1 -1 0 0 2 3 1 8 9 16 1C 9 a 4 1 3


16 9 9 0 6 B 6 6 11 12 20 IB 14 9 B 6 3 4
15 10 14 It 14 1-i IS 1? 14 18 19 14 13 0 0 7 4 2


J7 17 13 • 14 17 13 IS ID 12 9 B 6 O 3 2 1 -0 -2
13 13 11 14 14 13 11 II 0 B 0 3 3 1 « -0 -4 -4






1 | | o e -6 -« -I -3 -3 -3 -4 -3 ' -4 -4 -4 -4 -4'




19 W 51 23 23 34 23
.1 .1 .1 -.1 .<• -.9 -.1
-.3 -.3 -.0 -.1 -./ -.3 -.*

-.B -.6 -.6 -.7 -1.3 -1.2 -l.«
1.2 '-1.0 -1.8 -1.0 -1.0 -.G -I.«
•l.i -l.i -.8 -.6 -.6 -.2 -.ft
-.9 -.w -.6 -.6 -.3 -.7 -. i
-.7 -.0 -.u -.0 .« -.« .1
-.6 -.6 i . z -.v .1 -.0 -.1
-.* -.if -.4 .z -.* .* .-

.2 -.3 .2 -.1 -.1 -.8 .»





.a -.2 .a .a .* ,* .1
.1 .a-.o .a .0 .;i
.4 .Z .1 -.0 ,2 .1 .1
.2 .2 .1 . I -. i .2 . >
,2 .2 .1 ,1 . 1 -.1 .2

.4 .1 .» .t. ... .» .->
.1 .1 ,o -.6 -.0 -.0 .«i


-i -i -o -e i i o
e o i i 3 2 i
2 1 2 1 5 1 4
2234966

9 4 7 6 6 C 6
2*6784
3 4 B 1 3 1
e -o -i -i -i -e »
-a -i a -i -o o -o
-2-2-0 0-1 -0 -1

-4 -5 -4 -4 -3 -4 -»
-4 -4 -4 -4 -4 -4 — *


'
-4 -4 -4 -4 -4 -4



e -» -o -o -o -e -n
*See Table 24 for a description  of  the  experimental  conditions.

-------
                        Table 21
PREDICTED CHANGES III WIND SPEED A!!D  DIRECTION FOR CASE 6*
                     (a) Hind Speed

29
24
fcs
22
21
20
19
la
17
14
IB
14
18
12
11
U
V
B
T

«
a
a
1


29
94
23
22
21
29
19
la
17
16
IS
14
12
II
10
,
B
7
t
B
4
It
»
i
i
-i

-.«
-•>
.0
1.3
1.9
2.6


2.1
1.2
.B
. 1

-.S



.1

.*



1
-2
-12
-IB
r7
4
84
32
03
61
66
96

37
29
22
16
12
7
4
1
-a
-2
-9
-•
« II
-.1 -.9

-.a -.1

.4 .»
1.3 1.9
2.2 2.7
U.I 3.6





I.I 2.3
.9 1.9
.» 1.0

.0 I.O
.6 1.4
.0 1.2

.B 1.0

.3 .«

2 C
e i
-11 -4
-10 -10
-9 -13
2 -I
91 23

92 CO
SO 92
69 91
Bl 47

86 33
20 26
21 19
16 19
12 12
t 7
4 4
1 1
-1 -1
-1 -1
-0 -0
-• 1

-.6 -.T -.T -.» -.9 .1 i

-.2 -.3 -.1 .• -.1 .0 1
.1 .4 .6 .0 .9 I.O 1
1.3 I.O 1.6 1.7 1.7 1.9 *
a. a B.9 B.I 3.4 3.7 3.e 8
9.2 3.7 4.3 0.0 O.I 4.2 9
4.2 9.0 9.9 7.0 7.7 6.2 6,










2.2 3.1 3.0 4.9 6.3 6.3 0.


1.4 I.S 2.6 2.6 3.0 3.0 S.

1.1 1.7 i.O 2.6 2.4 3.2 2.
0)
406789
1 1 I 1 1 *
-0 1 3 6 9 13
-S 0 4 11 16 17
-7 1 0 14 23 29
-2 2 11 10 38 87
25 22 23 23 31 42


90 49 47 01 01 42
49 01 49 46 41 35.
47 43 40 35 33 20

2B 26 24 23 22 . 10
23 23 19 IB 17 IB
17 17 16 14 12 12
16 13 13 12 11 B
10 11 9 6 3 -6

3 2 0 -1 -4 -8 -
0 -0 -0 -2 -4 -4
-1 -1 -1 -2 -a -B
-o -o -o -i -a -3
e i o o i »




.2
.9


.2
.7









6
2


S

9
w-
10
7

31
41
47
47


40
29
24

17
13
0
2
-6

-10
-0
-6
-4
*

11
2.0

1.3
2.3
S.I
4.O
9.4
„.!










4.6


2. B

2.D
ind
U
10

so
01

63


32
29
10


9
B
-2
-12

-14
-II
-8
-0
-1
„ -P
12 13 14 IS 16 17 1Q 19
a. 7 2.9 2.3 .B -.2 -.9 -.B -.9

2.3 1.6 -,2 -I.O -2.7 -3M -2.0 -2.7
2.4 l.o -.3 -1.2 -2.9 -4.6 -3.S -3.6
a. 6 2.2 -.2 -1.3 -3.1 -4.4 -4.6 -4.3
4.6 a. 4 .9 -.6 -2.7 -4.3 -3.7 -3.8
3.2 4.3 2.2 -.3-1.9-4.0-4.7-3.3
0.7 5.0 4.2 1.4 -.9 -3.7 -4.6 -3.1










3.0 1.0 2.2 2.7 2.7 2.7 2.2 1.2


2.1 1.4 1.4 1.7 1.6 1.0 1.3 I.O

2.0 1.6 1.4 1.3 1.2 1.1 .9 .6
Direction
12 19 14 IB 16 17 10 19
13 20 28 13 ft -3 -7 -6

BB 62 07 02 39 22 19 14
60 68 64 63 01 36 80 24

61 G8 &2 61 40 39 33 20
02 40 42 43 40 02 24 24

26 29 22 10 14 9 B B
21 16 12 7 4 -2 -9 -7
12 9 6 2-3 -13 -IB -13




-7 -14 -14 -17 -19 -23 -24 -23
-17 -19 -IB -IB -19 -22 -23 -23

-17 -19 -IB -17 -17 -17 -19 -19
-13 -19 -14 -13 -13 -14 -13 -16
-10 -11 -9 -0 -9 -10 -12 -13
-6 -6 -4 -4 -4 -6 -7 -9
-2 -1 -0 -0 -1 -I -2 -4

20 21 22
-.3 -.0 .1
-1.4 -1.4 -.3
-2.0 -2.5 -2.3
-3.1 -3.0 -2.9
-3.7 -3. 1 -2.9
-3.4 -2.4 -1 .v
-2.4 -1.9 -l.o
-1.8 -1.8 -1.4
-2.1 .0-1.3

-1.4 -.7 -.1
-.3 .1 -.2
.4 .9 .4
.6 .7 .9
.6 .8 .7



1.0 ,
-.2 .1 . 2
.2 .0 .7
.3 .0 .?
i.e .0
.9 .9 .0





.4 .1

- . 1 .0 .i,

23 24 23
C 4 2

30 26 2 a
46 44 30

43 42 41
39 si yo
10 13 IT
237
-0 -0 ft
-3 -3 -3


-21 -29 -2»
-21 -21 -21
-23 -23 -2*
-23 -25 -25

-20 -2» -2.!
-10 -28 -21
-19 -19 -2*
-18 -18 -1^
-0 -9 -3


-------
                                Table  22
        PREDICTED CHANGES  IN  HIND  SPEED  AND  DIRECTION FOR CASE 7*
                             (a) Wind  Speed










13



10
«
0
T
6
0
4
3
2
1



23
21
20
19
10
17

14
13

10
9
0
T
e
9
4
3
2
1










-.1 -.2


-.1 -.2
-.8 -.1
-.0 -.1
-.8 -.8
-.8 -.«
-.» -.8
-.0 -.1

-.8 -.0
-.0 -.e
-.0 .0

1 2

2 B
3 -3
1 -13
-0 -17
-3 -11
-1 -4

» e
-0 -0

0 1
e i
-B 8
« 8
-9 -1
» -«
« 0
• 8
• 8
« 0










-.2


-.2
-.z
-.:





-.0
-.0
.8

3

B
-0
-20
-22
-12


-e
0

2
I
1
e
-i
-8
«
»
1











-.2 -.1 .» .2



-.z -.1 -.1 .1

-.2 -.3 -.3 -.1
-.2 -.3 -.3 -.1
-.1 -.2 -.2 -.2


-.1 -.1 -.1 '-,1
-.6 -.0 -.0 -,fl
.0 .0 .0 .0

4007

3278

-13 -I 9 17
-10 -4 3 13
-13 -12 -7 -0
-6 -0 -II -18

-8 -8 -8 -1
1 2 z »

3329
2 1-1-3
1 1-1-2

-9 » 3 5
0134
1234
1 2 3 C
2346
1122
6 » l»





I -



.2 -.9 -.r -



.2 -.0 -.0 -I

-.1 -.2 -.2 -

-.1 -.9 -.2 -
-.1 -.1 -.2 -

-.8 -.0 -.0 -

.» .6 .1
0) Wind
B 9 10

-I -0 -2

19 16 19
12 10 0
e -i e
-B -6 -6

-2 -1 5
8 2 1

-1-3 3
-6 -1 9
-3 1 11

3 1 2
604
B.e r.
679
6 0 »

II





1



.a



.0

.6
.4

.4

.0
.1
•'
12 13





-.4 -.4
-.0 -.9


-.8 -1.4


-1.3 -.7
-.3 -.1

-.0 -.9

-.0 -.7
-.4 -.2



.1 .1
14





-.9



-1.3



-.0

-.7







IS 16 IT





-1.2 -I. 8 -2.4
-2.6 -1.7 -2.2


-l.i -1.0 -. I


-.6 -.5 -.1
-.6 -.2 .1

-.3 .« .0






.z .2 .2
10 19





-1 .u -.6



1.2 i.n



.3 1.4
.9 1.4
.0 1.2







20 21









2,7 2.3



1.9 2.0

1.1 l.i







22 23 24






1.5 i. i i.o


2.1 l.-r 1.2



i.3 i.2 1.3
!.-» ..J 1-2
l.U .7 1.0

.7 -H .tJ
.0 .0 .7



.0 ,e .0




„•




„



u
.7
.7
.0
.3

•*

•2
•^

Direction
n

-8

9

-1
-9

,
0

19
20
20

0
7
10
11
11

12 13

-7 -0

11 12

-8 -14





19 10
19 14
21 1C.
20 22
19 26
16 23
19 26
10 17
14 1C

14

-9
13
12
-0
-17


-6


7
11
18
10
29
27
23
29
to
9
19 16 17

-10 -12 -7
T B 13
779

-35 -40 -30


-7 -ie -ir


0 -4 -3
720
6 1 -S
11 6-0
16 12 4
26 19 1 1
24 21 19
21 20 IB
19 19 IT
fO 10 9
10 19

-7 -10
B -2
2 -U

-S4 -10


-22 -21


-0 -9
6 4
-4 -B
-l e
2 6
6 3
14 12
IO 17
15 13
8 7
20 21

-13 -10
-6 -5
-9 -2

-14 -9


-17 -II

14
-B -4
-1 -2
11 7
-4 -8
7 7
3 4
9 6
17 7
14 16
7 B
22 23 24

-0 I -5
-2 -O -I
063
-402
-5 -2 -0
-4 -2 -0

-4 -3 -2

960
-3 -1 -2
-4 -3 -1
1 1 I
-2-1 0
9 3 1
4 2 1
1 2 1
1062
14 11 6
099
25

-*
_,
2
O
0
0

n
0
'
a
-i
0
0
6
0
0
n
0
0
*See Table  24  for  a  description  of  the  experimental  conditions-.

-------
                                Table 23
         PREDICTED  CHANGES  IN WIND  SPEED AND  DIRECTION  FOR  CASE 8*
                             (a) Wind Speed

£9



20
10
16
19
14

II
It
»



4
2


23
24
22

IV
10
17
19
14
13
12
II
ie
9
B
T
«
I
«
<
a
i
i
.2



-l.»

1.7
1.9

2.7


2.0



1.0


1
e
-3
3

2
-0
1
4
<
«
B
4
4
4
4
3
t
1
1
0
t>
-0

.2



-1.4
.6
3.7
4. 1

8.S


5.2



2.2


2
4
-4
-3

21
13
11
10
11
9
12
B
G
4
«
Z
2
•
-1
-2
-2
-a

.0




1.3
2.9
4.1

5,7






2.4


3
3
e
-13

20
21
16
14
13
13
11.
7
3
3
2
0
-1
-3
-B
-6
-6
-0





1.2
2.2
4.4
0.9








2.6


4.
1
0
-»

29
24
31
22
10
14
»
6
2
3
-0
-2
-4
-7
-»
-»
-10
-6

-.6



I.Q
3.2
is. a
9.9

6.0

6.0




3.8
...

6
-e
-i
-3

20
2B
26
24
If!
14
0
a
2
0
-0
-5
-T
-11
-13
-13
-H
-0

-.»




4.2
6.9
6.9

7.4






3.0


6
-1
-a
i

29
27
£9
23
10
10
9
4
2
-0
-2
-6
»I2
-13
-IB
-16
-10
-11








v.4

7.4






3.6


T
-2
-3
B

31
20
37
21
17
13
0
4
1
-0
-B
-10
-14
-16
-18
-20
-22
-14
0 » IS 11 12
-.2 .2 .6 .8 .2

910


o.v 4.7 B.O 4.0- 4.2
7.9 7.0 7.3 6.7 6.2
B.4 7.1 7.0 6.3 6.2




7.9 7.7 6.3 4.0 2.9



4,3 4.u 4.6 *.2 a,i)

13
-.2




4.0
7.0.
6.2


] „





2.9
...
14
-.4
15







2.7 .e

7.4

3.9
2.7

2.4





(b} Wind Direction
0 9 19 11 12
-5 -6 -B -7 -6
-0 -D -0 -0 7
13 19 3f 42 53

37 40 44 62 44
33 43 47 40 84
34 01 29 22 16
19 15 12 6 0
15 12 0 2-3
12 9 6 0-0
0540-4
4 2 0-3-0
-0 -I -4 -7 -14
-1 -4 -18 -16 -21
-9 -13 -20 -2T -33
-14 -19 -26 -30 -33
-19 -23 -27 -31 -36
-2» -23 -27 -31 -33
-23 -23 -Zt -32 -36
-24 -20 '31 -35 -30
-26 -30 -34 -30 -41
-17 -20 -23 -28 -26
13
-0
29
63

4I
26
Iff
-|
-e
-9
-7
-10
-22
-30
-36
-39
-40
-39

-39
-41
-23
14
13
no
ea

34
24
13
-4
-11
-12
-11
-13
-22
-31
-36
-39
-41
-40

-39
-40
-24
C.9
6,n


3.3

2.0



3.6
3.6

15
6
20
B5

34
23
a
-9
-16
-17
-10
-20
-27
-34
-30
-40
-41
-41

-39
-aa
-23
16
-.3




-2. 1
2.7
3,<


2.3
3.3




3.6
2.9

16
g
7
30

30
20
4
-IB
-21
-24
-24
-27
-33
-37
-40
-43
-43
-42

-39
-37
-23
17
,1




-4.B
-.4
1.1
I.H

1 .7





2.7
2.7

17
-0
-3
17

21
10
-3
-27
-32
-33
-33
-33
-40
-43
-43
-46
-44
-43

-39
-07
-21
18
.3



-4.3
-3.3
-1.9
-.4
.a

- . t
.2




2.2
2.4

10
-2
-6
11

12
3
-6
-36
-42
-43
-44
-41
-46
-40
-49
-00
-47
-43

-39
-36
-20
19
1.0




-4.0
-2.0
-1.2
•'

-.0
-.3





1.9

19
-4
-0


11
9
-10
-39
-40
r-49
-49
-47
-Cl
-G6
-54
-03
-49
-46

-39
-33
-19
2d
1.3



-4.0
-3.0

-2.2
-.9

-.9
-.5




,U


20
-6
-9


13
7
-14
-41
-G3
-G4
-53
-49
-63
-61
-57
-C4
-BO
-46

-39
-34
-17
21
1.6



-3.3
-2.7
-2.6
-2, i
-1 .0

-1. i
-.7







21
-6
-0


17
0
-20
-40
-33
-06
-53
-D4
-CO
-66
-60
-33
-52
-46

-37
-31
-14
22
1.4



-3.0
-3.3

-1.9
-1 .9

-1.6
-l.i




-.7


23
-6
-S


10
-7
-211
-46
-60
-63
-64
-62
-67
-71
-63
-07
-01
-42

-34
-26
-11
23




-2.8

n -
-2.6
-2.3
_o •>
-1.6
-1.0




-1.5


23
-3
0


-2
-10
-37
-09
-63
-09
-74
-76
-77
-70
-72
-60
-DO
-39

-27
-20
-7
24
•*





-2.0
-3. I


-2.n
-2.3




-2.0
-.5

24
-2
"


-20
-37
-43
-66
-71
-00
-93
-9 I
-06
-no
-01
-6*
-46
-32

-16
-11
-n
23
-.0



-1. I
-2.0
-2.1
-!> -
-Z. J

-- 1
-2.0




— 1 .0
- . y

23
0
4
a

-13
-14
-1 I
-14
-13
-IB
-!•>
-ir
-13
-H.
-14
-11
-f.
-4
-a
-i
-u
0
*See Table 24 -for a  description of the experimental  conditions.

-------
                                                                        119
                                    Table 24
                 CONDITIONS REPRESENTED IN TABLES 16 THROUGH 23
                 Time
                                     D - D
0
               		r_T_    Grid Level
Case   Table    6 a.m.   3 p.m.    Eg. (26)  Eg. (27)   k = 1  k = 5
1
2
3
4
5
6
7
8
16
17
18
19
20
21
22
23
X
X
X



X

                       X
                       X
                       X
  X
  X


  X
  X
  X
  X
                                                         X
                                                         X


                                                         X
                                                         X
                                                                X
                                                                X
                                                                X
 Boundary
Conditions
    = 06
  X
  X
  X
  X
  X
  X
                                                                                X
                                                                                X
     In  addition to the general observations given above, we can also make the
following specific comments:
     >  Perturbations  obtained  at  ground  level  from  the  use of  Eq.  (26)
        to estimate  D  -  D^  are  larger  than  those  generated from the use
        of Eq.  (27)  (compare  Tables  16 and  17  and Tables  19 and 20).
        The opposite is  true  at the  top of  the  region  (compare Tables
        16 and  18  and  Tables  19 and  21).
     >  When Eq.  (27)  is  used to estimate D -  Dfl,'the  perturbations
        aloft are  much larger than those  predicted at  ground  level
        (compare Tables  17  and  18  and  Tables 20 and  21).
     >  The choice of  boundary  conditions employed does  have  some
        influence  on the  magnitude of  the predicted  changes in wind
        speed and  direction.  As expected,  this influence is  great-
        est near the boundary (compare Tables  18  and 22  and Tables
        21  and  23).

-------
                                                                         120
To explain the first two observations cited above, we note that the forcing
function in Eq. (23) follows the same pattern of behavior.  This can be illus
trated by considering the ratio of the forcing functions employed in each
case.  If we let AD26(K) and AD27(k,K) represent the values of D - DQ calcu-
lated using Eqs .  (26) and (27), respectively, then we can write the ratio of
the forcing function corresponding to the comparisons made above as follows:
                          AD27(1,K)
                          AD27(K,K)    2K
 where K is the number of vertical layers of grid cells.  In this study, K has
 a value of 5.

     The experience gained in this brief study indicates that the use of
 algorithms similar to those given in the previous section provides a viable
 means of producing mass-consistent wind fields.  Although such algorithms are
 relatively simple to employ,  they are deficient in the treatment of momentum
 and energy balance relationships.  However, until  complete planetary boundary
 layer models  suitable for predicting flow fields  over urban areas can be
 developed  and validated,  photochemical  modeling efforts will  undoubtedly
 continue to rely  on wind  fields  derived from actual  field measurements.
Thus, the  use of  mass-consistent wind algorithms  should be considered as
an interim means  for removing excessive convergence  and divergence ef-
fects in the  flow field.   The need for such usage may also be enhanced
by the inclusion  of wind  shear in the airshed model, since the extent of
convergence and divergence in the predicted flow  field aloft may be larger
than that  previously experienced near the ground.

-------
                                                                        121
     In future work, we recommend that mass-consistent wind algorithms be
employed in conjunction with interpolation procedures for predicting flow
fields over an urban area where a reasonably dense meteorological network
has been established.  In this way, tests can be designed to evaluate the
performance of the methodology.  The RAPS study in St. Louis may provide
such a data base.  In addition, further consideration should be given to
the manner in which the quantity D - DQ is estimated.  Examination of the
characteristics of flow fields over urban areas may provide some guidance
in this matter.
D.   ADOPTION OF AN IMPROVED ALGORITHM FOR ESTIMATING
     TURBULENT DIFFUSIVITIES
     Pollutants are dispersed through advection and turbulent diffusion.
 In the horizontal  directions, the advective mass flux is usually much
 larger than the diffusive flux.   However, vertical transport is often
 dominated by turbulent diffusion.  The usual  means for treating vertical
 diffusion is through the assumption that the turbulent mass flux, F., is
 proportional to the gradient of the mean concentration field.  That is,
                                   v  9z
where
        K   = turbulent diffusivity,
         = mean concentration.

Turbulent diffusivities are extremely difficult to measure in the field,
and their parameterization has been the subject of numerous studies.  Upon
reviewing the algorithm we employed in the 1969 validation study to calcu-
late K ,  discussed in Roth et al.  (1971),  we found that there was sufficient
justification to  formulate a new algorithm.   This  new algorithm includes
important atmospheric parameters,  heretofore omitted, that are known to have
a significant effect  on the value  of the diffusivity.  Specific criticisms
of the diffusivity algorithm that  we used  previously are as follows:

-------
                                                                        122
     >   The  diffusivity  is  assumed  to  depend  only  on  the  wind  speed.
        Using  measured diffusivity  data  reported by Hosier (1969),
        Eschenroeder et  al.  (1972)  found that the  diffusivity  does
        not  correlate well  with  wind  speed  alone.   This  finding  is
        not  surprising,'since  we would expect that, for  a given  wind
        speed, the  value of Ky for  stable atmospheric conditions  would
        be much less than its  value under unstable conditions.
        Clearly,  an algorithm  for K  must include  the effect of  at-
        mospheric stability.
     >   Surface roughness effects are  not explicitly  included  in  the
        formulation of K .   Recent  studies  by Lissaman (1973)  and
        Ragland (1973) indicate  that  ground-level  pollutant concen-
        trations  are significantly  influenced by the  value of  the
        surface roughness.

     In reviewing previous  efforts  to  parameterize the diffusivity reported
in the  literr.ture,  we found that guidelines appear to exist that  are  suffi-
ciently well  developed for  use in estimating  the value of K  in  the surface
layer (up to about  100 m).   However,  for the  remaining portion of the plane-
tary boundary  layer above an urban  area, we have not  found a definitive
treatment of the  diffusivity that is  both general  and simple enough to include
in an airshed  model. Also  of  concern  is the  objective for multiday simula-
tions of defining the vertical extent  of the  modeling region to  include  the
inversion layer,  if present.   The "trapping"  effect of the elevated tempera-
ture inversion would be  treated  through  the use of the vertical  diffusivity
profile.   Thus, a relatively sophisticated  treatment  of  K is  required aloft,
a region of  the.planetary boundary  layer where few measurements  are generally
available.

     Realizing that a completely satisfying treatment of KV may  not be attain-
able at the  present time, but  also  recognizing the need  to improve the algor-
ithm previously employed in the  SAI airshed model, we initiated  efforts  to
develop an algorithm for K   that includes,  at a minimum,  both  atmospheric

-------
                                                                         123
stability and surface  roughness  effects.   Several  schemes for computing K
have been proposed in  the literature^  including those described by Blackadar
(1962),  Wu (1965), Hino (1968),  Pandolfo  et al.  (1971),  Eschenroeder et al.
(1972),  Ragland  (1973), Bergstrom and  Viskanta  (1973), and Shir and Shieh
(1973).   However,  each of these  approaches is  to some extent heuristic, and
their validity is  somewhat uncertain.

     To  alleviate  the  difficulties associated with basing a diffusivity algor-
ithm on  field measurements, we developed  a methodology that uses the predic-
tions of a sophisticated numerical planetary boundary layer model  developed
by Deardorff.  Although the present K   algorithm is applicable for only
neutral  and slightly unstable atmospheric stability regimes, the methodology
can be extended  to other regimes.  For a  more  detailed discussion  of this
algorithm, we refer the reader to Chapter II of Volume III.

E.   MODIFIED TREATMENT OF THE INVERSION  LAYER
     IN  THE AIRSHED MODEL

     In  previous studies, the modeling region  has  been defined to  extend from
the ground level to the base of  an elevated temperature  inversion.   However,
a major  difficulty arises when using this approach for multiday simulations:
A significant amount of pollutants can be reintroduced into the mixed layer
from aloft as the  inversion is eroded  away during  each daytime period.   Unless
the pollutants that are trapped  in the inversion on the  previous day are re-
tained in the modeling region, it will  be difficult to account properly for
their reintroduction into the mixed layer on a  given day.  As an example of
the DO levels that have been observed  aloft, we present  in Figure  21 a  cross
section  of the pollutant distribution  in  a portion of the Los Angeles basin  on
the morning of 11  July 1973 (Jerskey et al., 1975).

     As  an alternative definition of the  modeling  region, we propose to include
the portion of the atmosphere bounded  below by  the terrain and bounded aloft by
the top  of the inversion layer.   All governing  equations and coordinate trans-
formations used  previously still  apply, except  that the  term AH should be
interpreted as

-------











1 	
QJ
>
01
— I
rtf
O)
OO

f
rtf
O)
t£-~t
QJ
>
O
•£3
c3-
j—
0
• r— •
•4-^
rcf
OJ

1 1 1












4SOO.
•J700.
4000.
4000.
4400.
4P,00.
4200.
4100.
4000.
3900.
OGOO.
3700.
3COO.
0000.
0-1-00.
3COO.
0200 .
3100.
3000.
2900.
2000.
2700.
2600.
2000.
2-: oo.
2COO.
2200.
2100.
2COO.
1900.
1000.
1700.
1600.
1500.
•I'iOO .
inoo.
1200.
1 100.
1COO.
900.
BOO.
700.
000.
000.
400.
300.
200.
100.
0
  ~"~-> 	 	. - __pjJ2	. nt: —,
  03
  05
  Ou
  07
  03
  07
  03
  10
  1 1
  11
  1 1
  11
  "16 '
  IV
  19
  .05
  .05
  .00
  .06
  .07
  .00
  .09
  .or
  .07
   09_
  . 12
  . 13
                            !0
                                   ->..
  29
  20
  26
   13
   1-V
   16
  20
  19
  10
      ^  —  ^0-14
        X /
      X X /
       1'
.03
.05
.04
.04
.00
.07
.07'
.07
.09
TTi 	
. 10
. 10
. 13
.09
. 10
. 10
. 10
.10
1 2
. 12
. 13
. 13
. 15
. 14
. 10
. 14
.[5 _
. 16
-^->. —

0., CONCENTRATIONS (ppm)
o





-. 	 . 	 	
~~o.fo — — — 	 __











^-- INVERS
0.10 
-------
                                                                        125
                        AH = Ht(x,y,t)  •-  h(x,y)
where
        Ht(x,y,t)  = elevation of the top of the inversion layer,
        h(x,y)     = terrain elevation.

The effect of trapping pollutants below the inversion layer can be accounted
for through the height dependence of the vertical  diffusivity.   Whereas  rela-
tively large values of K  are used in the mixed layer, the values in the
stable inversion layer are much smaller, reflecting the suppression of turbu-
lent mixing.

-------
                                                                       126
          IV   EVALUATION  OF ALTERNATIVE TECHNIQUES
      FOR  INTEGRATING THE SPECIES CONTINUITY EQUATIONS

                            James  P.  Meyer
A.   INTRODUCTION
     In essence,  the  SAI atmospheric photochemical  simulation program is
based on the solution of the nonlinear, multidimensional species transport
equation
                1 + v  • vc. = V •  Kvc.  + R.  + S.     ,             (29)
              9t
which, for convenience, has been transposed [Reynolds et al .  (1973)]  into
the form

           3(AHc.)
                  ++^^c.}  + »~(«c.)

                             8c             8c             8c
                          •H""3p / f 5F W + IF  (\M;
                                         + S.AH     .             (30)
In general,  no  closed-form analytical  solution  exists for this highly com-
plex partial  differential equation for all  possible  initial and boundary
conditions.   Hence, one is forced to resort to  approximation techniques,
most notably  finite difference schemes,  to  find a solution.

-------
                                                                             127
     The choice of an appropriate numerical  technique for inclusion in the
airshed model  involves two primary considerations.  First, the accuracy of
the solution obtained must be such that the  error in the predicted concen-
trations is predominately the result of errors in model  inputs rather than
errors introduced by the numerical technique itself.  Second, the final
choice between alternative techniques capable of solving the governing
equations to a specified error tolerance should be based on minimizing com-
puting costs.   In view of these considerations and the variety of numerical
techniques available for solving the equations of interest, care must be
taken to choose a method that offers an optimal blend of numerical accuracy
and computational efficiency.

     Currently, a finite difference approach termed the  method of fractional
steps [Yanenko (1969)] is employed in the SAI model.  The basic feature of
this method is that the  four-dimensional governing equation in (<;,n,p,T)  is
split into three two-dimensional equations in (?,T), (n,T), and (P,T).  The
details of this analysis are given elsewhere [Reynolds et al. (1973)].  With
this type of approach, errors are introduced into the solution in the follow-
ing ways:
     >  Through  the  introduction  of truncation  errors  caused by the
        finite  differencing  of the partial  derivatives in the trans-
        port equation.
     >  Through  the  decomposition of a  three-dimensional  equation
        into a  sequence  of three  two-dimensional  equations.

     As  an  example of truncation  error  effects, Harlow and Amsden  (1970)
showed that for  the  one-dimensional  advection equation
                                                                         (3D

-------
                                                                              128
a numerical  solution involving a first-order finite difference approximation
introduces an error on the order of
                                                                         (32)
                                       2    ?
into the calculation.  Since the term 3 c/9x  appears, this error has been
called "numerical  diffusion."  Its effect is to smooth an initially peaked
distribution over a large portion of the modeling area.  This reduces the
resolution of the solution so much that in extreme cases it becomes nonex-
istent.  In an attempt to reduce the truncation error effects in the SAI
model arising from the treatment of the horizontal advection terms, we pre-
viously carried out numerical experiments using various second- and fourth-
order difference approximations.  Although we observed some reduction in
truncation error using the higher order methods, many of these techniques
also had the undesirable property of producing negative concentrations in
the vicinity of steep concentration gradients.  We finally selected an
uncentered second-order method described by Price et al . (1966), which is
somewhat more accurate than the first-order advection approximations, and
which, at the same time, presented no difficulties with regard to the pre-
diction of negative concentrations in the initial application of the model
to the Los Angeles basin.

     In the fractional step technique,  the decomposition process for the
n,c, and p directions introduces a sequence of higher order partial deriv-
atives that would not normally appear in the transport equation.  Although
the effect of these  terms  is difficult to quantify a priori, their impact
on model  predictions can be examined by comparing, predicted pollutant con-
centrations with known analytical solutions of the governing equations.
In the discussion presented in Volume I of the validity of the grid and
trajectory model concepts, we noted that errors introduced into the grid
model  predictions  by the numerical integration technique can be as  large as
50 percent  in some situations.  These errors are mainly due to the finite

-------
                                                                           129
difference treatment of the horizontal  advection terms.   Thus, the objec-
tive of the present study is to test and assess various  alternative numer-
ical approximations to the transport terms in the governing equations.   In
this analysis,  our aim is to provide recommendations regarding the course
of future efforts to improve the numerical integration procedure employed
in the airshed  model.

     In the work described next, the emphasis was on the development of an
analytical solution to the diffusion equation and on a comparison of the
analytical results with the corresponding results obtained from various ap-
proximate integration  schemes.   Because of the difficulties involved in
developing solutions to the diffusion equation, only a simplified one-
dimensional, linear, time-dependent result could be obtained.   Thus, we
are able to assess the errors associated with various numerical  methods
for a one-dimensional  flow problem in which the pollutant is allowed to
undergo a first-order  chemical  reaction.  Clearly, this  test situation  is
not completely  representative of a full  photochemical airshed simulation.
However, numerical techniques incapable of producing sufficiently accurate
results in a one-dimensional linear problem cannot be expected to perform
better in a multidimensional nonlinear application.

     Since it was not  possible to carry out the tests for photochemical
pollutants, we  are unable to assess the effect of inaccuracies introduced
in the treatment of the transport terms on error propagation, especially
when nonlinear  chemical interactions are taking place.  In addition, the
test results do not illustrate the errors caused by using the fractional
step methodology to treat a multidimensional  problem.  In spite of these
limitations, however,  we have been able to delineate two numerical methods
that seem to represent a significant improvement over the finite difference
scheme currently employed in the SAI airshed  model.

-------
                                                                            130
B.    AVAILABLE  METHODS

     A variety  of methods  have  been  developed to  solve partial  differential
equations.   In  general,  they fall  into  two  categories:  finite  difference
schemes and particle  techniques.   The former, which  are well  developed,  in-
clude the work  of Price  et al .  (1966),  Fromm (1969), Crowley  (1968),  and,
more recently,  Boris  and Book (1973).   In  contrast,  particle-in-cell  methods
are relatively  current;  they include the contributions of Sklarew et  al.
(1971) and Egan and Mahoney (1972).

     For the purpose  of  analysis,  a  simplified solution of the  diffusion
equation was developed and the  results  of  this calculation were compared to
the results of  the suitably programmed  approximation schemes.   The equation
chosen for this work  was the one-dimensional  transport equation,
                                2
                        8C _ n 8 C     8C   ,
                        — r - U - ^ -  U — -  KC    ,
                        3t       2     3X
in which u, D,  and k were considered constant.   The following  boundary  con-
ditions were imposed:

     >  Initially, no material  is  in the modeling region,  i.e.

                             c(05x)  = 0    .                              (34)

     >  There is zero concentration  gradient  of infinity,  i.e.,

                           ff (t,-)  = 0    •                              (35)

     >  There is a uniform concentration at the inlet, i.e.,

                             c(t,0)  - 1     .           .                   (36)

-------
                                                                             131
We evaluated the following methods:

     >  Price scheme
     >  Crowley second- and fourth-order methods
     >  SHASTA method
     >  Galerkin method
     >  Particle-in-cell techniques
     >  Egan and Mahoney method.

In'the following subsections we briefly describe each method.

1.   The Price Scheme

     Currently, the SAI model uses a method proposed by Price, Varga, and
Warren (19661).  For the test problem selected, this method has the finite
difference form
  cn+1 = cn +        c?,, - 2cn + cn .  - ^ (3cn. - 4cn . + cn    - k5tcn
   J      J       2 V j+1     j    j-}    ^6X    j     j-1    j-
                                                                          (37)
This approximation has errors that are first order in time (fit) and second
                      2
order in distance (6x) .

     Since the solution is explicit in time, definite limits of stability
exist.  These limits can be developed by assuming that the solution of the
transient equation can be written in the complex.Fourier form

                      c(t.jAx) = *(t) eijAx     ,                          (38)

where
                               i = /T    ,                               (39)

-------
                                                                              132
After considerable algebraic manipulation and stipulation of the requirement
                               At)
                                    < 1
                                                      (40)
we obtain
         k6t < 2
                                                                         (41)


                                                                         (42)
                        26X    H
                                 (fix)1
                         4   2
                                                                         (43)
as the conditions required for stability to occur.


2.   The Growley Second- and Fourth-Order_Methods


     To enhance the accuracy of the finite difference approximations  used in
formulating the advective terms of the governing equations,  Crowley (1965)
developed both  second- and fourth-order centered difference  algorithms  for
these terms.   For the test problem, the second-order method  has  the expansion
          •r
  a ,  a
~ 2   2
c"   + (1  - 23 - a2 - k6t)c"
 J '    \           ..     /  J
                                         a
                                         2
                                             a"   n
where
                                  Ku6t
                                   n
                              a -
               U6t
               6X
                                                      (44)
                                                                         (45)
                                                                         (46)

-------
                                                                     133

The corresponding fourth-order expansion has the form
 n+1 _ /_a   
-------
                                                                             134
where
                      J- + u!/2  —
                           	»     j=l,...,n     ,          (49)
                               l/2
                            -  UJ
and where u.    refers to the velocity at the j-th location at time t + (fit/2)
           J
Completion of the antidnffusion step requires the expression
                 n+1  _  ~n+l    l/~n+l    ??n+l    ~n+l\     .                 (50)
                Cj    "  Cj    "    c    * ^C
     To account for cases in which material  may be advected either into or
out of the modeling region, the SHASTA method applies the following rules at
the end points:
     >  Left-hand side
        --If v-j > 0, then CQ and vfi, the upwind boundary conditions,
          must be specified.
        --If v, < 0, then 3c/5x = 0, and CQ = c, and VQ must be specified.
     >  Right-hand side
        --If vn > 0, then 8c/3x = 0, and c .,  - c  and v ., must be specified.
        --If vn < 0, then cn+^  and vn+-j , the incoming concentration and velocity,
          respectively, must be specified.

Since the SHASTA algorithm treats only the advective parts  of the continuity
equation,  the concurrent diffusion and kinetic steps of the governing equation
must be treated as subsequent operations.   .Hence, the system heavily relies on
the method of fractional  steps.

-------
                                                                            135
     For the  test  case,  the  advective  equation has the form
                                                                         (52)
        *n+l  _  ~*n+l   l/~*n+l     -*n+l    ~*n+l\
        Cj    '  Cj     " 8
The diffusive  and  kinetic  terms  become

                     +  (1  -  23   -  kfit)  S     +  3         ,                  (54)
               K,,,6t
           3  -    -      .                                                 (55)
4.    The Galerkin  Method

     Finite  element  methods  represent  a  significant departure  from finite dif-
ference techniques as  a tool  in  solving  partial  differential equations.   Unlike
finite  difference  equations,  which  approximate  derivatives  at  specific  loca-
tions,  finite  element  techniques  approximate  functions  over an entire  domain
[Zienkiewicz (1971), Finder  and  Gray  (1974)J.

     To develop  finite element solutions,  one must  follow  four steps:

     >   Subdivide  the  domain  of  interest into a  finite  number  of elements
        defined  by node points.
     >   Approximate  the dependent variables  in  terms  of their  unknown  node
        point  values within  each  element.  This  insures the continuity  of
        the  dependent  variable across  the  element.

-------
                                                                              136
     >  Minimize an appropriate measure of error such that a  set of
        simultaneous equations results.
     >  Solve the resulting set of equations for the node point values.

     The distinct advantages of finite element techniques are their ability
to model arbitrary geometric areas without a loss of convergence and  their
generally greater stability compared with corresponding finite difference
systems.

     Two distinct classes of finite element solutions exist.  The Rayleigh-
Ritz procedure requires the minimization of a function associated with  a de-
fining differential equation.  Although this is an extremely  useful method,
often one cannot determine the functional form associated with the differential
equation.  Thus, the method has limited applicability.  A more general, but
somewhat less mathematically elegant, approach is the Galerkin technique
[Keldysh (1964), McMichael and Thomas (1973)].  This method requires  merely
that the integral of the approximate solution be orthogonal to each of  the
basis functions spanning the solution space.  For example, the linear differ-
ential equation

                     O C ,    oCr^oC.i     r\                          f r c \
                     ni7_^ „ -I- I I  ,     I „_ _ -m	i _L I/ f~* — I 1                          I ^ r"i I
                        i  LJ    "" \J   ~f*r ~ |\ L*   w                          \*-^^/

is written in operator form as
and is assumed to have a solution of the form
                            c =  E  a.cf,,     ,                             (58)
where there are n nodes in the domain of c.  Then, the'Galerkin  procedure
requires that

-------
                                                                             137
               J  U6H.
dv = 0    ,    . i  = 1, 2, ..., n    ,             (59)
 so  that the coefficients  (a.) can be determined.

     As an illustration of the technique, consider  the  following  sample
 problem:
                                                                          (60)
 Assume  a  solution  of the form
                              n
                         c =  E  a.(t)  6.(x)     .                         (61)
 Multiply  the differential equation by ., and  integrate the result over  its
 entire  domain:
  .L  A              L  ^
    |f  *. dx + u  I   |£ f. dx -  D I  ^ *_. dx + |   kc*, dx -
 fo   3t  i        JQ  9x   i        JQ
                                                      i=l,  ..., n    .      (62)

 By  using Green's  theorem,
                                   3x
                                                 (63)
                              0
and by substituting  the  expanded  series  into the equation, we obtain

-------
                                                                            138
 f
n aa.
£ -rjr1- 4>. dx + u  I   I a.
 . at  Ti         I   .  i
                                dx
      I n
dx + ki i a ..d>. dx
                                                                0
                                                                  J
                   = D . Ea.
                       1   J
                                             1=1,2,  ....  n    .    (64)
Once the functions $. are selected, a series of matrix equations  result.
They are of the form
                               da
                             ~ dT + ~~ = ~
                                                                     (65)
where
                       £i, B  =  coefficient matrices,

                       a, S  -  column vectors.

These equations can be easily solved by using a Crank-Nicholson technique  to
approximate a between times t and t + St.  The initial conditions an must  be
specified.

     For the work described in this report, we selected chapeau functions  of
the form
                       "-7-    ,    0 < x < x,
                        X]                    I
                                 elsewhere
                                                                          (66)
                       v	v	   '    Xi 1 - X - X1
                       A •  - X •  T            I ~ I         I
                                                                          (67)
                        ,,,^T    '    xi
-------
                         - x
                             n-
                       xn -  xn-l
                  X  , < X < X
                   n-1 -   -  n
                       0     ,     elsewhere
Correspondingly,  the matrices A and B  had  the  tridiagonal  form
                                                                               139
                                                                            (68)
A = ^
                  1



                  0


                  0
4



1



0
1



4


1



0
                        0


                        1
0


0


1


1
     0



     0



...   0



 4     1
                                                 1 < i < n - 1     (69)
where
          B -
                -a



                0
                      -a
               D   ,  u    k6x

              H   2  "  ~r
          0



          -Y
                                  -a
                0


                0
                                             -Y
                                                     (70)
            -  2D    2K5X

              6x      3
            _  D     u

          Y  "  £T-  2~ ~2
while S was  zero  everywhere.   The  boundary condition yielded the terms
                                                                             (71)
                            an  =  an-l
                                                                             (72)

-------
                                                                              140
 5.   Particle-in-Cell  Techniques

     Harlow and Welch  (1965) first developed particle-in-cell methods for
 use in the analysis of free-surface fluid mechanical problems.  Since their
 initial development, these techniques have been expanded to include such
 variations as marker-in-cell (MAC) and HYDRO codes.

     An interesting adaptation of the particle-in-cel1 algorithm has been
 developed by Sklarew (1971) to mode] mesoscale air pollution problems.  In
 this variant, pollutant particles representing a fixed weight of material
 are generated in quantities proportional to the ambient pollutant concen-
 tration.   As time passes, the particle positions are tracked in space by
 determining the incremental changes in their locations caused by advective
 and diffusive forces.   Sklarew chose to rewrite the species transport equa-
 tion in the form

                      !§•+ v • (vc - Dvc) - 0    ,                       (73)
                      0 t        ~

 where

            v  =  mean velocity,
           vc
          D—  =  diffusive velocity.
With these definitions, it is possible to increment the radial position of
each particle during each time step by a corresponding contribution due to
mean fluid flow (v6t)  and diffusional  motion [(Dvc5t)/cJ.

     To account for photochemistry, one must assume that, within each cell,
the particle weights can be summed to  form a representative cell concentra-
tion and that the reaction occurs as if the material is homogeneously dis-
tributed throughout the cell.  At the  end of the reaction sequence, each
particle is  reweighted proportionally  to the change in the cell concentra-
tion of the  individual  species:

-------
                                                                             141
                     m.(t * At)  = m.(t)  -c(t)        >                   (74)

and the transport process is subsequently allowed to occur.

     Hotchkiss and Hirt (1972)  improved  the modeling of the  diffusional  part
of the transport process  at Los  Alamos.   Their  contribution  was the represen-
tation of the diffusive movement as a random particle motion of the form

                        6xDIFF  = /4D6t $     ,                            (75)

where ^ is a randomly distributed Gaussian variable.  Their  work indicates
that their method results in substantially better agreement  than the method
of Sklarew in areas of strong concentration gradients near point sources.
This modification overcomes the  difficulties in computing the finite differ-
ence approximation needed by Sklarew in  calculating the gradient of the  con-
centration.

     Fundamentally, the problem  with all  particle-in-cell methods is the
essential  question of exactly what a particle represents and over what area
should it be considered to have  domain—classically an  Eulerian-Lagrangian
paradox.   A recurrent problem in using this type of analysis is the back-
ground noise that must be accommodated when a particle  leaves one cell and
enters another.   This quantum jump can be smoothed to some extent by volume-
averaging  the particle over the  adjacent  cells  it intercepts.  However,  such
a procedure  may  well  extend the  domain of a pollutant into regions that  it
does not  actually represent.  To circumvent this problem, one can always in-
crease the number of particles associated with  a problem, but at the added
expense of dramatically increasing computer storage and computational  time.

6.   The  Meth_od_ of Egan and Mahoney

     One  of the  more interesting developments in the analysis  used  in air
pollution modeling has been the  work of  Egan and Mahoney  (1970,  1971, 1972).
In essence, their approach is to follow  air parcels as they move within a

-------
                                                                            142
grid network,  taking  into  account  the  zero,  first,  and second moments of
the pollutant -distribution.   With  this  type  of analysis,  it is possible
to maintain  extremely high resolution  and  to eliminate almost entirely
the numerical  diffusion  caused  by  errors associated with  approximations
for the advection  terms.

     Unfortunately, the  method,  which  is owned proprietarily by Environmental
Research and Technology  in Lexington,  Massachusetts, is only paraphrased
in the open  literature.  Hence,  the  analysis presented here is cursory and
represents only a  superficial evaluation of  the utility of this method.

C.   A TEST  PROBLEM

     To provide a  common basis  of  comparison for each of  the methods, we
posed the following two-dimensional  problem  (x - t) .

     Consider  a semi-infinite strip  extending, from  zero to infinity over
which the species  transport equation is assumed to  hold and a first-order
irreversible reaction occurs:
                                      2
                          ,    8C    n  8 C   ,                              i-,r\
                          +  u  — =  D  — ~ -  kc    .                        (76)
Specify that all  parameters  (u,  D,  and  k)  are constant,  and impose the fol-
lowing boundary conditions:

     >  Initially no  material  is  in the modeling region, i.e.,

                     c(0,x)  =0     ,     0  <  x < -    .                    (77)

     >  There  is  zero flux of  infinity, i.e.,

                              (t,-) = °    •                              (78)

-------
                                                                              143
     >  There is a uniform concentration at the inlet, i.e.,



                             c(t,0)  = 1     .                              (79)



     To develop a solution, take the Laplace  transform of the defining

differential equation


                                       2-
                        sc+ u {j|«  D^f- kE    .                       (80)
                                      dx


Then rearrange the equation in the form

                         2-
                      D^f - u £- (k + s)c = 0    .                    (81)
                        dx^     dx


Next, solve for c:


                  E = Ae[(Pe/2) - *JA + Be[(Pe/2) +  *Jx    ^             (82)



where


                          Pe = ^    ,                                   (83)



                                    ,                                    (84)
                                                                         (85)
By imposing the boundary condition



                         --=0    at x = -    ,                          (86)


we find that

                                 B = 0                                   (87)

-------
                                                                              144
and, hence,
                         -c . fle[(Pe/2) - TJ.    _                         (88)

At the leading edge of the system,

                        c=j=A    at x = 0    ,                        (89)

and the complete solution in transform space becomes

                             [(Pe/2) - 
-------
                                                                             145
Instead of using  x  -  t  space, which  involves the derivative of integrals,
it is  simpler  to  compute  the derivative  in x - s space and invert the ob-
tained transform.
     The inversion  of the  derivative
                  c (0 s) = -
                 dx IU'SJ   L
                                                    s)
   2s
is given by
If (t'0) =^F
oX         Lu
                            ^m
                                      +Jjferf(nt)|
                                             (95)
                                             (96)
where
Consequently,  the  total  flux
N(0,t)  - uc - D
                                            9X
is represented  by
                                                                        (97)
                                                                        (98)
                  N(0,t)  -
                                             (99)
for a uniform concentration  of one at the origin.
     During  the  time  interval  t  to t + at, the amount of material entering
the first  cell ,
                       ±+5t
                             l(0,t')'dt  -  Q    ,
                                             (100)

-------
                                                                              146
can be approximated by


                        QO Lx ,  f U/
                    = u T +^o
                                    [erf (/n(t + 6t))+ erf(^t)]     .      (101)
Note that at long times the inflow approaches the quantity

                               /  .      \
                           Q - (^ + ,/Dn)6t    ;                          (102)


and if k = 0,

                               Q = u6t    ,                              (103)


and pure advection occurs.  One effect of having the reactive term is to
enhance the inflow above the purely advective amount.

     In Section D, we present figures in  which the analytical solution is
always represented by continuous  curves.

D.   RESULTS

     To test each method under conditions similar to those encountered in
atmospheric modeling, we decided  to allow the Peclet number (uL/D) and the
kinetic rate constant to vary over a wide range of values.  For each run,
the incremental spatial distance  was set  at 2 miles, and the total length
of the region was assumed to be 50 miles.  Each hour was subdivided fnto
12 five-minute segments.  In all  cases, the free-stream velocity was held
at 4 miles per hour, and the diffusivity  was allowed to vary as shown in
Table 25.

-------
                                                                           147
                                Table  25
                VALUES  OF  DIFFUSIVITY  AND  PECLET  NUMBER
                         FOR  THREE  CASE STUDIES
                               Diffusivity
                                  2    -1
              Case               (m sec  )     Pec let  Number
                1                  200            720

                2                  700            206

                3                  2000            72
These runs  varied  from almost  a  square  wave  propagation  (Pec  -  720)  to  a
smooth diffusion problem (Pec  =  721).   For each  method  tested, we  executed
a series  of 12  runs.

     The  corresponding kinetic values  associated with  these transport con-
ditions are given  in  Table  26.
                                Table  26
                    KINETIC  CONSTANTS  FOR  EACH  CASE

                                   Kinetic Constant
                    Case                (sec"  )
                      1                    0

                      2                 10"4

                      3                 10"3

                      4                2 x  10~3

-------
                                                                              148
     These values include the cases of both no reaction (k = 0),  and a
     tively fast react'
were also considered.
                                     o
relatively fast reaction  (k  =  2  x 10~  ).   Two intermediate reaction rates
     In the presentation of the data, we included only those cases in
which no reaction occurs (k = 0; Pec = 720, 206, and 72) and those of
highest Peclet number (Pec = 720, k = 10~4 and 10"3).  We chose these
cases because they are somewhat representative of the range of conditions
that can occur in mesoscale modeling systems.  The following subsections
present a brief synopsis of the performance of each numerical scheme.  In
each figure presenting our results, the analytical solution is given at
3,  6 and 9 hours from the start of the test.

1•   The Price Method

     As used in the SAI model, the Price scheme is inadequate for accur-
ately modeling mesoscale phenomena.  In all cases, the method overpredicts
the actual ground-level concentration and transposes the wave to the left
because of phase shift, as shown in Figures 22 through 23.  Although some
improvement occurs in cases having high Peclet numbers, this agreement is
not substantial enough to reduce dramatically the errors involved.  Thus,
we  rated the method as poor.

2.   The Cro_wT_ey Second- and Fourth-Order Methods

     The accuracy of prediction can be substantially increased by using
either the Crowley second-order or the Crowley fourth-order approximation,
as shown in Figures 24 through 25.   In cases where an extremely strong con-
centration gradient appears (Pec = 720), the second-order scheme exhibits
some rather erratic results near the top of the wave.  Aside from such
cases,  both methods provide essentially the same results.

     In  the implementation  of these methods in actual  simulation programs,
some observers  have noticed that higher order methods occasionally predict
negative concentrations  in  regions  having large concentration gradients.

-------
                                                                             149
Although this result did not appear in our work, one should keep it in mind
as a limitation when using these methods.

3.   The SHASTA Method

     One of the simplest and yet most efficient methods of solving the spe-
cies transport equation is the SHASTA method.   Figure 26 presents its  per-
formance results.   Not only does the method exhibit a relatively high  degree
of accuracy, but also, unlike many of the  alternative finite difference
methods, it never predicts negative results.  Thus, it is the best choice
available of an explicit solution algorithm.

4.   The Galerkin Method

     Of all the methods tested, the Galerkin technique provided the most
accurate results over the widest range of  conditions selected in this  study,
as shown in Figure 27.  The predicted results, were always within 1  percent
of the analytical  solution, and for many individual points in the analysis,
the results exhibited zero error.  Not only could the method be used to model
situations in which extremely strong concentration gradients appeared, but
also it could accurately treat cases involving very fast reaction schemes.
Although the method is implicit and hence  iterative in solution, its execu-
tion time appears to be comparable to a corresponding implicit Price scheme
as currently used on the SAI model.

     We thus recommend that this technique be  used for cases where high
resolution is desirable, even at the expense of increased computing time
and programming effort.

5.   Particle-In-Cell Methods

     Accuracy in particle-in-cell methods  is a strong function of the  number
of particles used.   In this study, as the  particle size was reduced from 80
to 40 to 20 weight units, the average error was reduced from 9.6 to 6.3 to

-------
                                                                            150
3.3 percent,  respectively.   Figure 28 presents the results obtained using
this type of  method.   These methods proved successful  at simulating both
reactive and  nonreactive systems,  and they were able to treat steep gra-
dients exceptionally  well.

     A rather interesting aspect of this analysis is that such techniques
show greater  accuracy at lower rather than higher Peclet numbers, as would
normally be expected  to occur.  The reason for this phenomenon is probably
the following:   As the diffusivity is increased, the diffusive component
in the Hotchkiss-Hirt analysis displaces the particle by an amount propor-
tional to the square  root of the diffusivity.   For large values of the dif-
fusion coefficient, this displacement can extend well  over several cells.
Hence, the method is  best applied  to those cases in which the diffusivity
                  2    -1
is less than  200 m  sec  .

     One drawback of  particle-in-cell methods  is the amount of computing
time required to solve a particular problem for a given accuracy.  Since
a random number must  be generated  for each particle at each step, comput-
ing costs can be exorbitant as the number of particles increases.

6.   The Method of Egan and Mahoney

     For the  strictly advective case, the Egan and Mahoney method gener-
ates an extremely accurate solution with virtually no error attributable
to numerical  diffusion.  This accuracy is particularly notable because
advective phenomena have been extremely difficult to simulate using com-
puting methods.  Figure 29 presents the results obtained using this method.
Unfortunately,  we could not incorporate the diffusion step in this analysis
because of the  absence of any clear explanation of the treatment of this
process in the  open literature.  Without this  link, it is difficult to form
an overall critical appraisal of the technique.  In light of this limita-
tion, this method should continue  to be investigated as more material
becomes available.

-------
               " u—cr
O-
o.
 I
 I
n3
S-
-P
c
O)
o
E
O
o
         0.8
         0.6
         0.4
         0.2
                              o
                        10           20            30

                                 Downwind Distance—miles


                                 (a)  Pec = 720, k = 0 sec
                                                            o
-1
       40
50
                    FIGURE 22.   CONCENTRATION AS A FUNCTION OF DOWNWIND
                         DISTANCE  FOR THE EXPLICIT PRICE SCHEME

-------
          1.0
CL
Q.
 I
 I
fO
i-
-p

53
u
E
•O
                                       20            30


                                  Downwind Distance—miles


                                  (b) Pec = 206, k  = 0 sec
      40
-1
                   FIGURE 22.   CONCENTRATION AS A FUNCTION OF  DOWNWIND

                    DISTANCE FOR THE EXPLICIT PRICE SCHEME (Continued)
50
                                                                                                       en
                                                                                                       rv>

-------
         l.O
         0.8 -
         0.6 -
Q.
CL
 i
4J
C
cu
u
c:
o
0.4 -
         0.2 -
                         10           20            30            40


                                  Downwind Distance—miles


                                  (c) Pec = 72, k = 0 sec"1


                     FIGURE 22.   CONCENTRATION AS A FUNCTION OF  DOWNWIND
                      DISTANCE FOR THE EXPLICIT PRICE SCHEME (Continued)
                                                                     50
                                                                                                         tn
                                                                                                         OJ

-------
        1.0-
Q-
O-
 I

        0.8
        0.6
        0.4
        0.2
                        10           20            30

                                Downwind Distance—miles

                                (d) Pec = 720, k = 10"4 sec'1
40
50
                   FIGURE 22.   CONCENTRATION AS A FUNCTION  OF  DOWNWIND
                    DISTANCE FOR THE EXPLICIT PRICE SCHEME  (Continued)

-------
         1.0 e
         0.8
         0.6
CL
CL
         0.4
O>
O
O
O
         0.2
                         1234

                                 Downwind Distance—miles

                                 (e) Pec = 720, k - TO'3 sec"1


                  FIGURE 22.  CONCENTRATION AS A FUNCTION  OF  DOWNWIND
                   DISTANCE FOR THE EXPLICIT PRICE SCHEME  (Concluded)
                                                                                                     en
                                                                                                     en

-------
   7.0,  o  O
   0.8
   0.6
CO
S-
0)
0
c
o

0  Q.4
   Q.2
                      o
o
                  10           20             30
                  /

                          Downwind Distance—miles


                          (a)  Pec = 720, k = 0 sec"-1
                                    40
50
       FIGURE 23.  CONCENTRATION AS A FUNCTION OF DOWNWIND  DISTANCE
                          FOR THE IMPLICIT PRICE SCHEME
                                                                                                       en
                                                                                                       cr>

-------
   1,0
   0.8
   0.6
ra
S-
-*->
c:
01
o


8  0.4
   0.2
                  10           20           30

                         Downwind Distance—miles

                         (b) Pec = 206, k = 0 sec
        40
-1
50
       FIGURE 23.  CONCENTRATION AS A FUNCTION OF  DOWNWIND  DISTANCE

                   FOR THE IMPLICIT PRICE SCHEME  (Continued)

-------
   1.0
   0.8
   0.6
(O
s_
CD
O
O
O
   0.4
   0.2
                   10
20
30
40
                         Downwind Distance—miles

                         (c) Pec = 72, k = 0 sec"
       FIGURE  23.   CONCENTRATION AS A FUNCTION OF DOWNWIND  DISTANCE
                   FOR THE IMPLICIT PRICE SCHEME (Continued)
50
                                                                                                     en
                                                                                                     OD

-------
   1.0
   0.8
   0.6
4-5
tO
s-
4J
C
O)
o
c:
o
   0.4
   0.2
                   10
20
30
40
'  50
                         Downwind  Distance—miles

                         (d)  Pec = 720,  k  =  10~4 sec"1
       FIGURE 23.  CONCENTRATION  AS  A  FUNCTION OF DOWNWIND DISTANCE
                   FOR THE  IMPLICIT  PRICE SCHEME (Continued)
                                                                                                      en

-------
                  Downwind Distance—miles

                  (e) Pec = 720, k = 10~3 sec'1
FIGURE 23.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
            FOR THE IMPLICIT PRICE SCHEME (Concluded)
                                                                                             CTl
                                                                                             O

-------
       10
   20

(a)  Pec
30
40
50
                        = 720, k = 0 sec
                                        .-1
FIGURE 24.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
             FOR THE CROWLEY SECOND-ORDER SCHEME

-------
1,0
   0.8 -
O)
o
o
O
   0.4
    0.2
                                     ^ O
                  10           20           30

                          Downwind Distance—miles

                          (b)Pec = 206, k = 0 sec"1
                                                  40
50
          FIGURE 24.   CONCENTRATION AS A FUNCTION OF  DOWNWIND DISTANCE
                 FOR THE CROWLEY SECOND-ORDER SCHEME  (Continued)
                                                                                                       en
                                                                                                       ro

-------
          10
20
30
                Downwind Distance—miles
                (c) Pec = 72, k = 0 sec
                                       -1
40
50
FIGURE 24.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
       FOR THE CROWLEY SECOND-ORDER SCHEME (Continued)
                                                                                           CTi
                                                                                           CO

-------
1.0
                             20
30
                      Downwind Distance—miles
                      (d)  Pec = 720, k = TO"4 sec"1
40
50
      FIGURE 24.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
             FOR  THE CROWLEY SECOND-ORDER SCHEME (Continued)
                                                                                                  CTi

-------
   1.0
  0.8
   0.6
O)
o
   0.4
   0.2
                                2             3

                        Downwind  Distance—miles

                        (e)  Pec =  720,  k  = 10"3 sec"1
        FIGURE 24.  CONCENTRATION AS A  FUNCTION  OF  DOWNWIND DISTANCE
               FOR THE CROWLEY  SECOND-ORDER SCHEME  (Concluded)

-------
   1.0
                                            A
   0.8
c  0,6
o
fO
o
c.
o
   0,4
   Q.Z
                          A
                  10            20             30

                         Downwind Distance—miles

                         (a) Pec = 720,  k  = 0 sec
        40
-1
50
       FIGURE 25.  CONCENTRATION AS A  FUNCTION  OF  DOWNWIND DISTANCE
                       FOR THE  CROWLEY FOURTH-ORDER SCHEME

-------
   1,0
   0.8
c:
o
   0.6
CD
CJ
O
o
   0.4
   0.2
                               20
30
                         (b)  Pec = 206, k = 0 sec
                                                 -1
40
50
      FIGURE 25.   CONCENTRATION AS A FUNCTION OF DOWNWIND  DISTANCE

                FOR  THE  CROWLEY FOURTH-ORDER SCHEME  (Continued)
                                                                                                      CTi

-------
   1.0
  0.8
o  0.6
(T3
i-

-------
   7.0
   0.8
   0.6
ra
s~
0)
o
O

O
   0.4
   0.2
                                20
30
40
                        Downwind  Distance—miles

                        (d)  Pec  = 720,  k = 10~4 sec"1
50-
     FIGURE 25.  CONCENTRATION  AS  A FUNCTION OF DOWNWIND DISTANCE

           FOR THE  CROWLEY  FOURTH-ORDER SCHEME (Continued)
                                                                                                   01

-------
  1.0
  0.8
  0.6
fO
s_
OJ
u
  0.4
  0.2
                  _L
I
                  1            2'3             4

                       Downwind Distance—miles

       (e) Pec = 720, k = 10~3 sec"1,  6t  =  300  sec,  6x =  2 miles
     FIGURE 25.  CONCENTRATION AS A FUNCTION OF DOWNWIND  DISTANCE
           FOR THE CROWLEY FOURTH-ORDER SCHEME  (Concluded)

-------
           10           20             30
                  Downwind Distance—miles
                  (a) Pec = 720, k = 0 sec
         40
50
-1
FIGURE 26.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
                     FOR THE SHASTA METHOD

-------
           10
20
30
                  (b) Pec =206, k = 0 sec
                                          -1
40
50
FIGURE 26.  CONCENTRATION AS A FUNCTION OF  DOWNWIND  DISTANCE
               FOR THE SHASTA METHOD  (Continued)
                                                                                              IX)

-------
1.0
                                         30
40
50
                     Downwind Distance—miles
                     (c) Pec = 72, k = 0 sec'1
    FIGURE 26.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
                  FOR THE SHASTA METHOD (Continued)
                                                                                               CO

-------
   1.0,
O)
o
O
O
                   10
20
30
                         Downwind  Distance—miles

                         (d)  Pec = 720,  k  =  10~4 sec"1
40
50
       FIGURE 26.   CONCENTRATION  AS  A  FUNCTION OF  DOWNWIND DISTANCE
                      FOR THE SHASTA METHOD  (Continued)

-------
1.0
                12            3             4
                     Downwind Distance—miles

      (e) Pec = 720, k = 10"3 sec"1, st = 300 sec, 6x = 2 miles
    FIGURE 26.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
                  FOR THE SHASTA METHOD (Concluded)

-------
   7.0
        -e	a a D D
                        a a
a on a  a  ,_, a  GL
   0.8
   0.6
rtf
s_
sr
Of
   0.4
   0.2
                   D
                            '   20            30

                         Downwind Distance—miles

                         (a) Pec = 720, k = 0  sec
                    -1
                                         50
      FIGURE  27.   CONCENTRATION AS A FUNCTION  OF  DOWNWIND DISTANCE

                         FOR THE GALERKIN METHOD
                                                                                                      en

-------
fC
S-
O)
O
   0,2 -
                               20           30
                         Downwind Distance—miles
                         (b) Pec = 206, k = 0 sec
        40
50
-1
       FIGURE 27.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
                   FOR THE GALERKIN METHOD (Continued)

-------
   1.0
   0.8 -
c  0.6
o
C
CD
O
   0.4 -
   0.2 -
                                20
30
                         Downwind Distance—miles
                         (c) Pec = 72, k = 0 sec
                                                -1
40
50
       FIGURE 27.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
                   FOR THE GALERKIN METHOD (Continued)
                                                                                                   CD

-------
    1.0
    0.8  -
tO
O
O
    0.2 -
                                              30
                         Downwind  Distance—Miles

                         (d)  Pec  =  720,  k  -  10~4  sec"1
40
50
       FIGURE 27.  CONCENTRATION AS A  FUNCTION  OF  DOWNWIND  DISTANCE
                   FOR THE  GALERKIN METHOD  (Continued)

-------
1.0®
                             2            3

                     Downwind Distance—miles

                     (e) Pec =  720, k = 10~3 sec"1
   FIGURE 27.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
               FOR THE GALERKIN METHOD (Continued)
                                                                                                OD
                                                                                                o

-------
    1.0
   0.8
c.
o
   0,6
O)
u



I   0.4
    0,2
                          •o
                                          o  o
                   10
                                               o
                                                  o\
20
30
                        Downwind  Distance—miles

                        (a) Pec =  720,  k  = 0  sec"1
40
50
      FIGURE 28.  CONCENTRATION AS A FUNCTION OF  DOWNWIND  DISTANCE

              FOR THE PARTICLE-IN-CELL  (SMOOTHED)  METHODS

-------
    T.Q
   0.8
c:
o
c:
O)
o
   0.6
   0.4
   0,2
                  10            20            30

                        Downwind Distance—miles

                        (b) Pec = 206, k = 0 sec
         40
50
-1
      FIGURE 28.  CONCENTRATION AS A FUNCTION OF DOWNWIND  DISTANCE
        FOR THE PARTICLE-IN-CELL (SMOOTHED) METHODS  (Continued)
                                                                                                     co

-------
   1.0
                                     o
   0.8
o
   0.6
i-.

c:
O) '
o
c:
o
   0.4
   0.2
                   10            20           30

                        Downwind  Distance—miles

                        (c) Pec = 72,  k  =  0  sec'1
40
50
      FIGURE 28.  CONCENTRATION AS  A  FUNCTION  OF DOWNWIND DISTANCE

        FOR THE PARTICLE-IN-CELL  (SMOOTHED)  METHODS (Continued)

-------
    1.0
   0.8
    0.6
(tf
S-
-M
C
cu
u


J  0.4
    0.2
                    10            20           30


                         Downwind Distance—miles

                         (d)  Pec = 720, k = 10~4 sec"1
40
50
      FIGURE  28.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
         FOR THE  PARTICLE-IN-CELL (SMOOTHED) METHODS (Continued)
                                                                                                    co

-------
    7.0
    0.8
cr
o
rcf
S-
4->
cr
o;
u
c:
o
o
    0.6
0.4
    0.2
                     o
                                  2            3


                         Downwind Distance—miles

                         (e)  Pec = 720,  k = 10"3 sec"
                                                -1
      FIGURE 28.   CONCENTRATION  AS A FUNCTION OF DOWNWIND DISTANCE
        FOR THE PARTICLE-IN-CELL (SMOOTHED)  METHODS (Concluded)
                                                                                                    co
                                                                                                    en

-------
           10           20            30

                  Downwind Distance—miles

                  (a) Pec = 720, k = 0 sec
        40
50
-1
FIGURE 29.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
               FOR THE EGAN AND MAHONEY METHOD
                                                                                             CO
                                                                                             en

-------
1.0
                10
20
30
                      Downwind Distance—miles
                      (b) Pec = 720, k - 10~4 sec"1
40
50
    FIGURE 29.  CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE
               FOR THE EGAN AND MAHONEY METHOD (Continued)
                                                                                                 co

-------
                   Downwind Distance—miles

                   (c) Pec = 720, k = TO"3 sec"1
FIGURE 29.   CONCENTRATION AS A FUNCTION OF DOWNWIND DISTANCE

          FOR THE EGAN AND MAHONEY METHOD (Concluded)
                                                                                             Co
                                                                                             Co

-------
                                                                           189
7.    Computational  Time

     Concurrent with  any appraisal  of the  accuracy associated  with  alterna-
tive solution techniques must be a  comparison  of the  computing times  required
for these methods.  Although  a particular  method may  be  extremely accurate,
the computational  time it requires  may be  so  large that  a  less accurate, more
efficient algorithm would be  a better choice.   Table  27  lists  the computing
times for the various methods surveyed in  this study.
                                Table 27
        COMPUTING TIME REQUIRED FOR ALTERNATIVE  SOLUTION  METHODS

                                              Computing  Time
           	^Method	          	(sec)
           Price—explicit                          7.50
           Price--implicit                         11.10
           Crowley--second order                    7.40
           Crowley--fourth order                    7.40
           SHASTA                                   7.95
           Galerkin                                13.2
           Egan and Mahoney                         1.10
           Particle-in-cell                         68.2
     Note that all  of the explicit finite difference methods  use approximately
the same amount of computing  time (approximately 7.5 seconds).   Of the implicit
schemes, only a slight difference exists  between the Galerkin and the  Price
methods.  Obviously,  the accuracy more than compensates  for the larger compu-
tational time.   Finally, the  particle-in-cell  methods are extremely costly in
computing time and should be  used only as  a last resort.

-------
                                                                           190
E.    CONCLUSIONS

     The selection  of a  solution  algorithm for a  set  of  partial  differential
equations should  be based  on  considerations  of both speed and  accuracy.   For
the methods  surveyed in  this  analysis,  we  drew the following conclusions:

     >  For  a rapid explicit  scheme,  the  SHASTA technique should be
        chosen.   Not only  is  the  method accurate  and  efficient,  but
        also it  is  guaranteed not to  predict negative concentrations
        dn areas  having  steep concentration  gradients.   This latter
        quality  greatly  enhances  the  appeal  of the SHASTA method
        over competing schemes.
     >  In those  cases in  which extremely  high resolution is desir-
        able, it  is advisable to  develop a Galerkin algorithm  for
        the  transport equation.   The  increase  in  accuracy, stability,
        and  ease  of modeling  irregularly spaced regimes  more than off-
        sets the  increased cost in computational  time.
     >  As more  information becomes available  in  the  open literature,
        the  Egan  and Mahoney  method should be  explored as a possible
        supplement  or replacement for either the  SHASTA  or the Galerkin
        scheme.
     >  Finite difference  techniques  introduce a  considerable  amount
        of numerical  diffusion into the calculation,  producing an over-
        prediction  of pollutant concentrations downwind  from the source.
        The  effect  is most pronounced using  the Price scheme and is
        somewhat  smaller using the Crowley second- and fourth-order
        systems.  The Crowley fourth-order scheme tends  to be  more ac-
        curate than the  corresponding second-order scheme in regions
        having a  steep concentration  gradient, though the fourth-order
        scheme frequently  predicts negative  concentrations in  regions in
        which complex flow fields  exist.   Regardless  of  the technique
        used,  finite  difference methods are  inaccurate for systems in
        which extremely  fast  reactions  occur.

-------
                                                                           191
     >   The  particle-in-cell  technique  developed  by  Sklarew  (1971)  is
        accurate  for  both  reactive  and  nonreactive systems,  provided
        that a  sufficient  number of particles  is  used  in  the simula-
        tion.   However,  the  time required  for  the simulation for  a
        given accuracy can be prohitively  excessive  and can  thus  in-
        validate  the  use of  the  technique.

     In  conclusion, we wish  to caution  the reader about interpreting the
results  obtained  in this study:   These  results  were  developed  for a simple
one-dimensional,  time-dependent  problem in which  a simple first-order  reac-
tion occurs. In  a  real  situation,  this idealized model can  easily  be  invali-
dated by a complex  flow  field, a set of nonlinear reactions, or a complicated
source emissions  pattern.  In essence,  this analysis focused on one aspect of
the complete problem:  the identification  and  assessment  of  the errors  asso-
ciated with  the solution of  the  one-dimensional advection-diffusion equation.
The study did not treat  problems that are  associated with the  method of frac-
tional  steps, nor did it consider systems  in which nonlinearities occur (as
they frequently do  in the  real world).   Yet, since the numerical  diffusion
associated with finite difference techniques is considerable,  the results of
this study serve  as a benchmark  for identifying those  schemes  that  are  the
most accurate in  a  one-dimensional  sense.   If  one can  assume that this  ac-
curacy is maintained  throughout  the entire solution, then the  application of
the most promising  of these  techniques  to  the  current  SAI model will most
likely produce—but cannot guarantee—an improvement in the  results.

-------
                                                                       192
     V    AIRSHED  MODEL  MODIFICATION FOR MULTI DAY  SIMULATION
                           Steven D. Reynolds
                                Jody Ames
A.    INTRODUCTION
     In the  pa-st,  the application of the SAI airshed model has been limited to
the simulation of  one daytime period (usually 5 a.m. to 3 p.m.).  During the
present study, we  adapted the model for multiday runs.  The two most important
benefits to  be derived from multiday simulations are the following:

     >  Treatment  of multiday episodes. A primary objective of adapting
        models to  perform multiday computations is  to provide the basis
        for  evaluating the effectiveness of air pollution control strategies.
        For  example, difficulties  in specifying initial conditions for some
        future year can  be averted by  performing a  multiday run, since the
        predictions on the second  and  subsequent days are generally less
        sensitive  to the choice of the initial concentration distribution
        input to the model.  Also  of interest is the short range prediction
        of the ground-level pollutant  concentrations for strategies such as
        those that might be put into effect when meteorological conditions
        conducive  to severe pollution  episodes occur.

    .>  Identification of errors.  Multiday simulations will be instrumental
        in establishing  possible sources of error in the airshed model.  Errors
        incurred in a short term simulation (say, less than 12 hours) would
        not  accumulate to the extent that they would over a two- or three-day
        period.  As an example, suppose that the predicted concentrations of
        total nitrogen oxides were much higher than measured values after a
        simulation of several days.  This might suggest that either NOX emissions
        are  too high or  sinks of NO  have not been  properly accounted for in
                                  A                    •
        the  model.

-------
                                                                        193
 One obvious difficulty that might preclude usage of the model for multiday
 runs is the accumulation of errors introduced by the numerical integration
 scheme.   But as noted in the previous chapter, several promising numerical
 techniques are available that—if implemented in the model—should reduce
 numerical  error propagation significantly.

     In modifying the SAI model, we considered the following aspects of
 multiday usage:

     >  Treatment of photochemistry at night.
     >  Definition of the modeling region.
     >  Use of a grid with variable resolution.
     >  Generalization of the finite difference solution technique for
        use on a grid with variable vertical  resolution.
     >  Modification of the computer codes.

     Furthermore, to obtain some experience in the performance of multiday
 runs,  we prepared a set of emissions, meteorological, and air quality inputs
 applicable to the Los Angeles basin on 29 and 30 September 1969.   These are
 two days that we studied  under a 'previous EPA contract (68-02-0339).  Using
 these  days, we were able to compare results from the multiday simulation with
 those  obtained from the corresponding 5 a.m.  to 3 p.m.  runs made previously.
 Of particular interest is the comparison of the two sets of predictions at
 3 p.m.  on  30 September to determine to what extent the two sets of predictions
 agree.

    In the following sections, we summarize  our efforts in each of the pursuits
listed  above.

-------
                                                                        194
B.   MODEL REFINEMENTS

1.   Treatment of Photochemistry at Night

     The primary objectives of this study were to examine the suitability of
the kinetic mechanism employed in the airshed model for performing nighttime
simulations and to determine whether chemical reaction effects can be ignored
during a portion of the nighttime period to reduce computing costs.  In addition,
we wished to obtain some experience in running the model at night, since previous
efforts were limited to the simulation of single daytime periods of 10 hours in
duration.  Because of the difficulties we experienced in incorporating the ex-
panded 36-step mechanism in the airshed model, we decided to try  to use the
original version of the airshed model, which treats the kinetics using a 15-step
mechanism.

     To determine the applicability of the original 15-step kinetic mechanism
employed in the airshed model  (see Hecht, 1972) for use at night, we performed
several "numerical" smog chamber simulations with photolysis rate constants set
to only a small fraction of their nominal values.  Since sunlight is one of the
most important driving forces  in the mechanism, we expected the photochemical
processes to be slowed considerably after sunset.  We set k-, (the N02 photolysis
rate constant) equal to 0.01 min   [the remaining rate constants and stoichiometric
coefficients were assumed to be equal  to those employed in the 29 September 1969
validation study (see Reynolds et al., 1973)] and employed the following initial
conditions:
                                             Initial Concentration
                    Species                  	(ppm)	

                     RHC                             0.4
                     NO                              0.5
                     N02                             0.15
                     CO                             15.0

The model  predicted  the  following  concentrations  after eight hours:

-------
                                                                        195
                                            Concentration
                  Species                      (ppm)	
                   RHC                          0.30
                   NO                           0.32
                   N02                          0.32

These results clearly indicate that substantial chemical conversion is predicted
in the absence of strong sunlight.

     Since we expected the predicted concentrations to change only slightly from
the initial conditions, we hypothesized that the HN02 steady-state assumption
was responsible for the large change in concentrations.  The RHC, NO, N02, and
Oo concentrations predicted by the mechanism are independent of the value of kj
(the HN02 photolysis rate constant) used when HN02 is assumed to be in a pseudo-
steady state.  We then carried out a second simulation, which was similar in all
respects to the first  except that HN02 was not assumed to be in a steady state.
The results of this simulation after eight hours were as follows:
                                            Concentration
                  Species                      (ppm)	
                    RHC                         0.38
                    NO                          0.45
                    N02                         0.17

These results indicate that considerably less chemical conversion is realized
when it is assumed that d(HN02)/dt ? 0.

     As a final  check on the old SAI mechanism employed above, we performed an
eight-hour simulation using the new SAI mechanism currently being validated.  The
purpose of this  test was to use the best available kinetic mechanism to obtain
an estimate of how much chemical  reaction takes place in the absence of intense
sunlight.   Assuming RHC to be entirely propylene, k^ to be equal to 0.01  min
(other photolysis rate constants  were scaled accordingly), and initial condi-
tions  to be the  same as those cited previously, the new mechanism predicted the
following  concentrations after eight hours:

-------
                                                                        196
                                              Concentration
                Species                           (ppm)
                  RHC                             0.38
                  NO                              0.45
                  N02                             0.18

These results reinforce our initial belief that the HN02 steady-state assump-
tion is responsible for the observed conversion of NO to N0? in the 15-step
mechanism.

     It appears from the results of these tests that the 15-step kinetic
mechanism previously employed in the airshed model with HN02 in a steady state
will not be suitable for carrying out photochemical  calculations at night.  Since
considerable effort would be required to remove the HN02 steady-state constraint
from the old airshed model and, furthermore, since we are replacing the 15-step
mechanism with the new expanded mechanism (in which HN02 is not assumed to be
in a steady state), we decided to defer further study of the treatment of photo-
chemistry at night.  This effort should be resumed,  however, when further exper-
ience is obtained in using the new mechanism in actual airshed simulations.

     After reviewing the results of the smog chamber runs cited above, we found
that we may be able to drop some, or perhaps all,  of the reaction terms in the
governing equations during a portion of the nighttime period.  If this is possible,
then computing requirements can be reduced significantly.  And since we are
concerned with multiday runs, it is especially important to find ways of reducing
the costs of such simulations.  To study further the possibility of modifying the
treatment of chemistry at night, we need to perform appropriate nighttime simu-
lations, both with and without chemistry in the model, to determine whether
and when chemical reaction effects can be ignored.  If the chemistry cannot
be completely omitted from the model, perhaps the mechanism can be simplified.

2.   Definition of the Modeling Region

     To minimize errors resulting from the need to specify pollutant concentra-
tions at points of transport into the modeling area, one should choose boundaries
of the  region such that either background levels or actual measurements can be

-------
                                                                       197
used to estimate the boundary conditions.  In previous simulations of pollu-
tants in the Los Angeles basin, we exercised particular care to account properly
for pollutants transported into the region from areas over the ocean and aloft
(i.e., contaminants originally residing in the inversion layer and subsequently
injected into the mixed layer as the inversion was eroded by convective heating).
Although significant amounts of pollutants are often carried out over the ocean
at night, it is usually difficult to estimate the concentration levels in the
returning off-shore air mass because of the absence of appropriate measurements.
To employ background concentrations as the boundary conditions, one must model
both the urban area of interest and a portion of the surrounding environs
(suburban, rural, and ocean areas).  In addition, the upper boundary of the
modeling region should be defined at that elevation where background levels
generally exist (1 to 2 kilometers should be sufficient).  Thus, we modified
our original treatment of the vertical extent of the model from the region
between the ground and the inversion base to the region between the ground
and a user-specified surface aloft.  As an example, one could define the top
of the region to correspond to the top  of the inversion layer.  The trapping
effect of an elevated inversion layer within the model is treated through the
z-dependence of the vertical diffusivity.

3.   Use of a Grid with Variable Resolution

     For efficient modeling of an urban area and a portion of its surroundings,
a grid with variable resolution should be used.  Choosing the appropriate degree
of resolution in a particular area of the airshed depends primarily on the spatial
characteristics of gradients in the concentration field.  In areas where gradients
are large, a relatively fine grid should be used; where gradients are small, a
relatively coarse mesh spacing may be adequate.  With respect to horizontal grid
resolution, the mesh spacing in the outlying areas could be, say, two to four times
that used over the urban center.   Because many sources are located at "ground
level," the vertical concentration gradient is often greatest near the surface.
Thus, it may be advantageous to use fine vertical spacing near the ground and
coarse spacing aloft.   Since the numerical technique currently employed in the
model  is not readily adaptable for variable horizontal grid resolution, we devel-
oped only variable vertical  grid resolution capabilities during this study.

-------
                                                                        198
However,  inclusion of a variable horizontal  mesh in the model, possibly using
a nested  grid approach, should be considered in future efforts to incorporate
improved  numerical solution techniques in the model.
4.   Modification of the Finite Difference Equations

     Because the finite difference equations, previously employed in the model
were derived for an equally spaced grid, it was necessary to modify the differ-
ence expressions involving derivatives of the concentration field in the z (or p)
direction.  [See Reynolds (1973) for a discussion of the numerical integration
procedure.]  In particular, changes were required in the advective and diffusive
flux terms  in  Step III of the numerical integration technique.  As an illus-
tration of the nature of the changes made, Eqs.  (44) through (54) in Reynolds
(1973) become:
   cn+l    _   **      i , j +    AT       Fn+l
    i>0,k " A isj,k AHn+l   2Ap ARn+l  U
                           F**      \ ,    AT   / Rn+1     Hn+l
                i,j,k+% " A ri,j,k+% J   2AHn+l  UKi,j,k   i ,j
               R**    AHn  \  + _^_  / sn+1  ,  AHn+1  +  Sn .  ,  AHn .      ,   (104)
                                ?AH     I   1"3'    1'J      1>J»k   ls
where
   Fn+l      _  'P-i.j.k-ia L    rn+l      ,
  oi  T  I/  1  ~"    ' 9      '  '-  "  -  -  '- n
  A- 1 5 J 5 K~-^      c
n+l      / cn+l     _  cn+l  \
                                                 ,   k = 2, 3, .... K    .     (105)
                                                             '

-------
                                                                       199
          **           ^i  i  k_% /     **                   **
         F         =   —' >J »K  2 I v   r        + i-\   -\ \  r
               -'               A   L-  + u •      L
                                   **          **
                        yi,J,k-Js  (£Ci,j,k-l  "  £Ci,j,k      '     k = 2,  3,  ..., K

                                                                            (106)
                      0.5
                           Kn
                             i»j.k-is -        ^                         (108)
                          Ap.
                 \-
In view of the  discussion  given  in  the  previous  section, p is now defined by
the following relationship:
                         -,. z " h(x^!
                        Hlx,y,t)  - h(x,y)
where
          h(x,y)     =   terrain  elevation,
          H(x,y,t)   =   elevation  of  the  upper  boundary of the modeling  region,
          Ap,        =   the  dimension less height  of  the k-th  level grid  cell.

     The  boundary  conditions  at the  ground  and aloft  are the following:

     (1)   P  =   0
                             Fn+l     =   Qn+l
                            d'-i T  Is     O^n  \     »
                            * 1 >J »-2     * 1 ,J
         where k =  1/2  is equivalent to  P = 0.

-------
                                                                       200
     (2)   p   =  1
                                            1f
                                 **             ,n
if    *i,j,K+Js  >
                                                            0
          where  k  =  K + j is  equivalent  to  p  =  1.

     Changes  in  the  matrix expressions given  in Eqs.  (55)  through  (70)  of
Reynolds  (1973)  follow directly  from  the difference  equations  given  above.

5.    Modification  of the  Computer Codes

     To  carry out  multiday simulations,  we  modified  several  portions of the
computer programs.   These changes essentially make the codes  more  general.
Furthermore,  the main code now allows the user  to  use a grid  with  variable
vertical  resolution, where the spacing interval  is treated as  a  model  input.
As  an example,  in  the Los Angeles simulation, which  is discussed in  the next
section,  we  used a  grid with  10  vertical  levels and  the following  mesh  spacing:

                                          Grid  Spacing
                    	Level	         (feet)
                    10 (top)                  1625
                     9                         825
                     8                         425
                     7                         225
                     6                         125
                     5                          75
                     4                          50
                     3                          50
                     2                          50
                     1 (ground level)"           50

-------
                                                                       201
     Thus, the modeling region is assumed to extend from the ground to an
elevation of 3500 feet above the terrain.  In addition, we implemented
appropriate changes, corresponding to the discussion given in the previous
section, in the coding of the finite difference equations.

     In addition to the above-mentioned changes, we restructured the input
data deck setup to operate in the following manner.  First, all parameters
global to the run--i.e., those parameters that would not be expected to vary
from day to day—are input.   Then, the remaining inputs are arranged in
daily packets, one packet for each day to be simulated.  When the simulation
reaches midnight, the input packet for the next day is read from the input
file.  After the first day,  some daily parameters can be omitted from the
input packet, and the values used on the previous day can be used again.

     We also included provisions that allow the user to establish multiple
emission files for the input of emissions data to the program.   For example,
one might establish two sets of emissions, one applicable to weekdays and the
other suitable for weekends.  Once such a set of files is established for an
urban area, multiday runs consisting of any pattern of weekdays and weekends
can be simulated.  Table 28 illustrates the deck setup and lists some of the
main parameters included as  part of the global  and daily inputs.

C.   MULTIDAY SIMULATION OF  THE LOS ANGELES BASIN

1.   Preparation of Emissions and Meteorological Inputs

     Since previous applications of the airshed model  were limited to the simu-
lation of a 10-hour daylight portion of each of six days in 1969,..little con-
sideration was given to the  definition of emissions and meteorological  inputs
for use at night.  Thus, to  gain experience in  the performance of multiday runs
with the SAI  model, we carried out a necessarily limited effort to estimate
meteorological  and emission  inputs that would apply during the period from
3 p.m.  on 29  September to 5  a.m.  on 30 September 1969 (previous simulations
were carried  out for the 5 a.m.  to 3 p.m.  period on both 29 and 30 September
1969).   In  particular, we performed the following tasks:

-------
                                                                       202
                                Table 28
                     ORGANIZATION  OF  MULTIDAY  INPUT

               Global  data
                 Run heading
                 Simulation options  and grid definition
                 Start and  stop times and dates
                 I/O units
                 Print options for maps
                 Region definition
                 Stations and landmarks
                 Integration parameters and stoichiometric  coefficients
                 Activation energies
                 Initial  Conditions
               Day 1 packet
                 Date, emissions type, input controls
                 Rate constants
                 Light intensity factors
                 Deposition velocities
                 Concentrations aloft
                 Point source emissions
                 Boundary conditions
               Day 2 packet
                 Date, emissions type, input controls
                 Rate constants*
                 Light intensity factors*
                 Deposition velocities*
                 Concentrations aloft*
                 Point source emissions*
                 Boundary conditions*
               Day 3 packet
it
  Can be omitted after Day 1.

-------
                                                                       203
     >  We used the SAI automated wind field analysis package to generate
        hourly wind speed and direction maps spanning the period from 5
        a.m.  on 29 September to 3 p.m. on 30 September 1969.
     >  We employed the SAI automated inversion analysis program to esti-
        mate  hourly mixing depth maps, using actual  observed  mixing depths
        available for the daytime periods to the extent possible.   We
        examined nighttime temperature profiles measured by Meteorological
        Research, Incorporated in the Los Angeles basin during the  summer
        of 1973 and estimated that pollutants would  typically be mixed
        throughout a depth of about 60 to 70 meters  at night.
     >  We prepared a set of fixed-source emission maps for hydrocarbons
        and NO  that are applicable from 6 p.m. to 6 a.m. the following
              A
        morning; the original SAI fixed-source maps  were dervied for the
        complementary 12-hour period.  Using data presented in Appendix A
        of Roth et al.  (1971), we estimated that about 25.5 tons of NO
                                                                      /\ •
        and 30 tons of low reactivity hydrocarbons are emitted in the
        modeling region between 6 p.m. and 6 a.m.
     >  We specified boundary conditions at points of horizontal inflow
        into  the model  between 3 p.m. on 29 September and 5 a.m. on 30
        September.  Boundary conditions at other hours were available from
        the model validation studies reported in Reynolds et  al. (1973).

     At this  point, it is appropriate to note that the paucity of meteorological
and emissions data applicable specifically during the nighttime hours makes the
estimation of mixing depths and emission rates highly uncertain. The primary
objective of  our present effort was simply to assemble a set  of "reasonable"
inputs that can be used in tests of the multiday version of the SAI airshed
model.  Further efforts should be made to refine the temporal  distribution of
surface street and freeway traffic activity and the  spatial and temporal distri-
butions of the HC and NO  emissions from stationary  sources at night [see Roberts
et al., (1971)].

2.   Discussion of the Multiday Simulation Results

     To test  the various modifications made in both  the structure of the model
and the computer codes, we performed a multiday simulation of pollutant concen-

-------
                                                                       204
trations in the Los Angeles basin.  As noted in Section B-l,  we decided that
the 15-step chemical kinetic mechanism employed in the model  is inappropriate
for use at night.  Thus, the simulation reported here was carried out for CO
alone.   When an appropriate set of emissions inputs suitable  for usage with
the new mechanism can be developed, multiday photochemical  simulations should
be undertaken.

     Plots of predicted and measured CO concentrations at the downtown Los
Angeles, Long Beach, West Los Angeles, Burbank, Reseda, Whittier, and Azusa
stations are given in Figures 30 through 36, respectively.  The simulation
extends from 5 a.m. PST on 29 September to 3 p.m.  PST on 30 September 1969.
The results for the 5 a.m. to 3 p.m. period on 29  September  are very similar
to those obtained in the previously reported SAI validation effort.   (Dis-
crepancies in the two sets of predicted results are due to  the manner in
which the meteorological .variables were specified—automatically in  the
former case and manually in the latter.)  Of greater interest is an  examina-
tion of the remaining results, which are best approached by  considering the
5 p.m.  to 5 a.m. nighttime period and the following 5 a.m.  to 3 p.m. daytime
period separately.

     In general, the nighttime predictions are reasonably accurate consider-
ing the fact that neither vertical temperature measurements nor refined
temporal distributions for motor vehicle activity  were available to  estimate
the corresponding meteorological and emissions inputs.  Results at the end
of this period (5 a.m.) often fell within a few parts per million of the
measured concentrations, as shown in Table 29.  Two notable exceptions, how-
ever, are illustrated in the downtown Los Angeles  and Burbank predictions
(Figures 30 and 33, respectively).  Upon further examination  of the  results
for these two stations, we made the following observations:

     >   Downtown Los Angeles.  A rather substantial build-up  in the  CO
        concentration was observed to occur from 9 p.m. to midnight, but
        it was predicted to take place two hours earlier, from 7 p.m. to
        10 p.m.   Since the magnitudes of predicted and'measured concen-
        trations during the early morning hours of 30 September agree

-------
  40
   30
s
Q.
CL.

I
c
o
O)
   20
   TO
                          PREDICTED CO CONCENTRATION

                          MEASURED CO CONCENTRATION
               D
                                                                                           D
                                                                                        D
                                                                                     n
                              12      15        18

                           29 September 1969  	
21
0
    6        9

30 September 1969
12
                                                     Time—hour
15
                         FIGURE 30.  COMPARISON OF PREDICTED AND MEASURED HOURLY AVERAGED
                                      CO  CONCENTRATIONS AT DOWNTOWN LOS ANGELES
                                                                                                                    no
                                                                                                                    o
                                                                                                                    Gi

-------
   40
   30
Q.
CL
 l
 I


J 20

to
t-
O)
u
sr
o
O
   10
                         PREDICTED CO CONCENTRATION

                         MEASURED CO CONCENTRATION
                                                                                        D
                              12        15        18

                           29 September 1969	
                                                        21
0
    6        9

30 September 1969-
12
15
                                                  Time—hour
                          FIGURE 31    COMPARISON OF  PREDICTED AND MEASURED HOURLY AVERAGED
                                           CO CONCENTRATIONS AT LONG BEACH
                                                                                                                     ro
                                                                                                                     o
                                                                                                                     CTl

-------
   40
   30
0.
d.
   20
S-
o>

-------
  40
   30
CL
CL
I
I
   20
O)
o
cr
o
   10
                   	   PREDICTED  CO  CONCENTRATION

                   D      MEASURED CO CONCENTRATION
                                                                                       D
                                                                                    D
                                       D
 9         12        15       18

:	29 September 1969	
21        0
                                                                          3        6        9        12       15

                                                                               30 September 1969 - •
                                                  Time—hour
                         FIGURE 33.   COMPARISON  OF  PREDICTED AND MEASURED HOURLY AVERAGED
                                           CO  CONCENTRATIONS AT BURBANK
                                                           ro
                                                           o
                                                           CO

-------
  40
  30
0.
Q.
I
I
*  20
O)
O
   10
                   	   PREDICTED  CO  CONCENTRATION

                    D     MEASURED CO CONCENTRATION
                                                                                        D
                           D
                     9       12        15       18       21
                    	-29 September 1969	
 3         6         9         12
	30  September 1969	
                                                     Time—hour

                         FIGURE  34.   COMPARISON OF PREDICTED AND MEASURED HOURLY  AVERAGED
                                            CO CONCENTRATIONS AT RESEDA
15
                                                                                                                   ro
                                                                                                                   o
                                                                                                                   i-D

-------
   40
   30
ex
CL,
 I
 I
cr
o
o
tr
o
o
   20
   10
                      PREDICTED CO  CONCENTRATION

                      MEASURED CO CONCENTRATION
                                                                                     D
   D   .


D     D  D
D
                                                                                  n
9        12       15       18

	29 September 1969	
                                                          21
 369

	30  September 1969
                                                   Time—hour
                                                                                           12
                                                                                        15
                          FIGURE 35.  COMPARISON OF PREDICTED AND MEASURED HOURLY AVERAGED
                                           CO CONCENTRATIONS AT WHITTIER
                                                                                                         ro

                                                                                                         o

-------
   30
CL.
Q.
 I
 I
*  20
O)
u
cr
o
o
   10
                    	   PREDICTED CO CONCENTRATION

                    D      MEASURED CO CONCENTRATION
-nan   •  n
                                                                D
                                 D
                                                                  n  n
                             12       15       18

                           29 September 1969	
                                                   21
                                                   Time—hour


                          FIGURE 36.   COMPARISON OF PREDICTED AND MEASURED HOURLY AVERAGED
                                             CO CONCENTRATIONS AT AZUSA
                                                                                                          n  n
 3         6         9        12

	30  September  1969	
15
                                                                                                                     ro

-------
                                                                        212
         fairly well, the discrepancy in the time of occurrence of the
         build-up may well be the result of inaccuracies in the temporal
         distribution of motor vehicle emissions (the only source of CO
         in the model) during the period from 7 p.m. to midnight.  Thus,
         the total loading seems correct, but the temporal distribution
         appears to be in error by about two hours.  Traffic activity at
         5 a.m. also seems to be greater .than that predicted by the SAI
         emissions model.
      >  Burbank.  Of all the results presented in Figures 30 through 36,
         those calculated for the Burbank station are the poorest.  Pre-
         dicted concentrations after 5 p.m. on 29 September are consistently
         low by as much as 10 ppm.  Upon examining the results for Reseda
         (also located in the San Fernando Valley), we noted that, although
         predictions are often low, the discrepancy in the predicted and
         measured concentrations is at most only 3 ppm.  Meteorological data
         for Burbank indicate very light winds (1 mph) from the north.   The
         high concentrations at this station may thus be the results of local
         emissions from the major interstate freeway situated to the north
         of the station.  A shallow mixing layer coupled with near-calm con-
         ditions would certainly limit the extent to which local freeway
         emissions would be dispersed.

Basically,  the  nighttime  predictions  do  not  indicate  any  systematic  errant
behavior in the  model.   In  fact,  considering the  absence  of  key  meteorologi-
cal  and  emissions  data,  the  5  p.m.  to  5  a.m. results  are  about  as  good  as
might  be expected  under  the  circumstances.

     Turning now to an examination of the daytime results for 30 September,
we note that at many stations  the multiday predictions are very similar to
those previously reported in the single-day test of the model.   Since the
meteorological  and emission inputs for these two runs were not significantly
different,  discrepancies in the two sets of predictions can be attributed
primarily to differences between the multiday  and single-day CO concentra-
tion distributions at 5  a.m.  PST on 30 September 1969.  In the single-day
run, which  began at 5 a.m., the initial  CO concentration field, shown in

-------
                                                                       213
                                Table 29
         PREDICTED AND MEASURED HOURLY AVERAGED CO CONCENTRATIONS
          AT THE END OF THE 29 TO 30 SEPTEMBER NIGHTTIME PERIOD*

                                  Predicted            Measured
                                Concentration       Concentration
         	Station	      (ppm)               (ppm)
         Downtown Los Angeles        13                  13
         Azusa                        6                   8
         Burbank                      7                  16
         West Los Angeles             7                   6
         Long Beach                   6                   3
         Reseda                       7                   6
         Pomona                       3                   3
         Lennox                       7                   8
         Whittier                     7                   7
         *
          The  figures  presented  in  this .table were averaged from
          4  a.m.  to 5  a.m.  PST on 30 September 1969.

Table 30, was estimated from the  appropriate measurements  reported by the
Los Angeles  and Orange County Air Pollution Control  Districts.   The concen-
tration field on the grid at 5 a.m.  in  the multiday  run is,  of course, the
result of a  continuous simulation started at 5 a.m.  on the previous morning.
The ground-level concentration map for  this case  is  illustrated in Table 31.
Over much of the modeling region, the two sets of predictions  agree within
2 or 3 ppm.   However,  the discrepancies are much  higher in the  Pasadena,
downtown Los  Angeles,  and Burbank areas.   The high CO levels  in these areas
given in Table  30 are  the result  of  manual  interpolation of the measured
values reported at downtown Los Angeles and Burbank  (the Pasadena  station
did not report  CO levels on 30 September).   In general, the single-day results
for downtown  Los Angeles and Burbank are  better than  the corresponding multi-
day predictions.  At other stations, the  model  predicted CO levels reasonably
well, especially in view of the significant impact that local  roadways may have
on measured  concentrations during the peak traffic hours in  the morning.

     It is also interesting to note  the extent to which errors  have accumulated
throughout the  multiday run.   In  Table  32,  we give both the predicted and  mea-

-------
                                                              Table 30
                                        MULTIDAY GROUND-LEVEL CO CONCENTRATION MAP
                                             AT  5 a.m.  PST ON  30  SEPTEMBER  1969
                                          CO   GROUND LEVEL CONCENTRATIONS  (PPM 1 AT  500.00  PST
                                                     10   U   12    13    1*   15   16   17    18    19   ZQ   Zt    22    23   2»   25
 25   7.0  7.0  7.0   7.0  7.0	8^0	9iP_lg_..0_ll. P_ 11.. 0_l I . Q_

 24   7.3  7.3  7.J   7.3 '8.3  9.0 13.3  12.3  13.0 13.3 12 ^Q_
                                                                                               SAN  GABRIEL HTNS
      7.0  7.0  7. 0   7.0  8.0 10. 01 ?,. 0_J._i_-jL 1 5 • P_J 5 . fl_J£i. P_JLis.P_J.Oj Q __ ;
                RESEDA
 22   7.0  7.0  7.0   7.0  e.O 10.0 12.0  14.0  16.0 16. 0_16.J)_l._0_\t>.3  16,3J5.3 14.0J2.J 10.iJ_8.p _ 8. 3_8. J_.8.0_ J. 0 __ 6.0_5.3 _*.0
              Si.'lTA  KCNICA KTNS
 19   5.0  6.0  7.0 _ 7.J> _ 7.0  8.0   9.0  II. 0  13.0 15.0 16.0  16.0  16.0_16.p_15. 0 J2.0_IJI.O _8.0  _8.0_  8.0_8..Q _ 7. p _ 6. 0__5.0 __4.0
                                                                                            ~
 IB   ;.Q  6.0   7.0  7.0  7.0  7.0   8.0   9.0 11 .0 13 ._0 15 . 0"_ 1 6 . 0__1 6,0. 16 .0 1 6 .0 13.0  11.0   8.0 _ 8 . 0 _ 8 .0 _ 8_. Q _ 7..0 _ 6_. 0 _ .5.0 _ >, . 0
                                                 "•     "   "D'QKNTOHN'LA          "                     -_...-.      .
 17 __ 5.3 _ 5..0 _ 6.p_ 6._D_6._0  J. P _ 7. 0 _8 ,0_lp.JI 12.0 15.0_16.0  16.0 16.0 16. 3_13- 0_U .0_  8 .0  _ 8.0 _8.0 _ .8. 0 _ 7.0   6.0  5.0  «.0
                                KfST  LA
 16  ___ ; ___  g^O  6.0 _ 7.0_ J.O   0.0 10.0 12.0 ^^.0_16.p_2t.O 16.0 16.0_l>.0_ll.p_ 8.0 __ 8.0_8.0 _ 8.0_7.0   6.0 _ 5.a_«.0
                         "~ .......... ..... " ........... '" ......  ""  "  "   "COKMcilCE  "~  ". ..........
 JJ _ SiO  6.0   7.0   8.0 10.0 12.0 K.O  I5.0_16. Q l&.O 15.0 12.0  10.0 _ 8.0   S.O  6.0 _ 8..0__7,p _ t>.0_ _ 5.0 _ J..?
 '.4 _ ___ 6_.^ _7.0 _ 9.0 lP-P_12.0_12.a_J2.0_l2.p_l.?.p_ll.?_11.0_10..0_^8.p _8.P _ 8 ,0_ 8. 0 _t-0_4.0 _5.0 _4.p
                                          LENNOX
js ___    ___ 6.0   a.p  10. p ip. o ip.p_ig.o  ip.p_iq.o_io.o ip.oj^.p __ 9.0   e.p   S.O_T.O  6.p   s.o_ s.o_5.o  5.0
             "       "          ""    ...... ~ ..... "" '      "' " "    .......  .          '     " 'HHITTJER  LA" HAORA
 iz _ s.o   7.0  7.0 _ euo _ e.o   s.o'  8 ._p _ _B .0 _ e.j _ a.o  7.0   7.0 _ 7..°__6.- P _ 6 . o _ 5 . o _ 5 . p _ 5. o _ s. p

 1 1. __ 3_;P _ A.O _ 5 .^ _ 5^0 _ 6.0 _ 6.3 _ 6 .3 __6 . 3 ._. 6 .3 _ 6.JD __ 6.0 _ 6 . 0 _ 6 . 0 __ 5...0_ 5 . 0 _. 5 . 0_5 . 0 _ 5 .0 _ 5 . 0

 10 ___ ^ __ J ___ }•?  __l'?_*-?_*'? _ -L-? _ .?•£.  5«°  5.0  5.C _ 5^0 _ 5,0 _ 5.0_5.p__5.p_5.0   5.0_5.0_5.J _ 5.0
                                                "" ...........  "LUNG'BEACH ........ "   ""     "    ANAHEIM
  •3 _ 3^0 _ 3i.p_JV._0 _ 4.0  5.0 _ 5..0 _ S..O _ 5.._0 _ 5.0  S.O  S.O   3.0   S.O  5.0 _ 5.0 _ 5.0 _ 5.0 _ 5...0 _ 5...0

  8 ____ 3.0   3.J  3.D _ i_.0__A.O _ 5.0__5.J_5.p _ 5.0 _ 5-.p_5.0 _ 5.0_^.0 _ ?.p _ 5.0   5.0_5.0_5.0 _ 5.0
                                       PA LOS "VEROCS
  7 ___ 1-J°__1- 5_3 --° _ L-.O _ ?_- P__i :° _ * -.P _ f>r.9__i.-.9 _ V..9 _ 5.-.2 _ *• ° _ *• °— * .-A_ 5 .. 0 ._ 5 . 3 __ 5 -0._5.3_5.D

  .6 __ 3.0   3.0  3.0  3.0  3.0   3.0  3.0  3.0  3.0  3.0  3.0   '4.0 _ A.O  A.O  5.0   5.0 _ 5.0 _ 5.0 _ 5^0

  $  __             ____ 3_.0 _ 3.0 _ A_,0 __ ^0 _ 5.0_  5.0   5.0_ 5.0  5.0
                 p'A'cri='iC~ot'EAN                                                             "                 "    "  SANTA'ANA
  

_ 3.0_ __ 3.0 _ i'3_Jr..O_5.p__5.0 5.3 5.0 __J __ 3.0 3.0 3.0 4.0 *• . 0 _ 5 .0 5.0 _. 2 ........ ^ ... ......... ... ________ . ____________ __ ________________ 3.0 3.0 A.O A.O S.O __ 5.0 __ 1 ________________ . _____ , _____ >-P._5r0_. *?<>.._ 4,0 ....4,0


-------
                                                                Table  31
                                          SINGLE-DAY  (GROUND-LEVEL  CO CONCENTRATION  MAP
                                                AT 5 a,m'.   PST  ON 30 SEPTEMBER 1969
                                            CO   GROUND'L^VEX CONCENTftATrO/^S (PPM )  AT  SOO.OO nw 6909JO
                                   6     7    fl    9    10   )l   12   13   1«   15   16   17   18   19   20    31    22    ->3    '<"•   ,ff>

1
i —
i
^
25
2*
23
22
21
20
5.3
S.I
5.4
5.8
6.0
4.9
.5.2
5.8
6.3
6.5
*.8
..5-2
5.9
SFSE
6.9
7.0
5.Z
s-s. .
6.5
DA
7-9
6.7
6.1
.6.7
S.4
9-6
4.6
6.7
.. 7.2.
8.5
10.3
9.7
5.5
; SANTA MONICA MTwS
7.1 7.6
,._7.4 ;. 7.2
7.9 7.9
7.4 7.7
B. 6 8.0 7.2 7.0
10.2 8.9
10.8 10.8
7.9 _?.9

7.6 7.1
DURBAN*
8.8 7.7
8-4 7.4
\
8.0
7.9
"7.4
6-7
6.8
7.5



SAN GA0RI£X MTNS
8.0 8.1
7.6 8.0 8.2. 8-Z 8.0
7.3 7.7 6.4 fl. 7 8.5 8.5 ?.a 9.0 8.1 6.6 5.1 4.1 3.5 3.1
PASAO'-Nft A?US3
6.8 7.5 8.3 9.0 9.1 9.1 9.5 8.9 7.) 6.2 5.5 4.5 -3.6 J.I

!    19   3.7  3.6  3.3  3.0  3.8 -<..0   5.9  8-7  10.a  8.8  8.4  7.7  7.6  8.2.  S.8  9.1  9.3  9.2  8.0  6.6  6.5   6.2   4.8   3.6   3.Z
:_.  	 .  _.   	   _  	   .        j.                                   _            EL MONTE                  	
|_ ..lfl.. 2-8 _?.8  E.8 _3.0  3.9  ft .5   5.9 ..6.*  10.Q  10.6  10.7  10.0  9.0  8.9  9.4 lo.O 10.4 10.1  8.8  8.1   3.3   7.6   fc.7   4.1   3-S
|   "   "    	   	                POMNTOtfW l_ft  	    ....                     .                POHON*
•	 lT_"..Id"_i>*_3-^.1_l?'J_i-T__«-?   6'?.^?-?.  '*•? ''?'S  H-O  I?-5 ll-|_ll.V"lpVo"_^.8"l£.7 J0.&_ '_9.i""8. I_l7.8	3.3	?.9	.'r.?_3-3_
                                  wfST"~l."»"	~	~	"'   '"          ""	      '    "     	~	""
!    16                  4.7  6. Z  6.9   7.6  7.8  10.0  13.6  15.'t  H.a 14.0 13.6 11.1 10.5 11.0 10.5  9.1  7.<)   7.2   7.2   6.2   "..••   3-3
                                                                     COMMERCE.
15
14 ~
13
12
10
9
8
7
6
4
3
2
1
5.6 6.6 7.5 7.8 9.1 11.6 14.1 13.0 12.3 I3..
0 4
S 4
IARRA
9 "5
7 9
1 10
9 10
4 1U
4 7
9 ?
t 4
5 3
1
.3
.6
b
S
.9
• 4
•S
5.9. 4
S.'6 " 4
5.3 4
5.4 4
.2 7.8 6.1 -
ANA»|eif1
.1
. 4
.0
.6
.4
.6
.&
e
10
"
7
S
.3
.1
. it
.Z
.0
SANTA
* .0
a
1
.8
. t.
<>.3 4
7.6 5
8-5 5
6.1 4
4.4 3
3.5" 3
2.6 i
.3 3.3
.3 3.3
.2 3.3
". 3 " 3 . 3
.7 3.4
. »" 3.3 " ' '
.5 3.4
.6 3.5
.9 3.7
.7 3.4
.3 3.2
.? -2.9
2.0 *-3 £.6

-------
                                                                      216
sured hourly averaged concentrations  over the last hour of the  simulation.
Overall,  the predicted results  tend to be lower than  the measurements  by 1  to
2 ppm; in only one instance is  the discrepancy greater than 2  ppm.   Since we
expect the model  to predict concentrations somewhat lower than  those measured
at stations situated on heavily traveled streets,  it  is difficult  to assess the
cumulative effect of meteorological,  emissions, and numerical  errors in  this
simulation.  Thus, it appears  that we may have to  carry out longer runs  (three
or four days) to  observe the build-up of modeling  errors clearly.
                                  Table 32
           PREDICTED AND MEASURED HOURLY AVERAGED CO CONCENTRATIONS
                 FOR THE LAST HOUR OF THE MULTIDAY SIMULATION

                                    Predicted            Measured
                                  Concentration       Concentration
           	Station	       (ppm)               (ppm)
           Downtown Los Angeles         5                   4
           Azusa                        3                   5
           Burbank                      4                   5
           West  Los Angeles             4                   3
           Long  Beach                   4                   6
           Reseda                       2                   3
           Pomona                       3                   6
           Lennox                       3                   5
           Whittier                     3                   4
 D.   RECOMMENDATIONS FOR FUTURE WORK

     During this study, we adapted the SAI airshed model for use in the pre-
 diction of inert pollutant concentrations over multiday periods.  The predic-
 tion of photochemical contaminant concentrations should be undertaken when the
 program containing the new kinetic mechanism is fully operational and suitable
 hydrocarbon emission inputs are developed.  To gain experience in multiday us-
 age, we simulated pollutant concentrations in the Los Angeles basin for the

-------
                                                                      217
34-hour period extending from 5 a.m.  on 29 September to 3 p.m.  on 30 September
1969.   In  general,  the results obtained from this  run agree reasonably well
with available measured pollutant concentration data.

     We recommend that the following  tasks be undertaken in the future:

     >   Assembly of an accurate data  base for both meteorological  and
        emissions inputs for a multiple-day period.
     >   Performance of photochemical  simulations as  soon as possible.
     >  Performance of CO (and eventually photochemical) runs over
        several consecutive days (say, four or more) to obtain a
        better understanding of the cumulative effects of meteoro-
        logical., emission, and numerical errors on our ability to
        exercise the model on a multiday basis.

Our present experience indicates that, given a suitable input data base, the
model should be capable of producing reasonably good predictions of inert
species concentrations for at least two consecutive days.  However, further
testing will be required to establish guidelines regarding the total number
of days that can be simulated before errors accumulate to unacceptable levels.

-------
                                          218
       APPENDIX
A USER'S  GUIDE TO

      David C. Whitney

-------
                                                                     219
                              APPENDIX
                    A  USER'S GUIDE TO  MODKIN
                          David C. Whitney
1.   INTRODUCTION

     Quantitative description of the rates of chemical  reaction  of atmospheric
contaminants is a vital  ingredient in the formulation of a  model  capable  of
accurately predicting ground-level concentrations of gaseous  pollutants.   The
formulation of a kinetic mechanism having general validity  is,  however, an
endeavor beset by several  inherent difficulties.   First, many stable  chemical
species are present in the atmosphere.   Most of these exist at  very low con-
centrations, thereby creating major problems of detection and analysis.   In
fact, a number of atmospheric constituents remain unidentified.   Second,  the
large variety of highly  reactive, short-lived intermediate  species and free
radicals further complicates the picture.  Finally,  the enormous  number of in-
dividual chemical reactions that these  species undergo  creates  an even greater
barrier to understanding.   Nevertheless, despite  our limited  knowledge of at-
mospheric reaction processes, it is essential that we attempt to  formulate
quantitative descriptions  of the processes that are  suitable  for  inclusion in
an overall  simulation model.

     The formulation and development of a kinetic mechanism that  is to be in-
corporated in any airshed  model  is both delicate  and exacting,  an undertaking
requiring a blend of science, craftsmanship, and  art.   On one hand, such  a
mechanism must not be overly complex because the  computation  times for inte-
gration of the continuity  equations in  which the  mechanism  is to  be imbedded
are likely to be excessive.  On  the other hand, too  simplified  a  mechanism
may omit important reaction steps and may thus be inadequate  for  describing
atmospheric reaction processes.   Therefore,  one major issue is  the requirement
that the mechanism predict the chemical  behavior  of  a complex mixture of  many

-------
                                                                       220
hydrocarbons, and yet do so with a paucity of detail.  Thus, in postulating
a mechanism, the formulator must strike a careful balance between compactness
of form and accuracy in prediction.

     As an aid in the development of a kinetic mechanism for atmospheric
photochemical reactions, we prepared a computer program that allows the user
to present his proposed mechanism via data, input cards in the same manner as
he would formulate it on paper--i.e,, as a series of chemical equations and
their associated rate constants.  Moreover, he can select the method of cal-
culation for determining the concentration of each chemical species in the
'mechanism from among the following choices:  the integration of coupled or
uncoupled differential equations, the solution of algebraic equations for
species in a steady-state, or the assumption of a constant concentration.
Either static or dynamic smog chamber observations can be simulated, and plots
of species concentration as a function of time are provided as part of the
printed output.  Reactions of similar species can be combined into a single
"lumped" reaction.  Changes in the reaction mechanism, rate constants, or
species type designation can be effected by simple input card replacement;
receding or recompilation of the program is not necessary.

     Since descriptions of the solution techniques and the development of the
chemical mechanisms have appeared elsewhere (Seinfeld et al., 1971; Hecht,
1972), we do not repeat them here beyond the degree necessary for an under-
standing of the computer program.  This appendix is designed to serve primar-
ily as a user's guide to program operation and as a programmer's guide for
such program maintenance and modification as may be needed in the future.

2.   USE OF THE PROGRAM

     The input to MODKIN consists of two control cards, a set of reaction cards,
a set of species cards, a set of flow cards, and a set of plot cards.  The card
formats are described in Table A-l , and additional comments regarding program
input are given below.

-------
          Table A-l


INPUT CARD FORMAT FOR MODKIN
Card No.
1
1
1
1
1
1
1
1
1
1
2
2
2
Column No.
1-12
16-20
21-25
26-30
31-35
36-40
41-45
46-50
51-55
56-60
1-10
11-20
21-30
Variable
Name
NTIT(J)
NRXN
NLMP
NDIF
NSTS
NUNC
NREP
.MINT
NFLW
NRAT
TINCR
TEND
HSTART
Item Units of
Format Measure
3A4
15
15
15
15
15
15
15
15
15
F10.0 min
F10.0 min
F10.0 min
                         	Comments	

                         Twelve^character title for run heading

                         Number of reactions in the mechanism (maximum 99)

                         Number of reactions that contain species to be
                         be replaced, i.e.,  are lumped (maximum 10)

                         Number of species to be solved in a coupled
                         differential equation (maximum 40)

                         Number of species to be solved in a steady-
                         state approximation

                         Number of species to be solved via uncoupled
                         differential equations

                         Number of species that are replacements  for
                         lumped species

                         Number of inert (constant concentration) species
                         (the total  count of the above five species types
                         cannot exceed 50)

                         Number of species flowing into the reaction chamber

                         If nonzero, reaction rates will  be printed

                         Time increment  for  printing and  plotting results

                         Ending time for run                                :

                         Initial  time step size
ro
IX)

-------
                  Table A-l   (Continued)
Card No.
2
2
2
2
3
3
3
4
4
5-
. Column No.
31-40
41-50
51-60
61-70
1-20
21-50
51-60
1-4
6-10
1-20
Variable
. Name
HMINF
HMAXF
EPSF
Q
NMRC(J.K)
COEFF(J,K)/
NMPD(J,K)
RK(K)
NTEST
NLOC
NMRC(J,K)
Item Units of
Format Measure
F10.0 min
F10.0 min
F10.0
F10.0 min"1
4(A4,1X)
3(F6.0, : —
A4)
F10.0 ppm\~n
min"
A4
15
4(A4,1X)
5


5
21-50


51-60.
COEFF(J.K)/  3(F6.0,
NMPD(J,K)    A4)
RK(K)
F10.0
                         min
 	Comments	

•Minimum time  step  size

 Maximum time  step  size

 Fractional  allowable  error for iterative solutions

 Dilution or flow rate (sampling and  leakage com-
 pensation)

 Reactant names,  up to four per reaction  (if
 lumped  reaction, species  to  be replaced
 must  appear first)

 Species coefficients  and  product names,
 up  to three per  reaction

 Rate  constant; n  is  the  number of reacting
 species

 Name  of species  to be replaced in the  lumped
 reaction

 Number  of reactions contributing to  the
 lumped  reaction  (maximum  10)

 Reactant names,  up to four per reaction
 (the  first  species name must  be the  replacement
 for the lumped species)

 Species coefficient and product names, up
 to  three per  reaction

 Rate  constant; n  is  the  number of reacting
 species
                                                                                                    rv>
                                                                                                    r\3
                                                                                                    ro

-------
                                Table A-l   (Continued)

Card No.
6
6
7
7
8
9
9
9
9
9
9
,.
Column No.
1-4
11-20
1-4
6-10
1-80
1-4
6-7
9
11-14
16-40 '
41-50
Variable
Name
NAME(L)
YAX(L)
NTEST
NTIM
FTIME(J.L)/
FLOW(J.L)
NTEST
NDAT
JSYMB
JFACT
JCONC(J)
CLOW
Item Units of
Format Measure
A4
El 0.0 ppm
A4
15
4(2F10.0) min ppm"1
A4
12
Al
A4
5(A4,1X)
F10.0 ppm
51-60
61-70
CHIGH
TLOW
F10.0
F10.0
                                       ppm
	Comments	
Species name
Species initial concentration
Flowing species name
Number of flow points (maximum 10}
Time of measurement and concentration of
flowing species
Name of species to be plotted
Number of input data for the plot (maximum 80)
Symbol to be used for the calculated data
Conversion factor for the label
Concentration labels for the y-axix
Minimum concentration value to be considered
for plotting
Maximum concentration value to be considered
for plotting
Minimum time value to be considered for
plotting
                                                                                                                 ro
                                                                                                                 IX>
                                                                                                                 CO

-------
                                            Table A-l   (Concluded)
                         Variable
Card No.   Column No.      Time
 10
 11
            71-80
1-80
1-4
            THIGH
TIME(J)/
DATA(J)

JBLANK
               Item
              Format

              F10.0
                                      Units  of
                                      Measure
4(2F10.0)   min ppfli
A4
                                                          -1
                                                Comments
Maximum time value to be considered for
plotting

Time and concentration input data to be
plotted

Blank in Columns 1-4 stops plotting
                                                                                                                            ro

-------
                                                                      225
     The  first  control  card contains title and parameter information for the
run.   The first field on this card is a 12-character title; the contents of
this  field will  be printed following "MODULAR KINETICS RUN NO." on the first
page  of the printout.  The number of reactions is given next; the program
expects this number of reaction cards to follow the control card.  The next
entry specifies how many of these reactions represent lumped reactions and
thus  need to be recalculated from sets of contributing reactions.  The fol-
lowing five entries are the counts of each of the different types of species:
differential, steady state, uncoupled, replacement, and inert.   Note that
there are limits on both the number of differential species and the total num-
ber of species.  The program expects to find one species card for every species
named on  the reaction cards; they must be ordered as shown above (i.e., all
differential species first, then all steady-state species, and so forth).  The
next-to-last entry on the control card is the number of species that are flow-
ing into  the reaction chamber; there must'be a set of flow cards for each of
these species.   The final entry is a request flag governing the printout of the
reaction  rates.

     The  first  two entries on the second control card are printout parameters.
The first one determines the time increments (e.g., every five minutes) for
which the current concentration of all species are to be printed and plotted;
the second specifies the time at which the kinetics run is to be terminated.
The next  four values are control parameters for the differential equation solu-
tion  routine.  In order, these parameters are the initial time step (normally
on the order of 10   min), the minimum allowable time step (normally about 10"
min), the maximum time step (about 1 to 10 min), and-the fractional error accep-
table for iterative solutions.  The final entry on the second control  card is
the rate  at which each species concentration would be reduced in the absence of
reaction.   This "dilution rate" primarily reflects the loss of material through
sampling;  if there is an inflow, it is presumed to occur at this same rate.

     The  set of reaction cards provides all the reactions and rate constants,
one per card.  Each card begins with a list of reactantS's which must appear in
consecutive fields, since a blank stops the scan.  For a lumped reaction, the

-------
                                                                       226
name of the species to be replaced must appear first; otherwise, the order of
reactant names -is immaterial.   Note, incidentally, that the reactants appear
in the printed output in reverse order.  If a species reacts with itself, it
must appear twice in the list of reactants.  The products, along with their
coefficients, follow.  Coefficients can be whole numbers or fractions; the
printout is rounded to two decimal places.  Again, products must appear con-
secutively, since a blank stops the scan, but their order is unimportant.
The final  entry on the card is the rate constant.   The order of the reaction
cards does not matter, except that lumped reactions must follow nonlumped
reactions.

     A set of contributing reactions consists of an identification card con-
taining the name of the lumped reactant (the species being replaced)  and the
number of contributing reactions, followed by the  list of contributing reac-
tions.  All of the comments offered above regarding reaction cards apply to
these contributing reactions, except that these reactions cannot themselves
be lumped ones.   The contributing reaction must have its reactants and prod-
ucts in the same relative location on the card as  they are on the lumped
reaction card; i.e., the replacement species must  appear first, and all  prod-
ucts must be shown, even those that -have zero coefficients.  However, the
order of the reactions within a set does not matter.  The order of the sets
of contributing  reactions must be the same as the  order of the lumped reac-
tions in the set of reaction cards described above, and there must be one
set of contributing reactions for each lumped reaction.

     The set of  species cards is used to identify  the species by type and to
initialize the species concentrations.  Each card  contains a species  name and
concentration.  The following is the order of the.species types:  differential,
steady state, uncoupled, replacement, and inert.   Within a given type, no par-
ticular order is necessary; in fact, some orderings of steady-state species
are clearly preferable to others in terms of elapsed computing time.

-------
                                                                       227
     A set of flow cards consists of an identification card containing the
species name and the number of input points, followed by cards specifying
the data points themselves.  The data are not interpolated; instead, the
inflowing concentration is changed to a new value whenever the progression
of time in the mechanism passes an input time.  Note particularly that the
concentration for the inflowing species is zero until its first input time
is passed.

     The plot cards control the pictorial representation of concentration as
a function of time for each species.  Instead of processing the output using
a plotter, the plot cards map the concentration-time profile onto a page-size
grid of the printout.  The first card, which is the plot control  card, con-
tains the species name, the number of experimental  data points to be read, the
symbol to be used to represent the calculated data (an asterisk is used for
experimental points), the conversion factor, the concentration labels, and the
grid limits.  These last three items require comment.

     The grid has been divided into four vertical sections and eight horizontal
sections.  Aesthetically, therefore, the time (horizontal) limits should be
chosen to give a span divisible evenly by eight (e.g., a limit of 0 to 400 will
result in the printing of a label every 50 minutes).  Similarly,  the concentra-
tion (vertical) limits should be divisible evenly by four, and they should be
the true rather than the scaled concentrations.  The labels for the vertical
axis are not calculated from the concentration span, but rather are read in
from the control card.  They can be any multiple of the true concentrations.
The scale factor, which appears in the figure caption along with  the run title
and species values, indicates what multiplier was used.  For example, if the
data prints were expected to range between 0.08 and 0.16 ppm, a scale factor of
"10+1"; concentration labels of "0.75", "1.00", "1.25", "1.50", and "1.75";
and concentration limits of "0.075" and "0.175" would give a plot containing
all  the points.  Note that no check is made among the labels, scale factor, and
limits to insure consistency.  Also, the limit values themselves  will not appear
on the plot;  plotted points must fall  within the grid boundaries.

-------
                                                                       228
     Experimental  data cards, if any, come after the control card.  If desired,
sets of plot cards can be stacked.   The end of the plot deck is denoted by a
card with a blank  species name; the program will then expect another MODKIN
run control card.

3.   PROGRAM DESCRIPTION

     The modular kinetics program consists of a main routine labeled MODKIN,
the subroutines LMPCAL, DIFSUB, and PLOT, which are called by MODKIN, and the
subroutines DIFFUN, MATINV, and PEDERV, which are called by DIFSUB.   Each rou-
tine is treated in detail below.  Listings and samples of program inputs and
outputs appear at  the end of this appendix.  Symbol, glossaries are included
within each routine that was written especially for this program.

a.   MODKIN

     The program begins by declaring a number of variables used by DIFSU3 as
being DOUBLE PRECISION.  All arrays are identified in DIMENSION statements and
variables needed by LMPCAL and DIFFUN are placed in COMMON.  The DATA declara-
tions include the  input and output units, a blank word, and the maximum sizes
of the various arrays used for holding user inputs.

     The control cards are read (note that this is a return point for stacked
data decks).  An initial page is written listing all the control card param-
eters.  The number of reactions is  checked, and the set of reaction cards is
read.  The number  of lumped reactions is checked, and the contributing sets
are read.  For each set, the number of contributing reactions and the lumped
species name are checked, the reaction counter is incremented and checked,
and the contributing reactions are  read.  The number of differential species
is checked, the total  number of species is calculated and checked, and the set
of species cards is read.  The number of flowing species is checked, the flow
variables are cleared to zero, and  the flow cards are checked and read.

-------
                                                                      229
     The  numerical  identifiers for each species and the uncoupled species
reaction  rates  are  cleared to zero.   The initial  concentrations of the
differential  species  are set to their input values, and the input errors
and initial  maxima  are set.   The numerical  species indentifiers for the
reactants and products of each reaction are cleared to zero.

     The  counters  for the lumped reactions  are set.  The reactions are ana-
lyzed,  and each species is identified;  if it does not match a species  list
name, a flag is set.   Numerical  identifiers are placed in the reaction and
species matrices to allow reference  by  reaction and location  within the
reaction  expression.   If this is the first  of a set of replacement reactions,
a message is printed.  The order of  the reactants is reversed, and the list
of chemical  reactions and rate constants is printed.  LMPCAL  is called to
adjust  the lumped  reactions  and species concentrations, and the list of
initial species concentrations,  broken  down by. type, is printed.   If the
name flag is set,  processing halts.   A  number of computation  parameters are
initialize-ek   Initial concentrations are saved, and the incoming  concentra-
tions of  any flowing  species are initialized.

     A  call  is  made to the differential equation solution routine DIFSUB
(which  in turn  calls  DIFFUN  for solution of the steady-state  equations);
note that this  is -a return point from the calculation loop.  The  time  values
are updated, and the  concentration values are saved and checked for negative
values.  If a negative value is  found,  the  time step is reset to  one-tenth
of its  former value (note that this  is  done only once and that the time step
must be greater than  the user-specified minimum value), the concentrations
from the  previous  time step  are restored, and the call to DIFSUB  is repeated.

     The  replacement  species concentrations are calculated using  the following
algorithm:
                                      J  /    K
Y-''^ = Yi(o) exp  '"' Z^\mxj  "   'k,
                                K=2
-ATV  RK,~|T Y.
                                                       (1  - QAT)

-------
                                                                       230
where

        Y.j(n\   =  the  new concentration  of replacement species  i,
        Y.j/Q\   =  the  old concentration  of replacement species  i,
        AT     =  the  change  in  time  since the previous calculation,
        RK.     =  the  rate constant of the j-th reaction of the set
          J'        of J contributing reactions that have species i  as
                  the  first reactant,
         Yk     =  the  concentration of the K-th reactant species in
                  the  set of  K reactants in contributing reaction  j,
         Q     =  the  dilution rate.

 LMPCAL is then called to adjust the  lumped reactions  and species  concentrations
 to reflect  the changes in the replacement species concentrations.

      The uncoupled species concentrations are calculated using the algorithm
 described below under DIFFUN.   Note  that the uncoupled species reaction  rates
 are averaged  with those derived during  the previous time step.  The  current
 uncoupled species reaction rates are saved.

      The time is  checked against having passed the user-provided  limit,  in
 which case  no plot points are saved. The incremental  time  since  the last
 printout is checked against  the user-specified value.   If appropriate, the
 current concentration values are saved  for plotting,  unless  the maximum  num-
 ber of plot points has been  exceeded.   The current concentrations  are then
 printed, regardless of whether  the plot points have been saved.   If  the  user
 has indicated that the reaction rates are to be printed, they  are  sorted from
 largest to smallest and are  listed five per  line.

      If the time  limit has been passed,  if a repeated  negative  concentra-
 tion has been encountered, or if an  error has occurred in DIFSUB,  the error
 flag is  printed and the PLOT routine is  called.   Control  then  passes back to
 the beginning of  the  program, where  another  set of input cards can be processed
 Otherwise, the  inflowing species are checked to  determine whether  any concen-
 trations  should be updatea;  if  any updates are made,  an  appropriate  message is
 printed.  Finally,  the current  time  and concentrations are  saved  (in case it
 is  necessary  to restart the  calculation  with a smaller time  step), and control
 returns  to the  call of DIFSUB for calculation of the  next time step.

-------
                                                                      231
b.    LMPCAL

    This subroutine is called by MODKIN to calculate the rate constants and
product species coefficients for the lumped reactions.  Two arguments are
provided to this routine by MODKIN:  the number of lumped reactions and the
number of contributing reactions for each lumped reaction.  COMMON is de-
fined as in MODKIN, array sizes are set with DIMENSION statements, and param-
eters are defined via DATA statements.

     The location and number of the current set of contributing reactions are
established, and the corresponding lumped reaction is identified.   The rate
constants,  product coefficients, and replacement species concentrations for
each contributing reaction are transferred to local arrays, and the replace-
ment species concentration is summed.   The concentration of the lumped species
is set to the sum of all the replacement .concentrations, and the mole fraction
of each replacement species is calculated.  The rate constant of the lumped
reaction is calculated as the sum of the rate constants for the contributing
reactions multiplied by the replacement species mole fraction.  The product
species coefficients are calculated as the sum of the coefficients for the
product species multiplied by the replacement species mole fraction for each
contributing reaction, weighted by the ratio of the rate constant for the
contributing reaction to that of the lumped reaction.  Note that as a final
step any product species coefficient below a minimum value is reset to zero
to avoid underflow problems in later computations.

c.    DIFSUB

     This subroutine, which is called  by MODKIN, is a copy of the program for
the integration of coupled first-order ordinary differential  equations that
was presented in the Collected Algorithms of the Association for Computing
Machinery (Gear, 1971).   Since the algorithm and program are described in de-
tail  in the cited reference, they are  not discussed further here.   The only
change likely to be needed is the alteration of the DIMENSION statement near
the beginning.

-------
                                                                       232
d.   DIFFUN

     This subroutine is called by DIFSUB to calculate the rates of change of
the differential  species.  It also includes, however, the algorithm for the
steady~-state calculations, since the reactions involving species in a steady
state are presumed to be fast relative to the time steps in DIFSUB and must
therefore be updated at every time trial.  The three arguments provided to
this subroutine by DIFSUB are the time, the differential species concentra-
tions, and the rates of change of these concentrations; these are all DOUBLE
PRECISION.  The arguments are sized in a DIMENSION statement, COMMON is de-
fined as in MODKIN, and several parameters are defined via DATA statements.

     The concentrations are transferred to a local array, and the convergence
loop is begun.  The reaction rate for each reaction is calculated by using
the following algorithm:
                            R  = RK  -fT Y.
                             1     ] j=l  J

where R. is the reaction rate and RK^ is the rate constant of the i-th reaction,
and Y. is the concentration of the'j-th reactant species in the set of J reac-
     \j
tants in reaction i.

     The steady-state concentrations are calculated by using the following
dynamic mass-balance algorithm:
                          I
                             R.C.-
                              i U
                 Y  =    '" '	         k J.  i
                 Yj        M       K             K f J
                      n 4- V vv  V  v
                      Q   mtl   m ^  Yk
where

-------
                                                                      233
               Y.
                J

       R.  and C . .
                J
           = the concentration of the j-th steady-state species,

           = the reaction rate of the i-th reaction and the co-
             efficient of species j in the i-th reaction, respec-
             tively, of the set of I reactions in which species j
             is a product,

           = the dilution rate,

           = the rate constant for the m-th reaction in the set of
             M reactions in which species j is a reactant,

        Y, = the concentration of the k-th reactant (except when
             k = j) in the set of K reactants in reaction m.
                Q

              RK
Note that  in  the  case  of  a  species  reacting  with  itself, Y. may be the same as

Yk even  though  k  ^  j ;  this  case  has  been  explicitly  programmed.


     If  the old and new steady-state values  agree within the  requisite tolerance,

a convergence counter  is  incremented;  in  any event,  the new value is saved.  If

the value  registered on the convergence counter equals the number of steady-

state species,  the  loop is  completed;  othei^vise,  another pass  is made.   If the

steady-state  concentrations do not  converge, a warning message is written and

processing continues.


     As  a  final step,  the differentials are  calculated according to the  follow-

ing algorithm:
            I
                               M
                  1J
           AY   =      R.C.,  -      R   +  Q(YF,  -  Y  )
                              m=l
                                   m
where
AY.
  R.  and C. .  =
   1       1J
  •R- ••->••
   m
                       the  change  in  concentration  of  the  j-th  species
                       with  time,

                       the  reaction rate  of the  i-th reaction  and  the
                       coefficient of species  j  in  the i-th  reaction,
                       respectively,  of the set  of  I reactions  in  which
                       species  j is a product,

                       the  reaction rate  of the  m-th reaction  in  the  set
                       of M  reactions in  which species j  is  a  reactant,

-------
                                                                       234
        Q          =  the dilution rate,
        YF.        =  the inflowing concentration of species j,
          J
        Y.         =  the concentration of species j.
         J

These differential  values are returned to the calling program.

e.   MATINV

     This subroutine is called by DIFSUB to perform matrix inversions.   Since
it is a standard matrix inversion routine taken from the utility subroutine
library at the California Institute of Technology, it is not described  here.
The only change likely to be needed is the alteration of the DIMENSION  state-
ment near the beginning.

f.   PEDERV

     This subroutine,-which is called by DIFSUB, is used to provide a Jacobian
matrix for the calculation of partial derivatives and is not necessary  in this
application.   Nevertheless, to preserve the integrity of the Gear routine and
to allow for  possible future use of partial derivatives, we retained the rou-
tine in the program as a dummy subroutine.  It does contain, however, a DIMENSION
statement; thus, any alteration of array sizes in DIFSUB should also be made  in
PEDERV.

g.   PLOT

     This subroutine is called by MODKIN following completion of the time-
concentration calculations.   It maps the results, along with any user input
data, onto a  page-sized grid of concentration as a function of time cells for
as many species as  the user wishes.

     The routine begins by providing DIMENSION statements for some of the argu-
ments and the local  arrays.   The vertical axis and vertical label are estab-
lished via DATA statements,  as are the I/O units, some symbols, and the maximum
array sizes.

-------
                                                                       235
     The  grid is  cleared  to  blanks,  and the control  card is read.   If the
name is  blank, control  returns  to MODKIN.   If there  are input data, the num-
ber is  checked and  the  data  are read.   The normalization factors are calcu-
lated,  and the numerical  labels are  placed on the axes.  The species is
identified;  if it is  misspelled, the plot is skipped.

     The  data points, if  any,  are scaled to the grid;  the points are checked;
and, ifithey are  acceptable, the appropriate symbol  is placed on the grid.
The same  procedure  is used for  the calculated points.   A page is skipped, and
the vertical labels and axis and the grid itself are printed.  The horizontal
axis and  labels and the figure  caption are printed,  and the routine returns
control  to MODKIN.

4.   LISTINGS AND SAMPLE  REPORT AND  OUTPUT

     The  following  pages  contain a complete listing  of the computer program,
including the main  routine MODKIN (Exhibit A-l) and  the subroutines LMPCAL,
DIFSUB-r DLFFUN, MATINV, PEDERV, and  PLOT (Exhibits A-2 through A-7).  These
routines  are all  written  in  ASA FORTRAN and should be  acceptable without
changes  for  any computer  system that supports FORTRAN.

     Following the  program listings  is an input deck describing a  typical
kinetics  mechanism  (Exhibit  A-8) and selected printout from the computer
run using this input  deck (Exhibit A-9).  We note that, except for the plot
(which  uses  an entire 11  x 15  inch computer printout page), the output is
contained on a standard 8^ x 11 inch page.   With relatively minor changes in
the PLOT  routine, the plot can  also  be reduced to this size.

-------
                     EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN
                                                                        236
C MAIN PROGRAM *«•«•*** MODKIN
C
C THIS PROGRAM READS  AND  ANALYZES INPUT FOR MODULAR KINETICS PROGRAM.
C  SETS INITIAL VALUES,  CONTROLS ITERATION PRINTOUT, AND CALLS THE
C DIFFERENTIAL EQUATION  SOLVING AND PLOT ROUTINES.
C
C WRITTEN BY D. C.  WHITNEY FOR SYSTEMS APPLICATIONS, INC.
C ORIGINAL DATE 31  AUGUST 1973, LATEST MODIFICATION 25 OCTOBER 1973.
C THIS PROGRAM AND  ALL SUBROUTINES (EXCEPT DIFSU8 AND MATINV) ARE THE
C PROPERTY OF AND COPYRIGHT BY SYSTEMS APPLICATIONS. INC.
C 950 NORTHGATE DRIVE; SAN RAFAEL, CALIFORNIA 94903.
C
C SYMBOL DESCRIPTIONS —
C
C COEFF    NUMBER OF  PARTICLES, ONE PER PRODUCT SPECIES PER REACTION
C DELT     DIFFERENCE BETWEEN TWO CONSECUTIVE TIME STEPS
C EPS      CONVERGENCE CRITERION, DOUBLE PRECISION, FOR DIFSUB
C EPSF     CONVERGENCE CRITERION
C ERROR    ESTIMATE OF ERROR IN CONCENTRATIONS, PPM, ONE PER SPECIES,
C              DOUBLE PRECISION, FOR DIFSUB
C ESUM     SUM OF THE EXPONENTIAL TERMS'FOR THE REPLACEMENT SPECIES
C ETERM    TERM IN  THE EXPONENTIAL CALCULATION OF THE REPLACEMENT SPEC.
C FLOW     SPECIES  INFLOWS, PPM/MIN,  10 PER SPECIES
c FTEST    TEMPORARY  FLOW TIME OP CONCENTRATION -FOR TESTING, MIN OR PPM
C FTIME    TIMES AT WHICH INFLOW IS MEASURED, MIN, 10 PER SPECIES
C H        NEXT STEP  SIZE, MIN, DOUBLE PRECISION, FOR DIFSUB
C HMAX     MAXIMUM  TIME  STEP, MIN, DOUBLE PRECIS-ION, FOR DIFSUB
C HMAXF    MAXIMUM  TIME  STEP, MIN
C HMIN     MINIMUM  TIME  STEP, MIN, DOUBLE PRECISION, FOR DIFSUB
C HMIN     MINIMUM  TIME  STEP, MIN.
C HSTART   INITIAL  STEP  SIZE FOR DIFSUB
C J        DO-LOOP  INDICES OR LOCAL POINTERS
C JBLANK   A HOLLERITH WORD OF FOUR BLANK CHARACTERS
C JFLAG    INDICATES  NEGATIVE SPECIES FOUND IN DIFFERENTIAL CALCULATION
C JSTART   INPUT FLAG, FOR DIFSUB
C K        DO-LOOP  INDICES OR LOCAL POINTERS
C KCOF     COEFFICIENT POINTERS, ONE  PER REACTION PRODUCT PER SPECIES.
C KFLAG    PERFORMANCE FLAG FOR DIFSUB
C KLMP     NUMBER OF  CONTRIBUTING REACTIONS TO EACH LUMPED REACTION
C KLOC     LOCAL POINTER  TO CONTRIBUTING REACTION
C KPRD     PRODUCT  POINTERS, ONE PER  PRODUCT SPECIES PER REACTION
C KRCT     RFACTANT POINTERS, ONE PER REACTANT SPECIES PER REACTION
C KRXN     REACTION POINTERS, ONE PER REACTION PER SPECIES
C L        DO-LOOP  INDICES OR LOCAL POINTERS
C LFLAG    INDICATES  INCORRECT SPECIES NAME IN REACTION -- STOPS JOB
C LRXN     POINTER  TO FIRST OF A SERIES OF REPLACEMENT REACTIONS
c M        DO-LOOP  INDICES OR LOCAL POINTERS
c MAXDER   MAXIMUM  ORDER  FOR DERIVATIVES, FOR DIFSUB
c MAXDIF   MAXIMUM  NUMBER OF DIFFERENTIAL SPECIES
c MAXFLW   MAXIMUM  NUMBER OF FLOW TIMES
C MAXLMP   MAXIMUM  NUMBER OF LUMPED REACTIONS
c MAXPNT   MAXIMUM  NUMBER OF SAVED TIME AND CONCENTRATION POINTS
c MAXPRD   MAXIMUM  NUMBER OF PRODUCTS
C MAXPRT   MAXIMUM  NUMBER. OF ENTRIES  ON PRINT LINE FOR RATES
ooooooio
00000020
00000030
00000040
00000050
00000060
00000070
OOOOOORO
00000090
00000100
00000110
00000120
00000130
00000140
00000150
00000160
00000170
000001RO
00000190
00000200
00000210
00000220
00000230
00000240
00000250
000002^0
00000270
00000200
00000290
00000300
00000310
00000320
00000330
00000340
00000350
00000360
00000370
000003*0
00000390
00000400
00000410
00000420
00000430
00000440
00000450
00000460
00000470
000004RO
00000490
00000500
00000510
00000520
00000530
00000540

-------
               EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN  (Continued)
                                                                        237
  MAXRCT
  MAXREP
  MAXPXN
  MAXSPC
  MF
  MRXN
  N
  NAME
  NDIF
  NFLW
  NIN
  NINT
  NLMP
  NLOC
  NMPD
  NMRC
  NOUT
  NP
  NPNT
  NPRT
  NR
  NRAT
  NREP
C NRESET
  NRXN
  NS
  NSTS
C NTEST
C NTIM
C NTIT
C
C
C
C
C
C
  NTOT
  Nil
  NUNC
  PSAVF.
  Q
  R
C RATE
C
C
C
C
C
  RK
  RPRT
  SAVE
  SAVCON
  SAVTIM
C T
C
C TCOUNT
C TEND
C TF
C TINCR
C TOL
  UNCOLD
  TOLD
  Y

  VAX
          MAXIMUM NUMBER  OF  PEACTANTS
          MAXIMUM NUMBER  OF  REPLACEMENT REACTIONS PER LUMPED REACTION
          MAXIMUM NUMBER  OF  REACTIONS
          MAXIMUM NUMBER  OF  SPECIES
          MFTHOD INDICATOR*  FOR  DIFSUB
          TOTAL NUMRER  OF  REACTIONS,  INCLUDING REPLACEMENTS
          DO-LOOP INDICES  OR LOCAL POINTERS
          SPECIES NAMES,  ONE PER SPECIES
          NUMBER OF  DIFFERENTIAL SPECIES
          NUMRER OF  FLOWING  SPECIES
          THE FORTRAN INPUT  UNIT (NORMALLY  5)
          NUMBER OF  INERT/CONSTANT SPECIES
          NUMBER OF  LUMPED REACTIONS
                                                              00000550
                                                              00000560
                                                              00000570
                                                              000005*0
                                                              00000590
                                                              00000600
                                                              00000610
                                                              00000620
                                                              00000630
                                                              00000640
                                                              00000650
                                                              00000660
                                                              00000670
C YCALC
C YIN
NUMBER OF CONTIPIJTING  REACTIONS  PERTAINING  TO  LUMPED  REACTIONO00006P0
PRODUCT N6MES, ONE PER  PRODUCT SPECIES  PER  REACTION           00000690
REACTANT NAMES* ONE  PEP REACTANT SPECIES  PER REACTION        00000700
THE FORTRAN OUTPUT UNIT NUMBER  (NORMALLY  6)                   00000710
LOCAL POINTER TO REACTION RATE TO  BE  PRINTED                  00000720
NUMBER OF SAVED TIMES  AND CONCENTRATIONS                      00000730
HOLDING AREA TO PRINT  OUT A  LINE OF NAMES OR NUMBERS          00000740
POINTER TO REPLACEMENT  SPECIES                                00000750
USER INPUT FLAG REQUESTING PRINT OF REACTION RATES            00000760
NUMRER OF REPLACEMENT  SPECIES                                 0.0000770
COUNTER FOR NUMBER OF  TIMES  STEP,SIZE IS  RESET SMALLER        00000780
NUMBER OF REACTIONS                                           00000790
POINTER TO REACTING  SPECIES                                   00000800
NUMBER OF STEADY-STATE  SPECIES                                OOOOOR10
SPECIES NAME FOR TESTING                                      00000820
NUMBER OF TIMES AND  FLOWS FOR A  SPECIES                      00000030
USER-INPUT TITLE FOR PRINTOUT, 3 FOUR-CHARACTER WORDS        ooooo«4o
TOTAL NUMBER OF SPECIES                                       00000850
LOCAL POINTER TO UNCOUPLED SPECIES                            00000860.
NUMBER OF UNCOUPLED  SPECIES                                   00000870
BLOCK STORAGE, NUMBER  OF SPECIES SQUARED, FOR  DIFSUB          000008RO
DEGRADATION RATE, /MIN                                        00000890
REACTION RATES, SEC, ONE PER REACTION                        00000900
LOCAL REPRESENTATION OF R, THE REACTION RATE                  00000910
REACTION RATE CONSTANTS, PPM-MIN,  ONE PER REACTION            00000920
HOLDING AREA TO PRINT  OUT A  LINE OF REACTION PATES            00000930
BLOCK STORAGE, 12 PER  SPECIES, DOUBLE PRECISION,  FOR  DIFSUB   00000940
SPECIES CONCENTRATIONS, PPM, ONE PER  SPECIES AT 80 TIMES      00000950
TIMES THAT CONCENTRATIONS ARE SAVED,  MIN, UP TO 80 VALUES     00000960
CURRENT REACTION TIME,  SEC,  DOUBLE PRECISION,                 00000970
    FOP AND FROM DIFSUB                                       000009RQ
NEXT TIME FOR OUTPUT,  MIN                                     00000990
ENDING TIME, MIN                                              00001000
PREVIOUS TIME OF DIFSUB CALL, MIN                             00001010
TIME INCREMENT FOR OUTPUT, MIN                                00001020
CONVERGENCE TOLERANCE  ON STEADY-STATE ITERATION,  PPM          00001030
PREVIOUS VALUES OF RATE OF CHANGE, PPM/MIN, ONE PER  SPECIES   00001040
TIME OF PREVIOUS CALL  TO DIFSUB                               00001050
SPECIES CONCENTRATIONS, 8 PER SPECIES,  DOUBLE  PRECISION,      00001060
    FOR AND FROM DIFSUB                                       00001070
SPECIES CONCENTRATIONS, PPM, ONE PER  SPECIES                  00001080
LOCAL REPRESENTATION OF YDOT, THE  RATE  OF CHANGE             00001090
SPECIES INFLOW RATES,  PPM/MIN, ONE PER  SPECIES               00001100

-------
                EXHIBIT A-l.   LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                         238
C YMAX      CURRENT  MAXIMUM CONCENTRATION VALUES, PPM, ONE-PER SPECIES,   00001110
C              DOUBLE  PRECISION, FOR AND FROM DIFSUB                     00001120
C YOLD      CONCENTRATIONS AT PREVIOUS CALL TO DIFSUB, ONE PER SPECIES    00001130
c YUNC      RATE OF  CHANGE OF UNCOUPLED SPECIES, PPM/MIN                  00001140
c                                                                        ooooiiso
C BEGINNING OF  PROGRAM.                                                   00001160
C                                                                        00001170
C DECLARE  VARIABLES FOR  DIFSUB AS DOUBLE PRECISION                       000011«0
C                                                                        00001190
     DOUBLE PRECISION HMIN» HMAX, EPS, YMAX, ERROR, H, SAVE, T, Y       00001200
C                                             .                           00001210
C DEFINE  VARIABLES  AND DIMENSIONS OF COMMON STORAGE WITH DIFFUN          00001220
C                                                                        00001230
     COMMON RK(99), R(99), YAX(50), YIN(50), COEFF(3999)                00001240
     COMMON KPCT(4,Q9)-i KPPD(3,99), KRXN(99,50), KCOF(99,50)            00001250
     COMMON Q, TOL, NRXN, NDIF, NSTS                                    00001260
C                                                                        00001270
C DEFINE  DIMENSIONS OF LOCAL ARRAYS                                      000012*0
C                                                    .                    000012^0
     DIMENSION Y(8,40), YMAX(40), SAVE(12,40), ERROR(40), PSAVE(1600)   00001300
     DIMENSION NMRC(4j99), NMPD(3,99)» NTIT(3), NPRT(IO), RPRT(IO)      00001310
     DIMENSION FTIME (10,50) »  FLOW(10»50)» SAVCON(50,80) , SAVTIM(80)     00001320
     DIMENSION YOLDC50), NAME(50), UNCOLD(50), KLMP(IO)                  00001330
C                                                                        0'0001340
C SET MISCELLANEOUS DATA VALUES                                          00001350
C                                                                        00001360
     DATA MAXRCT  /4/? NIN /5/, NOUT /&/, MAXPRD /3/, JBLANK MH    /    00001370
     DATA PTAXRXN  /99/«  MAXDIF /40/. MAXFLW /10/, MAXPNT /80/            000013/30
     DATA MAXSPC  /BO/,  MAXLMP /10/, MAXREP /10/, MAXPRT /5/             00001390
C                                                                        00001400
C READ  CONTROL  CARDS —  NOTE THIS IS RETURN POINT FROM PLOT CALL         00001410
C                                                                        00001420
10   READ (NIN,2,ENO=900) NTIT, NRXN, NLMP, NDIF, NSTS,  NUNC, NREP,     00001430
     fi,      NINT, NFLW, NRAT                                              00001440
     READ (NIN,4)  TINCR, TEND, HSTART, HMINF, HMAXF, EPSF, Q            00001450
C     '                                                               •    00001460
C PRINT HEADING PAGE AND CONTROL CARD INPUTS                             00001470
C                                                                        00001480
     WRITE (NOUT51002)  NTIT,  NRXN, NLMP, NDIF, NSTS, NUNC, NREP, NINT,  00001490
     &      NFLW, NRAT, TINCR,  TEND, HSTART, HMINF, HMAXF, EPSF, Q        00001500
C                                                                        00001510
C TEST  REACTION COUNT                                                     00001520
C           •                                                             00001530
     IF  (NRXN  .GT. 0  .AND. NRXN .LE. MAXRXN) GO TO 12                   00001540
     WRITE (NOUT,1001)  MAXRXN                                           00001550
     GO  TO 900                                                          00001560
C                                                                        00001570
C SET SPECIES NAME  FLAG  AND OVERALL REACTION COUNT AND READ REACTIONS    000015RO
C                                                                        00001590
12   LFLAG = 0                                                          00001600
     MRXN = NRXN                                                         00001610
     DO  15 K = ItNRXN                                                   00001620
     READ (NIN,1)  (NMRC(J,K), J = 1,MAXRCT),                            00001630
     &      (COEFF(J,K)»  NMPD(J,K), J = 1,MAXPRD), RK(K)                  00001640
15   CONTINUE                                                            00001650
C                                                         .               00001660

-------
              EXHIBIT A-l.  LISTING OF MAIN PROGRAM M.ODKIN (Continued)
                                                                      239
TEST NUMBER OF LUMPED REACTIONS

    IF (NLMP .LE. 0)  GO TO 22
    IF (NLMP .LE. MAXLMP) 60 TO 16
    WRITE (NOUT*1027) MAXLMP
    GO TO 900

READ AND TEST LUMPED SPECIES NAME AND NUMBER  OF  REPLACEMENT SPECIES

 16 DO 21 L = 1tNLMP
    READ (NIN,6)  NTEST, NLOC
    IF (NTEST .EO. NMRCd.NRXN - NLMP + L) ) GO TO  17
    WRITE (NOUT;1028) NTEST
    LFLAG = 1
 17 IF (NLOC .GT. 0 .AND. NLOC .LE. MAXREP) GO TO  18
    WRITE (NOUT»1030> MAXREP
    GO TO 900

SAVE NUMBER OF REPLACEMENT REACTIONS AND UPDATE  AND CHECK  TOTAL

 18 KLMP(L)  = NLOC
    MRXN = MRXN + NLOC
    IF (MRXN .LE. MAXRXN) GO TO 19
    WRITE (NOUT;100l) MAXRXN
    GO TO 900

READ REPLACEMENT  REACTIONS

 19 E>0- 2-9-- K- = 1*NLOC
    KLOC = K * MRXN - NLOC
    READ (NIN»1)  (NMRC(J,KLOC)* J = 1»MAXRCT), (COEFF(J9KLOC)*
   Si     NMPD(J»KLOC) t J = l.MAXPRDJ* RK(KLOC)
 20 CONTINUE
 21 CONTINUE

TEST NUMBER OF DIFFERENTIAL SPECIES

 22 IF (NDIF .LE. MAXDIF) GO TO 23
    WRITE (NOUT»1012) MAXDIF
    GO TO 900

SET AND TEST TOTAL NUMBER OF SPECIES

 23 NTOT = NDIF + NSTS + NUNC + NREP + NINT
    IF (NTOT .LE. MAXSPC) GO TO 25
    WRITE (NOUT.1011) MAXSPC
    GO TO 900
    IF (NFLW .LE. 0)  GO TO 50
    IF (NFLW .LE. NTOT)  GO TO 30
00001670
00001680
00001690
00001700
00001710
00001720
00001730
00001740
00001750
00001760
00001770
00001780
000017QQ
00001800
00001810
00001820
00001830
00001840
00001850
000018^0
00001870
00001880
0-0001890
00001900
00001910
00001920
00001930
00001940
00001950
00001960
00001970
00001980
00001990
00002000
00002010
00002020
00002030
00002040
00002050
00002060
00002070
000020PO
00002090
00002100
00002110
00002120
00002130
00002210
00002220

-------
                                                                          240
                EXHIBIT A-1.   LISTING OF MAIN  PROGRAM MODKIN (Continued)


      WRITE (NOUT, 1023)  NTOT                                              00002230
      GO TO 900                                                            00002240
C                                                                          00002250
C SET  FLOW PATES AND TIMES  TO  ZERO                                         000022f>0
      IF (NFLW .L.F.. 0) GO TO 50                                            00002210
      IF (NFLW .LE. NTOT) GO TO  30                                         00002220

-------
                EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                      241
      WRITF (NOUT,  1023)  NTOT
      GO  TO 900
C
C SET FLOW  RATES  AND  TIMES TO ZERO
C
30    DO  33 K  =  1»NTOT
      DO  32 J  =  1,MAXFLW
      FTIME(J,K)  =  0.0
      FLOW(J.K)  = 0.0
32    CONTINUE
33    CONTINUE
C
C READ FLOW CONTROL CARD
C
      DO  45 K  =  l.NFLW
      READ  (NIN,6)  NTEST,  NTIM
C
C IDENTIFY  SPECIES  NAME — EXIT IF NOT FOUND
C
      DO  35 L  =  1»NTOT
      IF  (NTEST  .EG.  NAME(D)  GO TO 40
35    CONTINUE
      WRITE (NOUT;1003> NTEST
      GO  TO 900
C
C CHECK NUMRER OF FLOW  INPUTS
C
40    IF  (NTIM .LE. MAXFLW)  GO TO 42
      WRITE (NOUT»1024) MAXFLW
      GO  TO 900
C
C READ FLOW INPUTS
C
42    READ  (NIN,7)  (FTIME(J»L)» FLOW(J,L), J = 1»NTIM)
45    CONTINUE
C
C CLEAR REACTION  POINTERS, FLOWS, AND UNCOUPLED RATES
C
50    DO  60 K  =  1,NTOT
      DO  55 J  =  1,MRXN
      KRXN(J,K) = 0
      KCOF(J,K) = 0
55    CONTINUE
      YIN(K) = 0.0
      UNCOLD(K) = 0.0
60    CONTINUE
C
C MOVE INITIAL DIFFERENTIAL  CONCENTRATIONS AND SET ERRORS AND MAXIMA
C
      DO  65 J  = 1,NDIF
      Y(l.J) = YAX(J)
      ERROR(J) =  0.0
      YMAX(J)  = 1.0
65    CONTINUE
C
C CLEAR SPECIES POINTERS
00002230
00002240
00002250
00002260
00002270
00002280
000022«0
00002300
00002310
00002320
00002330
00002340
00002350
00002360
00002370
00002380
00002390
00002400
00002410
00002420
00002430
00002440
00002450
00002460
00002470
000024RO
00002490
00002500
00002510
00002520
00002530
00002540
00002550
00002560
00002570
00002580
00002590
00002600
00002610
00002620
00002630
00002640
00002650
00002660
00002670
00002680
00002690
00002700
00002710
00002720
00002730
00002740
00002750
00002760
00002770
00002780

-------
                EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                        242



70


80
90
C
c
C




c
c
c
DO 90 K = 1
DO 70 J = 1
KRCT(J,K) =
CONTINUE
DO 80 J = 1
KPRD(J,K) =
CONTINUE
CONTINUE

»MRXN
.MAXRCT
0

.MAXPRD
0



SET LUMPED REACTION COUNTERS

LRXN = NRXN
KLOC = 1
DO 190 M =
DO 130 L =


«• 1

1»MRXN
1»MAXRCT

IDENTIFY RFACTANT SPECIES --


                               FLAG IF MISSING

      NTEST = NMRC(L.M)
      IF  (NTEST  .EQ.  JBLANK)  GO TO 140
      DO  110 K =  1»NTOT
      IF  (NTEST  .EQ.  NAME(K))  GO TO 115
110    CONTINUE
      WRITE  (NOUT»1003)  NTEST
      LFLAG = 1
      GO  TO 130
C
C FILL IN REACTION  AND SPECIES POINTERS AND COUNTERS
C
115
     KRCT(L,M) = K
     DO 120 J = 1,MAXRXN
     IF (KRXN(J,K)  .EQ. 0)
120   CONTINUE
1P5   KRXN(J,K) = M
     KCOF(J.K) = -1
     CONTINUE
                            GO  TO  125
          PRODUCT  SPECIES  —  FLAG IF MISSING
130
C
C IDENTIFY
C
140    DO  170 L  =  1»MAXPRD
      NTEST = NMPDCL.M)
      IF  (NTEST  .EQ.  JBLANK)  GO TO 180
      DO  150 K  =  1«NTOT
      IF  (NTEST  .EQ.  NAME(K))  GO TO 155
150    CONTINUE
      WRITE  (NOUT.1003)  NTEST
      LFLAG = 1
      GO  TO 170
C
C FILL  IN REACTION  AND SPECIES POINTERS AND COUNTERS
C
155
160
     KPRD(L,M) = K
     DO 160 J = 1»MAXRXN
     IF (KRXN(J,K)  .EQ. 0)
     CONTINUE
                            GO  TO 165
00002790
00002800
00002810
00002820
00002830
00002S40
00002350
00002860
00002870
000028«0
00002890
00002900
00002910
00002920
00002930
00002940
00002950
00002960
00002970
00002980
00002990
00003000
00003010
00003020
00003030
00003040
00003050
00003060
00003070
00003080
00003090
00003100
00003110
00003120
00003130
00003140
00003150
00003160
00003170
00003180
00003190
00003200
00003210
00003220
00003230
00003240
00003250
000032^0
00003270
000032HO
00003290
00003300
00003310
00003320
00003330
00003340

-------
                EXHIBIT A-l.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                        243
 165   KRXN(J,K) = M                                                      00003350
      KCOF(J,K) = L                                                      00003360
 170   CONTINUE                                                           00003370
 C                                                                        000033RO
 C SAVE NUMBER OF PRODUCT SPECIES FOR THIS REACTION                       00003390
 C                                                                        00003400
      L = MAXPRD + 1                                                     00003410
 180   NPRO = L - 1                 •                                      00003420
 C                                                                        00003430
 C IF REPLACEMENT FOR LUMPED REACTION, PRINT. MESSAGE  AND  UPDATE POINTERS 00003440
 C                                                                        00003450
      IF (M .NE. LRXN) GO TO 183                                         00003460
      N = NRXN - NLMP + KLOC                                             00003470
      WRITE (NOUT<10?9) KLMP(KLOC). N                                    00003480
      LRXN = LRXN + KLMP(KLOC)                                            00003490
      KLOC = KLOC «• 1                                                    00003500
 C                                                                        00003510
 C REVERSE ORDER OF REACTANTS FOR PRINTING                                00003520
 C                                                                        00003530
  183 DO 185 J = 1»MAXRCT                                                00003540
      K = MAXRCT - J * 1                                                 00003550
      NPRT(J)  = NMRC(K>M)                                                00003560
 185;  CONTINUE                                                           0.0003570
 C                                                                        00003580
 C PRINT SET OF REACTIONS                                                 00003590
 C                                                                        00003600
      WRITE (NOUT«1004> M» RK(M),  (NPRT(J), J=1*MAXRCT),                 00003610
     &.     (COEFF(J»M)» NMPD(JsM)» J = 1,NPRD)                           00003620
 190   CONTINUE                                                           00003630
 C                                                                        00003640
 C GET INITIAL  CONDITIONS FOR LUMPED REACTIONS AND SPECIES                00003650
 C                                                                        00003660
      IF (NLMP ,GT. 0) CALL LMPCALtNLMP. KLMP)                           00003670
 C                                                                        00003680
 C PRINT INITIAL SPECIES CONCENTRATIONS — EXIT IF FLAG SET               00003690
 C                                                                        00003700
   .   WRITE (NOUT;1013)                                                  00003710
      WRITE (NOUT»1005)                                                  00003720
      WRITE (NOUT.1006) (NAME(J)« YAX(J)» J=1»NDIF)                      00003730
      IF (NSTS .GT. 0)                                                   00003740
     &WRITE (NOUT»1025) (NAME(J+NDIF)» VAX(J+NDIF)»  J=1*NSTS)            00003750
      IF (NUNC .GT. 0)                                                   00003760
     &.WRITE (NOUT»1007) (NAME (J + NDIF+NSTS) »                              00003770
     5.     YAXCJ + NDIF + NSTS) , J = 1»NUNC)                                 000037RO
      IF (NRF.P .GT. 0)                                                   00003790
     &WRITE (NOUT?103l) (NAME(J+NDIF+NSTS+NUNC)9                         00003800
     81     VAX (J + NDIF + NSTS + NUNC) » J = 1»NREP)                             00003810
      IF (NINT .GT. 0)                                                   00003820
     &WRITE (NOUT;1008) (NAME(J+NDIF+NSTS+NUNC+NREP)t                    00003830
     81     YAXU + NDIF + NSTS + NUNC + NREP) » J = 1»NINT)                       00003840
      IF (LFLAG .EQ.-l) GO TO 900                                        00003850
C                                                                        00003860
C SET INITIAL  CONDITIONS                             '                    00003870
C                                                                        00003880
      HMAX  = HMAXF                                                       00003890
      HMIN  = HMINF                                                       00003900

-------
              .EXHIBIT A-l.  LISTING OF MAIN PROGRAM MODKIN  (Continued)
                                                                     244
    TCOUNT = TINCR
    EPS = FPSF
    TOL = EPSF
    H = HSTART
    .MF = 2
    JSTART = 0
    NPNT = 0
    TF = 0.0
    T = 0.0
    TOLD = 0.0
    MAXDER = 6
    JFLAG = 0
    KFLAG = 1
    NRESET = 0

SAVE INITIAL CONCENTRATIONS AND FLOWS

    DO 195 J = 1»MTOT
    YOLD(J) = YAX(J)
    IF (NFLW .EQ. 0) GO TO 195
    IF (FTIME(I.J)  .GT. 0.0) GO TO  195
    YIN(J) = FLOW(1»J)
195 CONTINUE

CALL DIFFERENTIAL SPECIES SOLVER — NOTE  THIS  IS  A  RETURN  POINT

200 CALL DIFSUBtNDIF, J, Y» SAVE* H? HMIN, HMAX,  EPS9  MF»
   &    YMAX, ERROR? KFLAG, JSTART* MAXDER, PSAVE)

UPDATE TIME AND SAVE CONCENTRATIONS, CHECKING  FOR NEGATIVITY

    TF = T
    DELT = TF - TOLD
    DO 210 J = 1,NTOT
    IF (J .LE. NDIF) YAX(J) = YU,J)
    IF (YAX(J) .LT.  0.0) JFLAG = 1
210 CONTINUE
    IF (JFLAG .EQ. 0) GO TO 230
    JFLAG = 0

NEGATIVE CONCENTRATION — RESET AND TEST  STEP  SIZE

    H = 0.1 * H
    IF (H .LT. HMIN) KFLAG = 0

TEST RESET COUNTER FOR RE-ENTRY, THEN SET TO PREVENT  SAME—TEST  FLAG

    IF (NRFSET .GT.  0)  KFLAG = 0
    NRESET = 1
    IF (KFLAG .NE. 1) GO TO 330

RESTORE OLD CONCENTRATIONS AND RECALL DIFSUB WITH SMALLER  STEP SIZE

    DO 220 J = 1,NTOT
    YAX(J)  = YOLD(J)
220 CONTINUE
00003910
00003920
00003930
00003940
00003950
00003960
00003970
00003980
00003990
00004000
00004010
00004020
00004030
00004040
00004050
000040*0
00004070
00004080
00004090
00004100
00004110
00004120
00004130
00004140
00004150
00004160
00004170
000041BO
00004190
00004200
00004210
00004220
00004230
00004240
00004250
00004260
00004270
00004280
00004290
00004300
00004310
00004320
00004330
00004340
00004350
00004360
00004370
00004380
00004390
00004400
00004410
00004420
00004430
00004440
00004450
00004460

-------
              EXHIBIT A-l.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                     245
    JSTART = -1
    GO TO ?00

TEST FOR AMD SET POINTERS TO REPLACEMENT SPECIES

230 IF (NREP .LE. 0)  GO TO 265
    LRXN = NRXN  + ]
    DO 260 M = 1,NREP
    NR = NDIF +  NSTS «• NUNC + M
    ESUM = 0.0

FIND REACTIONS CONTAINING REPLACEMENT SPECIES

    DO 250 L = LRXN» MRXN
    IF (KRCT(1»L) .NE. NR) GO TO 250

MULTIPLY TOGETHER ALL OTHER REACTANT CONCENTRATIONS

    ETERM = 1.0
    DO 240 K = 2»MAXRCT
    MS = KRCT(K»L)
    IF (NS .EG).  0)  GO TO 245
    ETERM = ETERM «• YAX(NS)
240 CONTINUE

MULTIPLY BY RATE CONSTANT AND ADD TO EXPONENTIAL  SUM

245 ESUM = ESUM  + RK(L) * ETERM
250 CONTINUE
CALCULATE NEW SPECIES CONCENTRATIONS

    YAX(NR)  = YAX(NR) * EXP(-ESUM * DELT)
260 CONTINUE
(1.0  - DELT
                                                           Q)
UPDATE LUMPED REACTIONS AND SPECIES PARAMETERS

    CALL LMPCALfNLMP* KLMP)

TEST FOR AND SET POINTERS TO UNCOUPLED SPECIES

265 IF (NUNC .LE,, 0)  GO TO 300
    DO 290 M = 1»NUNC
  ,  NU = NDIF + NSTS «• M
    YUNC = 0.0
    DO 270 L = ItNRXN
    J = KCOF(L.NU)
    K = KRXN(L»NU)

CALCULATE THE RATE  OF CHANGE OF THE UNCOUPLED SPECIES

    IF (J) 267. 280»  268
267 YUNC = YUNC - P(K)
    GO TO 270
268 YUNC = YUNC + R(K) * COEFF(J.K)
270 CONTINUE
00004470
000044RQ
00004490
00004500
00004510
00004520
00004530
00004540
00004550
00004560
00004570
00004580
00004590
00004600
00004610
00004620
00004630
00004640
00004650
00004660
00004670
00004680
Q0004690
00004700
00004710
00004720
00004730
00004740
00004750
00004760
00004770
00004780
00004790
00004800
00004810
00004820
00004830
00004840
00004850
00004860
00004870
00004880
00004890
00004900
00004910
00004920
00004930
00004940
00004950
00004960
00004970
00004980
00004990
00005000
00005010
00005020

-------
              EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                      246
280 YUNC = YUNC + 0 * (YIN(NU) - YAX(NU))

CALCULATE UNCOUPLFD SPECIES CONCENTRATION  AND  UPDATE  OLD  RATE

    YAX(NU)  = YAX(NU) * (YUNC «• UNCOLD(NU))  *  DELT  »  0.5
    UNCOLD(NU)  = YUNC
290 CONTINUE

CHECK TIME FOR END AND PRINTING AND SAVING OF  PLOT  POINTS

300 IF (TF .GT. TEND) GO TO 330
    IF (TF .LE. TCOUNT)  GO TO 340

INCREMENT TIME AND PLOT POINT COUNTERS

    TCOUNT = TCOUNT + TINCR
    NPNT = NPNT + 1

CHECK PLOT POINT COUNTER FOR OVERFLOW

    IF (NPNT .LE. MAXPNT)  GO TO 310
    WRITE (NOUT?1019) MAXPNT? TF
    NPNT = NPNT - 1
    GO TO 330

SAVE PLOT POINTS

310 SAVTIM(NPNT) = TF
    DO 320 J = 1»NTOT
    SAVCON(J,NPNT) = YAX(J)
3?0 CONTINUE

PRINT INTERMEDIATE RESULTS

330 WRITE (NOUT»1009) TF
    WRITE (NOUT-1005)
    WRITE (NOUT-1006) (NAME(J)» YAX(J)» J  =  1»NDIF)
    IF (NSTS .GT. 0)
   8.WRITE (NOUT'1025) (NAME (J+NDIF) « YAX(J+NDIF)» J=1»NSTS)
    IF (NUNC .GT. 0)
   &WRITE (NOUT;1007) (NAME(J+NDIF+NSTS)«
   &     YAXU + NDIF + NSTS) » J = 1»NUNC>
    IF (NREP .GT. 0)
   S.WRITE (NOUT»1031) (NAME ( J^NDIF + NSTS^NUNC) ,
   &     YAX(J+NDIF+NSTS+NUNC), J = 1,NREP)

CHECK RATE PRINT FLAG* PRINT HEADER? AND SET SORT PARAMETERS

    IF (NRAT .EQ. 0)  GO TO 339
    WRITE (NOUT;1032)
    N = 0
    RATE = -1.0
    NP = 0
    DO 338 M = 1»NRXN

FIND LARGEST RATE
000050 3D
00005040
'00005050
00005060
00005070
00005080
00005090
00005100
000051 10
000051PO
00005130
0000514-0
00005150
00005160
00005170
00005180
00005190
00005200
00005210
00005220
00005230
00005240
00005250
00005260
00005270
000052^0
00005290
00005300
00005310
00005320
00005330
00005340
00005350
00005360
00005370
00005380
00005390
00005400
00005410
00005420
00005430
00005440
00005450
00005460
00005470
000054BO
00005490
00005500
00005510
00005520
00005530
00005540
00005550
00005560
00005570
00005580

-------
              EXHIBIT A-l.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                       247
    DO 336 L = 1»NRXN
    IF (P(L> .LT. RATE) GO TO 336
    RATE = R(L)
    NP = L
336 CONTINUE

SAVE AND FLAG THIS RATE AND RESET SORT PARAMETERS

    N = N * 1
    NPRT(N) = NP
    RPRT(N) = RATE
    R(NP) = -1.0
    RATE = -1.0
    NP = 0

CHECK PRINT COUNT AND PRINT IF FULL LINE OR END  OF  LOOP

337 IF (N. LT. MAXPRT .AND. M .NE. NRXN) GO TO 338
    WRITE  (NOUT?1033)  (NPRT(K), RPRT(K)? K =  1,N)
    N = 0
338 CONTINUE

CHECK FOR ERROR  OR FINAL TIME PASSED? IF SO PLOT AND  GET  NEXT  SET

339 I-F C-T-F .LT.  TEND .AND. KFLAG .EG. 1) GO TO 340
    WRITE  (NOUT.1010) KFLAG
    CALL PLOTJNTIT,  NPNT, NTOT, NAME* SAVTIM, SAVCON)
    GO TO 10

CHECK FOR INFLOW UPDATES

340 IF (NFLW .LE. 0) GO TO 380
    DO 370 K = 1»NTOT
    DO 350 J = 1*MAXFLW
    IF (J ,EO. MAXFLW)  GO TO 360
    FTEST = FTIME(J+1»K)
    IF (FTEST .GT. TF)  GO TO 360
    IF (FTEST .LE. 0.)  GO TO 360
350 CONTINUE

UPDATE INFLOWS AND WRITE MESSAGE

360 FTEST = FLOW(J.K)
    IF (YIM(K) .EQ.  FTEST) GO TO 370
    YIN(K) = FTEST
    WRITE  (NOUT»10?6) NAME(K)* FTEST* TF
370 CONTINUE

UPDATE TIME AND  CONCENTRATION AND TAKE NEXT TIME STEP

380 TOLD = TF
    NRESET = 0
    DO 390 J = l.NTOT
    YOLD(J) = YAX(J)
390 CONTINUE
00005590
00005600
00005610
00005620
00005630
00005640
00005650
00005660
00005670
00005680
000056QO
00005700
00005710
00005720
00005730
00005740
00005750
00005760
00005770
00005780
00005790
00005800
00005810
00005820
00005830
00005840
00005850
00005860
00005870
00005880
00005890
00005900
00005910
00005920
00005930
00005940
00005950
00005960
00005970
00005980
000059QO
00006000
000060]0
00006020
00006030
00006040
00006050
00006060
00006070
00006080
00006090
00006100
00006110
00006120
00006130
00006140

-------
               EXHIBIT A-1.  LISTING OF MAIN PROGRAM MODKIN (Continued)
                                                                        248
                                A4), F10.0)
      GO TO 200
C
C END OF PROGRAM
C
900   STOP
C
C LIST OF FORMAT STATEMENTS
C
1     FORMAT (4(A4, IX), 3(F6.0,
2     FORMAT (344,  3X, 915)
3     FORMAT (A4, 6X,  F10.0)
    4 FORMAT (8F10.0)
6     FORMAT (A4, IX,  15)
.7     FORMAT (8F10.0)
 1001 FORMAT (33H PROGRAM CANNOT HANDLE MORE
     &     26H REACTIONS — JOB ABORTED.)
 1002 FORMAT (1H1,  20X, 25HMODULAR KINETICS
           29H TOTAL NUMBER OF REACTIONS = ,
               NUMBER  OF LUMPED REACTIONS = ,  I
                         DIFFERENTIAL SPECIES  =
                         STEADY STATE SPECIES  =
                                             THAN
                                                    14
    8,

    8,

    6,
    8,
    &
    &

    &
    8.
    &
    &
                                           RUN
                                            13,
30H
34H
34H
31H
33H
3QH
29H
36H
18H
15H
22H
21H
21H
25H
17H
NUMBER
NUMBER
NUMBER
NUMBER
NUMBER
NUMBER
OF
OF
OF
OF
OF
OF
     NO- , 3A4, ////,
      //,
     •> //<,
     . 13, //,
     , 13, //,
                        UNCOUPLED SPECIES = , 13, //,
                        REPLACEMENT SPECIES = , 13, //•
                         INERT OP CONSTANT
                         FLOWING SPECIES =
               REACTION RATE PRINT REQUEST
               TIME INCREMENT = , 1PE12.3,
               ENDING TIME = , 1PE12.3, 9H
               STARTING STEP SIZE =
               MINIMUM STEP SIZE =
SPECIES =
,  13, //,
                                                       13, //
                                          FLAG = . 13, //,
                                          9H MINUTES  , //,
                                          MINUTES , //,
                                     1PE12.3, 9H MINUTES
                                    1PE12.3, 9H MINUTES
                                            ,  //•
                                             //,
                                             //,
              MAXIMUM STEP SIZE = ,  1PE12.3* 9H MINUTES
              CONVERGENCE TOLERANCE  = , 1PE12.3, //,
              DILUTION RATE = , 1PE12.3, 12H MINUTES(-l), //,
    &      1H1,  26X,  18H LIST OF RE ACTION'S, //•
    81      14H      R.  CONST., 8X, 9HREACTANTS, 12X, 8HPRODUCTS,
1003 FORMAT  (14H  SPECIES NAME , A4,  21H NOT IN SPECIES LIST,
    8.      21H  JOB WILL BE ABORTED.)
1004 FORMAT  (13,  1PF11.3, 4(1X, A4), 2H =,
    8,      3(OPF6.2,  IX, A4M
1005
1006
1007
1008
1009
1010
1011
    81
     FORMAT d
     FORMAT (/
     FORMAT (/
     FORMAT (/
     FORMAT (/
     FORMAT (/
     FORMAT (/
          24H
              ',  4(18H SPECIES   VALUE  ))
               ,  18H DIFFERENTIAL(PPM) , //. (4(3X, A4,  1PE11.3)))
               ,  15H UNCOUPLED(PPM) , //, (4(3X, A4, 1PE11.3)))
               ,  20H INERT/CONSTANT (PPM) ,  //,  (4(3X, A4,  1PE11.3)))
               /.  20X? 8H TIME = ,  1PE12.3, 8H MINUTES, /)
               ,  33H THIS RUN TERMINATED WITH KFLAG =  , 13)
               /,  33H PROGRAM CANNOT HANDLE MORE THAN  , 14,
               SPECIES -- JOB ABORTED.)
                                                       14
1012 FORMAT  (//,  33H PROGRAM CANNOT HANDLE MORE THAN
    8.     30H  DIFFERENTIALS — JOB ABORTED.)
1013 FORMAT  (/,  1H1, 20X, 30HINITIAL SPECIES CONCENTRATIONS
1019 FORMAT  (/,  31H MAXIMUM NUMBER OF PLOT POINTS  , 13,
    R-     19H  HAS BEEN EXCEEDED., /. 15H POINT AT  TIME  ,  F8.2,
    &     21H  WILL NOT BE PLOTTED.)
1023 FORMAT  (//,  33H PROGRAM CANNOT HANDLE MORE THAN  ,  14,
    8,     22H  FLOWS — JOB ABORTED.)
1024 FORMAT  (//,  33H PROGRAM CANNOT HANDLE MORE THAN  ,  14,
                                                             /)
 00006150
 00006160
 00006170
 00006lflO
 00006190
 00006200
 00006210
 00006220
 00006230
 00006240
 00006250
 00006260
 00006270
 00006230
 00006290
 00006300
 00006310
 00006320
 00006330
 00006340
 00006350
 00006360
 00006370
 00006380
 00006390
 00006400
 00006410
 00006420
00006430
 00006440
00006450
00006460
00006470
000064RO
00006490
00006500
00006510
00006520
00006530
00006540
00006550
00006560
00006570
00006580
00006590
00006600
00006610
00006620
00006630
00006640
00006650
00006660
00006670
 00006680
00006690
00006700

-------
               EXHIBIT A-l.  LISTING OF MAIN PROGRAM MODKIN  (Concluded)
                                                                       249
    J.      ?7H FLOW TIMES — JOB ABORTED.)                               0000671.0
1025 FORMAT  {/,  IflH STEADY STATE(PPM), //,  (4(3X,  A4?  1PE11.3)))       00006720
1026 FORMAT  (/,  10H INCOMING i A4» 26H CONCENTRATION  CHANGED TO ,      00006730
    &      1PE11.3, 4H AT « 1PE11.3, 5H MIN.)                            00006740
1027 FORMAT  (//,  33H PROGRAM CANNOT HANDLE  MORE  THAN  t  I4»              00006750
    &      33H LUMPED REACTIONS -- JOB ABORTED.)                         00006760
1028 FORMAT  (16H  LUMPED SPECIES , A4- 24H IS NOT FIRST  SPECIES IN,     00006770
    f.      54H CORRFSPONDING LUMPED REACTION — JOB  WILL BE  ABORTED.)    00006780
1029 FORMAT  (/.  21H THE FOLLOWING SET OF ,  I3»                          00006790
    f,      4?H REACTIONS CORRESPONDS TO REACTION  NUMBER  ,  13, /)         00006800
1030 FORMAT  (//»  33H PROGRAM CANNOT HANDLE  MORE  THAN  9  14*              00006810
    8,      39H CONTRIBUTING REACTIONS — JOB ABORTED.)                   00006«?0
1031 FORMAT  (/,  17H REPLACEMENT(PPM)» //. (4(3X» A4»  1PE11.3)))         00006830
1032 FORMAT  (/.  15X, 44HREACTION  RATES (SORTED INTO DECREASING SIZE)*  00006640
    &,      //» 5(15H   NO.    RATE   )» /)                                 00006850
1033 FORMAT  (5(15, 1PE10.2))                                             00006860
    END                                                                 00006870

-------
                      EXHIBIT A-2.  LISTING OF SUBROUTINE LMPCAL
                                                                        250
C SUBROUTINE ****** LMPCAL «****«                                   00000010
C                                                                        00000020
C THIS SUBROUTINE CALCULATES THE CONCENTRATIONS  OF  LUMPED  SPECIES AND   00000030
C THE COEFFICIENTS AND RATE CONSTANTS FOR THE  CORRESPONDING  REACTIONS   00000040
C                                                                        00000050
C SYMBOL DEFINITIONS —        •                                          00000060
C                                                                        00000070
C ALPHA    LOCAL VALUES OF COEFF, THE PRODUCT  COEFFICIENTS               000000*0
C COEFF    NUMBER OF PARTICLES* ONE PER PRODUCT  SPECIES  PER  REACTION     00000090
C COLOC    LOCAL VALUE OF PRODUCT COEFFCIENT                             00000100
c COMIN    MINIMUM ALLOWABLE COEFFICIENT SIZE                            00000110
C J        DO-LOOP INDICES OR LOCAL POINTERS                             000001PO
C K        DO-LOOP INDICES OR LOCAL POINTERS                             00000130
C KCOF     COEFFICIENT POINTERS, ONE PER REACTION PRODUCT PER  SPECIES   00000140
C KLMP     NUMBER OF CONTRIBUTING REACTIONS TO EACH LUMPED REACTION      00000150
C KPRD     PRODUCT POINTERS? ONE PER PRODUCT SPECIES PER REACTION       00000160
C KRCT     REACTANT POINTERS, ONE PER REACTANT SPECIES PER REACTION      00000170
C KRXN     REACTION POINTERS? ONE PER REACTION PER  SPECIES               00000130
C L        DO-LOOP INDICES OR LOCAL POINTERS                             00000190
C LRXN     POINTER TO LUMPED REACTION                                    00000200
C M        DO-LOOP INDICES OP LOCAL POINTERS                             .00000210
C MAXPRD   MAXIMUM NUMBER OF PRODUCTS                                    00000220
C N        DO-LOOP INDICES OR LOCAL POINTERS                             00000230
C NDIF     NUMBER OF DIFFERENTIAL SPECIES                                00000240
C NLMP     NUMBER OF LUMPED REACTIONS                                    00000250
C NLOC     NUMBER OF REPLACEMENT REACTIONS FOR THIS LUMPED REACTION      000002^0
C NOUT     THE FORTRAN OUTPUT UNIT NUMBER (NORMALLY 6)                   00000270
C NR       POINTER TO REPLACEMENT REACTION                               000002RO
C NRXN     NUM-B-ER OF REACTIONS                             %              00000290
c NS       POINTER TO REACTANT SPECIES                     *              00000300
C NSTS     NUMBER OF STEADY-STATE SPECIES                '                00000310
C Q        DEGRADATION PATE? /MIN                                        00000320
C R        REACTION RATES? SEC, ONE PER REACTION                         00000330
c RK       REACTION RATE CONSTANTS? PPM-MIN, ONE PER REACTION            00000340
C RKLMP    LOCAL VALUE OF LUMPED RATE CONSTANT                           00000350
C RKLOC    LOCAL VALUES OF RK? THE REACTION RATE CONSTANTS               00000360
C SUM      SUM OF CONCENTRATIONS OF ALL THE REPLACEMENT SPECIES          00000370
C TOL      CONVERGENCE TOLERANCE ON STEADY-STATE ITERATION? PPM          000003RQ
C VAX      SPECIES CONCENTRATIONS* PPM, ONE PER SPECIES                  00000390
C YF       THE MOLE FRACTIONS OF THE REPLACEMENT SPECIES                 00000400
C YIN      SPECIES INFLOW RATES. PPM/MIN? ONE PER SPECIES                00000410
'C YLOC'     LOCAL VALUES OF YAX, THE SPECIES CONCENTRATIONS               00000420
C                                                                        00000430
C SUBROUTINE  ENTRY POINT                                                 00000440
C                                                                        00000450
      SUBROUTINE LMPCAL(NLMP, KLMP)                                      00000460
C                                                                        00000470
C DECLARE  COMMON STORAGE                              "                   00000480
C                                                                        00000490
      COMMON  RK(99),  R(99), YAX(50), YIN(50), COEFF(3,99)                00000500
      COMMON  KRCT(4,99)t KPRD(3,99), KRXN(99,50), KCOF(99»50)            00000510
      COMMON  0,  TOL»  NRXN, NDIF, NSTS                                    00000520
C                                                                        00000530
C SET DIMENSIONS                                         .                00000540

-------
              EXHIBIT A-2.  LISTING OF SUBROUTINE LMPCAL (Continued)
                                                                     251
    DIMENSION YLOC(10)» PKLOC(IO). ALPHA(3*10),  YF(10),  KLMP

SET DATA STATEMENT PARAMETERS

    DATA MAXPRD /3/» COMIN /o,oooi/

SET CONTRIBUTING REACTION POINTER AND NUMBER  OF  CONTRIBUTING

    NR = NRXN + 1
    DO 70 N = 1»NLMP
    NLOC = KLMP(N)
    LRXN = NRXN - NLMP + N

SAVE RATE CONSTANT AND SAVE AND SUM SPECIES CONCENTRATIONS

    SUM =0.0
    DO 20 K = 1»NLOC
    NS = KPCT(l.NR)
    RKLOC(K)  = RK(NR)
    YLOC(K) = YAX(NS)
    SUM = SUM + YLOC(K)

SAVE PRODUCT  COEFFICIENTS

    DO 10 J = l^MAXPRD
    ALPHA(J,K)  = COEFF(J,NR)
 10 CONTINUE

ADVANCE REACTION POINTER AND SAVE OVERALL SUM

    NR = NR * 1
 20 CONTINUE
    NS = KRCT(1«LRXN)
    YAX(NS) = SUM

CALCULATE THE MOLE FRACTIONS

    DO 30 K = 1?NLOC
    YF(K) = YLOC(K)  / SUM
 30 CONTINUE

CALCULATE LUMPED RATE CONSTANT

    RKLMP =0.0
    DO 40 K = 1
-------
                EXHIBIT A-2.  LISTING OF SUBROUTINE LMPCAL (Concluded)
                                                                         252
   50  CONTINUE
C
C NORMALIZE COEFFICIENT AND CHECK  FOR  UNDERFLOW
C
      COLOC = COLOC / RKLMP
      IF (COLOC .LT. COMIN) COLOC  =  0.0
      COEFF(J,LRXN) = COLOC
   60  CONTINUE
   70  CONTINUE
C
C END  OF PROGRAM — RETURN TO CALLER
C
      RETURN
      END
00001110
00001120
00001130
00001140
00001150
00001IftO
00001170
000011RO
00001190
00001200
00001210
00001220
00001230
00001240

-------
                     EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB
                                                                     253
C» THE  PARAMETERS  TO  THE  SUBROUTINE  DIFSUB  HAVE                          00000020
C* THE  FOLLOWING MEANINGS:                                               00000030
C*                                                                      00000040
         THE  NUMBER OF  FIRST  ORDER DIFFERENTIAL  EQUATIONS.   N           00000050
           MAY  BE  DECREASED ON  LATER CALLS  IF  THE  NUMBER  OF              00000060
           ACTIVE  EQUATIONS REDUCES, BUT  IT MUST NOT  BE                  00000070
           INCREASED  WITHOUT  CALLING WITH JSTART = o                     oooooo«o
         THE  INDEPENDENT  VARIABLE.                                       00000090
         AN 8 BY N ARRAY  CONTAINING  THE DEPENDENT  VARIABLES  AND          00000100
           THEIR SCALED DERIVATIVES.  Y(J+I,D CONTAINS                  00000110
           THE  J-TH DERIVATIVE  OF Y
-------
                EXHIBIT A-3.   LISTING OF SUBROUTINE DIFSUB (Continued)
                                                                       254
C*                       ARRAY WHERE M  IS  THE  VALUE  OF  N USED ON
C*                       THE FIRST CflLL TO THIS  PROGRAM.
C*                2    THE SAME AS CASF 1, EXCEPT  THAT  THIS
C*                       SUBROUTINE COMPUTES THE PARTIAL
C*                       DEVIVATIVES BY NUMFRICAL  DIFFERENCING
C*                       OF THE DEVIVATIVES.   HENCE  PEDERV  IS
C*                       NOT CALLED.
C»  YMAX AN ARRAY OF N LOCATIONS WHICH  CONTAINS  THE  MAXIMUM
C*         OF EACH Y SEEN SO FAR.  IT SHOULD NORMALLY BE SET  TO
C*         1 IN EACH COMPONENT BEFORE THE  FIRST  ENTRY.   (SEE  THE
C*         DESCRIPTION OF EPS.)
C*  ERROR  AN ARRAY OF N ELEMENTS WHICH CONTAINS THE ESTIMATED
C*         ONE STEP ERROR IN EACH COMPONENT.
C*  KFLAG  A COMPLETION CODE WITH THE FOLLOWING  MEANINGS:
c*                +1   THE STEP WAS SUCCESSFUL.
C*                -1   THE STEP WAS TAKEN  WITH H = HMIN,  BUT  THE
C*                       REQUESTED ERROR WAS NOT ACHIEVED.
C*                -2   THE MAXIMUM ORDER SPECIFIED WAS  FOUND  TO
C*                       BE TOO LARGE.
C*                -3   CORRECTOR CONVERGENCE COULD NOT  BE
C*                       ACHIEVED FOR H .GT. HMIN.
C*                -4   THE REQUESTED ERROR IS  SMALLER THAN  CAN
C*                       BE HANDLED FOR THIS PROBLEM.
C*  JSTART   AN INPUT INDICATOR WITH THE FOLLOWING MEANINGS:
C«                -1   REPEAT THE LAST  STEP WITH A NEW  H
C*                 0   PERFORM THE FIRST STEP. 'THE FIRST STEP
C*                       MUST BE DONE WITH THIS  VALUE OF  JSTflRT
C*                       SO THAT THE SUBROUTINE  CAN INITIALIZE
C*                       ITSELF.
c*                +1   TAKE A NEW STEP CONTINUING FROM  THE  LAST.
C*           JSTART IS SET TO NQ, THE CURRENT  ORDER OF  THE  METHOD
C*           DERIVATIVE USED, THIS RESTRICTS THE ORDER.   IT MUST
C*           BE LESS THAN 8 FOR ADAMS AND 7 FOP STIFF METHODS.
C*  PSAVF.  A BLOCK OF AT LEAST N**2 FLOATING POINT LOCATIONS.
C*
C*
C*
             DERIVATIVE USED, THIS RESTRICTS THE ORDER.  IT MUST
             BE  LESS THAN 8 FOR ADAMS AND 7 FOR STIFF METHODS.
    PSAVE   A  BLOCK OF AT LEAST N**2 FLOATING POINT LOCATIONS.
    ***
      DOUBLE  PRECISION A , D,E , H,R, T , Y , Rl , R2, BND,EPS, EUP, EDWN, ENO 1
     1 . *ENQ2, ENQ3 » HM AX » HMIN «HNEW» HOLD » SAVE? TOLD , YMAX , ERROR * RACUM
     2*SDOT1,SDOT2
      DIMENSION  Y(8»40)» YMAX(40), SAVE(12*40)» ERRORUO)* PSAVE(
      DIMENSION  A(H), PERTST ( 7 » ? , 3 ) , SDOT1(40)* SDOT2(40)
        00000550
        00000560
        00000570
        00000580
        00000590
        00000600
        00000610
        00000620
        00000630
        00000640
        00000650
        00000660
        00000670
        00000680
        00000690
        00000700
        00000710
        000007.PO
        00000730
        00000740
        00000750
        00000760
        00000770
       •00000780
        00000790
        00000800
        00000810
        000008 ? 0
        OOOOOH30
        00000840
        00000850
        00000900
        00000910
        000009PO
innnnnntQ 0000930
        00000900
        00000910
        00000920
innnnnni-00000930
KFLAG,  00000940
        00000950
        00000960
        00000970
        00000980
1600)   00000990
        00001000
C*  THE  COEFFICIENTS  IN PERTST APE USED IN SELECTING THE STEP AND
C*  ORDER,  THEREFORE  ONLY ABOUT ONE PERCENT ACCURACY IS NEEDED.
     DATA  PERTST
                 /2.0«4.5.7.333,10.42,13.7,17.15,1.0,
                  2.0,12.0,24.0,37.89,53.33., 70.08,87.97,
                  3.0,6.0,9.167,12.5,15.98,1.0,1.0,
                  12.0,24.0,37.89,53.33,70.08,87.97,1.0,
                  !.,!., 0.5, 0.1667,0.04133,0.008267,1.0,
                  1.0,1.0,2.0,1.0,.3157,.07407,.0139/
       00001020
       00001030
      * 00001040
       00001050
       00001060
       00001070
       00001080
       00001090
       00001100

-------
               EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB (Continued)
                                                                       255
      A(2)=-l.
      IRET = 1
      KFLAG = 1
      IF(JSTART.LF.O)
                      GO TO 140
C*
c»
C*
C*
c*
C*
    BEGIN  BY  SAVING INFORMATION FOR POSSIBLE RESTARTS  AND  CHANGING
    H  BY THE  FACTOR R  IF THE CALLER HAS CHANGED H.   ALL  VARIABLES
    DEPENDENT ON  H  MUST ALSO RE CHANGED.
    E  IS A COMPARISON  FOR ERRORS OF THE CURRENT ORDER  NO.   EUR  IS
                                                                      00001110
                                                                      00001120
                                                                      00001130
                                                                      00001140
                                                                nxnnn>0000 1 150
                                                                      00001160
                                                                      00001170
                                                                      00001130
                                                                      00001190
    TO  TEST  FOR  INCREASING THE ORDER, EDWN. FOR DECREASING  THE  ORDER.     00001200
    HNEW  IS  THE  STEP SIZE THAT WAS USED ON THE LAST CALL.                00001210
 100D0110I=1,N
      DO 110 J = 1 ,K
 110  SAVE(J,I)  = Y(J,I)
      HOLD = HNEW
      IF ( H.EO.HOLD) GO TO 130
 120  RACUM = H/HOLD
      IRET1 = 1
      GO TO 750
 130  NQOLD = NO
      TOLD = T
      RACUM = 1.0
      IF (JSTART.GT.O) GO TO 250
      GO TO 170
 140  IF (JSTART.EQ.-l)  GO TO 160
C*  ON  THE  FIRST  CALL*  THE ORDER
C*  DERIVATIVES  ARE  CALCULATED.
                                 IS SET TO  i  AND  THE  INITIAL
NQ =
N3 =
Nl =
N2 =
N4 =
N5 =
N6 =
CALL
1
N
N*10
Nl + 1
N«.»2
Nl + N
N5 + 1
DIFFUN
(T,Y»SDOT1)
 150
      DO  150  I  =  1,N
      Y(2»I)  =  SDOT1(I)*H
      HNEW  =  H
      K = ?
      GO  TO 100
C*  REPEAT LAST STEP BY RESTORING SAVED  INFORMATION.
 160   IF  (NO.EQcNQOLD
      IF(KFLAG.GE.-l)
      NQ  = NOOLD
      K = NQ  +  1
      GO  TO 120
                     i  JSTART = 1
                      T = T - HOLD
                                                                     00001230
                                                                     000012^0
                                                                     00001250
                                                                     00001260
                                                                     00001270
                                                                     000012*0
                                                                     00001290
                                                                     00001300
                                                                     00001310
                                                                     000013PO
                                                                     00001330
                                                                     '00001340
                                                                     00001350
                                                                     00001360
                                                                         1370
                                                                     000013*0
                                                                     00001390
                                                            00()01400
                                                                     00001410
                                                                     00001420
                                                                     00001430
                                                                     00001440
                                                                     00001450
                                                                     00001460
                                                                     00001470
                                                                     000014^0
                                                                     00001490
                                                                     00001500
                                                                     00001510
                                                                     00001520
                                                                     00001530
                                                      •innnnnnnnnnnnnni-00001540
                                                                     00001550
                                                                         1560
                                                                     00001570
                                                                     000015HO
                                                                     00001590
                                                                     00001600
C*
C#
C*
C*
                                                                     00001610
                     j#tttt#«tt*fttt«#«#«-tt«*#««tt
-------
               EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB (Continued)
                                                                       256
C»  STEP  IF  IT HAS NOT YET PEEN DONE (IRET =  1)  OR  SKIP TO A FINAL
C«  SCALING  BEFORE EXIT IF IT HAS BEEN COMPLETED  (IRET  = 2).
C«HHHHHHHHHHt tf«#^tttt#Ctt1»tt##tt##tta#fc##tt#«H><»««&«tttt«tttt-H-<
 170  IF(MF.EQ.O)  GO TO 180
      IF  (NQ.GT.6)  GO TO 190
      GO  TO  (221,222*223*224,225,226),NQ
 180  IF(NO.GT.7)  GO TO 190
      GO  TO  (211*212,213,214*215,216,217)9NQ
 190  KFLAG  =  -2
      RETURN
Q»ft»«*»*»«*»«»*»*««#*««*«»««»»«««»««»*»«»tt«*»»«»«*««»««»«-»«'»-»-»»-»a'«'{
C#  THE FOLLOWING  COEFFICIENTS SHOULD BE DEFINED  TO  THE MAXIMUM
C»  ACCURACY PERMITTED BY  THE MACHINE*  THEY  ARE  IN  THE ORDER  USED:
C»
c«  -i
C*
C*
c»
C*
C*
c»
C*
c»
C*
C*
c»
C*
   -3/8, -11/1?, -1/3, -1/24
   -251/720, -25/24, -35/72, -5/48 9 -I/ 120
   -95/288, -137/1 20 .-5/8, -17/96, -1/40, -1/720
   -19087/6 048 0, -49/40, -203/270* -49/1 92 » -7/1 44, -7/1 440, -1/5040
   -6/119-6/11, -1/11
   - 12/25 » -7/10 «-l/5 » -1/50
   -120/274, -225/274 , -85/274 ,-15/274? -1/274
   -180/441, -58/63 ,-15/36, -25/252,-3/252.-l/1764
211
212
213
214
215
      A(l)  =  -1
      GO  TO 230
      A(l)  =  -0
      A(3)  =  -0
      GO  TO 230
      Ad)  =  -0
      A(3)  =  -0
       A(4) = -
      GO  TO 230
      A(l)  =  -0
      A<3)  =  -0
      A(4>  =  -0
      A(5)  =  -0
      GO  TO 230
216
      Ad) =
      A(3) =
      A(4) =
      A(5) =
      A(6) =
      GO TO
      Ad) =
      A(3) =
      A(4) =
      A(5) =
      A(6) =
      A(7) =
      GO TO
: -0
' -1
: -0
: -0
• -o
230
• -0
 -1
 -0
 -0
 -0
 -0
230
.0

.500000000
.500000000

.4166666666666667
.750000000
0.166666666666667

.375000000
,9166666666666667
.3333333333333333
.0416666666666667

.34861111111111
.04166666666667
.48611111111111111
.104166666666666667
.00833333333333333

.3298611111111111
.14166666666666667
.625000000
.1770B33333333333333
.02500000000
.001388888888888889
00001*70
000016PO
    01690
00001700
00001710
000017PO
000017^10
000017^0
00001750
000017*0
    1770
00001780
00001790
00001800
00001810
00001820
00001830
00001840
00001850
00001860
00001870
00001HPO
00001890
00001900
00001910
00001920
00001930
00001940
    1950
00001960
00001970
000019«0
00001990
00002000
00002010
00002020
00002030
00002040
00002050
000020*0
00002070
000020*0
00002090
00002100
00002110
00002120
00002130
00002140
00002150
00002160
00002170
00002180
00002190
00002200
00002210
00002220

-------
               EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB  (Continued)
                                                                      257
 217
221

222


223



224




225
 226
 230
 240
A(l)
A(3)
A(4)
A(5)
A(6)
A(7)
A(8)
GO TO
            -0,
            -1,
            -0.
            -0.
            -0,
            -0.
            -0,
           230
           3155919312169312
           235000000
           7518518518518519
           2552083333333333
           0486111111111111
               ,0001984126984126984

               ,000000000
               ,6666666666666667
               ,3333333333333333

               ,5454545454545455
               I)
               , 09090909090909091
A(l! = -1,
GO TO ?30
Ad) = -0,
A(3) = -0,
GO TO 230
Ad) = -0,
A(3) = AC
A(4) = -0.
GO TO 230
Ad) = -0.480000000
A(3) = -0.7000000000
A(4) = -0.2000000000
A(5) = -0.0200000000
GO TO 230
Ad) = -0.437956204379562
       -0.8211678832116788
          3102189781021898
          05474452554744526
          0036496350364963504
      A(3)
      A(4)
      A(5)
      A(6)
      GO TO
      A(3)  =
              ,4081632653061225
              ,9206349206349206
              ,4166666666666667
               0992063492063492
               0119047619047619
               000566893424036282
      = -0
      = -0
      = -0
      230
      =. -0
       -0

A(5) = -0
A(6) = -0
A(7) = -0
K = NQ+1
IDOUB = K
MTYP = (4 -MF)/2
ENQ2 = ,5/FL.OAT(NO +  1)
ENQ3 = .5/FLOAKNQ +2)
ENQ1 = ,5/FLOAT(NQ)
PEPSH = EPS
EUP = (PERTST(NQ»MTYP,2)»PEPSH)**2
F = (PFRTST(NO,MTYP» 1)"PEPSH)**2
EDWN = (PERTST(NQ,MTYP,3)«PEPSH)**2
IF  (EDWN.EO.O) GO TO  780
BND = EPS«ENQ3/FLOAT(N)
IWEVAL = MF
GO TO (250 .680 )»IRET
C*  THIS  SECTION COMPUTES THE PREDICTED VALUES BY EFFECTIVELY
C*  MULTIPLYING THE SAVED INFORMATION BY THE PASCAL TRIANGLE
C*  MATRIX.
 250   T  =  T  +  H
      DO 260 J = 2.K
      DO 260 Jl = J.K
 00002230
 00002240
 00002250
 00002260
 00002270
 00002280
 00002290
 00002300
 00002310
 00002320
 00002330
 00002340
 000023SO
 00002360
 00002370
 00002380
 000023QO
 00002400
 00002410
 00002420
 00002430
 00002440
 00002450
.00002460
 00002470
 000024RO
 000024QO
 00002500
 00002510
 00002520
 00002530
 00002540
 00002550
 00002560
 00002570
 000025^0
 000025QO
 00002600
 00002610
 00002620
 00002630
 00002640
 00002650
 00002660
 00002670
 00002680
 000026QO
 00002700
      '10
 000027?0
 00002730
 00002740
                                                                       00002760
                                                                       00002770
                                                                       00002780

-------
               EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB (Continued)
                                                                       258
      J2  = K-J1  *  J - 1
      00  260  I  = 1,N
 260   Y(J2*I)  =  Y(J2,I)
   n
C*  UP TO 3 CORRECTOR ITERATIONS ARE TAKEN*  CONVERGENCE  IS TESTED
C«  BY REQUIRING CHANGES TO BE LESS THAN BND WHICH IS DEPENDENT  ON
C*  THE ERROR  TEST CONSTANT.
C*     THE SUM  OF  THf CORRECTIONS .IS ACCUMULATED IN THE ARRAY
C*  ERROR(I).   IT  IS EQUAL TO THE I-TH DERIVATIVE OF Y MULTIPLIED
C*  BY H**K/(FACTORIAL(K-1)*A(K))» AND IS THEREFORE PROPORTIONAL
C*  TO THE ACTUAL  ERRORS TO THE LOWEST POWER 'OF H PRESENT.   (H**K)
      DO  270  I  =  1,N
 270  ERROR(T)  =  0.0
      DO  430  L  =  1*3
      CALL  DIFFUN(T?Y?SDOT1)
+ PSAVE(I*(N3+1)-N3>
C»  IF THERE  HAS  BEEN A CHANGE OF ORDER OR THERE HAS BEEN TROUBLE
C*  WITH  CONVERGENCE., PW IS PE-EVALUATED PRIOR TO STARTING THE
C*  CORRECTOR ITERATION IN THE CASE OF STIFF METHODS.  IWEVAL IS
C*  THEN  SET  TO  -1  AS AN INDICATOR THAT IT HAS BEEN DONE.
IF  (IWEVAL.LT.l) GO TO 350
IF  (MF.EQ.2) GO TO 310
CALL PEDERV(T,Y»PSAVE,N3>
R = A(1)*H
DO 280 I = 1,M4
PSAVE(I) = PSAVE(I)*R
DO 300 I = 1»N
PSAVE(I*(N3+1)-N3) = 1.0
IWEVAL = -1
CALL MATINV(PSAVE7N3»N3»J1)
IF(Jl.GT.O) GO TO 350
GO TO MO
DO 320 I = 1,N
SAVE(9,I) = Y(1»I)
DO 340 J = 1,N
R = EPS*DMAX1(EPS-DABS(SAVE(99J)))
Y(1»J) = Y(1»J) + R
D = A(1)*H/R
CALL DIFFUN(T,Y»SDOT2)
00 330 I = 1»N
PSAVE(I+(J-1)*N3) = (SDOT2(I).-SDOT1 (I)
Y(1,J) = SAVE(9»J)
GO TO 290
IF  (MF.NE.O) GO TO 370
DO 360 I = 1»N
SAVE(9,I) = Y(2»I)-SDOT1(I)*H
GO TO 410
DO 380 I = liN
         = Y(2,I)-SDOT1(I)*H
 280
 290
 300
 310
 320
 330
 340

 350

 360

 370
 380
 390
 400
SDOT2(I)
DO 400 I = 1»N
D = 0.0
DO 390 J = l.N
D = D + PSAVE(IMJ-1)*N3>»SDOT2(J)
SAVE(9,I) = D
                                          00002790
                                          00002800
                                          00002810

                                          00002830
                                          00002840
                                          00002B50
                                          00002880
                                          00002870
                                          00002880
                                          00002890
                                        *»00002900
                                          00002910
                                          00002920
                                          00002930
                                          00002940
                                                                   00002^^0
                                                                   00002970
                                                                   00002980
                                                                   00002990
                                                                      103000
                                                                   00003010
                                                                   •00003020
                                                                   00003030
                                                                   00003040
                                                                   00003050
                                                                   000030^0
                                                                   00003070
                                                                   00003080
                                                                   00003090
                                                                   00003100
                                                                   00003110
                                                                   00003120
                                                                   00003130
                                                                   00003140
                                                                   00003150
                                                                   00003160
                                                                   00003170
                                                                   00003180
                                                                   00003190
                                                                   00003200
                                                                   00003210
                                                                   0000322C
                                                                   0000323C
                                                                   00003240
                                                                   0000325C
                                          0000 32 7C
                                          0000328C
                                          0000329C
                                          0000330C
                                          0 0 0 0 3 3 1 C
                                          0000332C
                                          0000333C
                                          0000334C

-------
EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB (Continued)
                                                       259
410  NT = N
     HO 420 I = 1,N
     Yd»I> = Y(l, I) «• Ad)*SAVE(9,I)
     Y(2.I) = Y(2,I) - SAVE(9,I>
     ERRORd)  = ERROR d)  + SAVE(9,I)
     IF (DABS(SAVE(9»I) ) .LE. (BND*YMAX (I) ) )
420  CONTINUE
     IF (NT.LE.O)  GO TO 490
430  CONTINUE
                                            NT  =  NT  -  1
C»  THE CORRECTOR ITERATION FAILED TO CONVERGE  IN  3  TPIESo   VARIOUS
C*  POSSIBILITIES ARE CHECKED FOR.  IF H IS ALREADY  HMIN  AND
C*  THIS IS EITHER ADAMS METHOD OR THE STIFF METHOD  IN  WHICH  THE
C*  MATRIX PW HAS ALREADY BEEN RE-EVALUATED, A  NO  CONVERGENCE  EXIT
C*  IS TAKEN.  OTHERWISE THE MATRIX PW IS RE-EVALUATED  AND/OR  THE
C*  STEP IS REDUCED TO TRY AND GET CONVERGENCE.
( (H.LE. (HMIN*1. 00001 ) ) .AND. ( (IWEVAL - MTYP ) .LT . -1 ) }
( (MF.EQ.O) .OR. (IWEVAL. NE.O) )  RACUM = RACUM**2*0 . 5
440  IF
     IF
     IWEVAL = M
     T = T - H
     IRET1 = 2
     GO TO 750
460  KFLAG = -3
•470  DO 4flO I =
     00 480 J =
480  Y(J*I) :• SAVE(J»I)
     H = HOLD
     NQ = N&OLD
     JSTART = NQ
     RETURN
                                                             GO  TO  460
                 ItN
                 1,K
C*  THE CORRECTOR CONVERGED AND CONTROL IS PASSED TO STATEMENT 520
C*  IF THE ERROR TEST IS O.K., AND TO 540 OTHERWISE.
C*  IF THE STEP IS O.K. IT IS ACCEPTED.  IF IDOUB HAS BEEN REDUCED
C*  TO ONE*  A  TEST IS MADE TO SEE IF THE STEP CAN BE INCREASED
C*  AT THE CURRENT ORDER OR BY GOING TO ONE HIGHER OR ONE LOWER.
C*  SUCH A CHANGE IS ONLY MADE IF THE STEP CAN BE INCREASED BY AT
C*  LEAST 1.1.   IF NO CHANGE IS POSSIBLE IDOUB IS SET TO 10 TO
C*  PREVENT  FUTHER TESTING FOR 10 STEPS
C*  IF A CHANGE IS POSSIBLE* IT IS MADE AND IDOUB IS SET TO
C#  NO * 1    TO PREVENT FURTHER TESTING FOR THAT NUMBER OF STEPS.
C*  IF THE ERROR WAS TOO LARGE* THE OPTIMUM STEP SIZE FOR THIS OR
C*  LOWER ORDER IS COMPUTED* AND THE STEP RETRIED.  IF IT SHOULD
C*  FAIL TWICE  MOPE IT IS AN INDICATION THAT THE DERIVATIVES THAT
C*  HAVE ACCUMULATED IN THE Y ARRAY HAVE ERRORS OF THE WRONG ORDER
C«  SO THE FIRST DERIVATIVES ARE RECOMPUTED AND THE ORDER IS SET
C*  TO 1.
 490   D  =  0.0
      DO 500  I  =  1 »N
 500   D  =  D  f  (ERROR (I)/YMAX(I) )
      IWEVAL  =  0
      IF (D.GT.E)  GO TO 540
      IF (K.LT.3)  GO TO 520
      DO 510  J  =  3,K
                                                         00003350
                                                         00003360
                                                         00003370
                                                         00003380
                                                         00003390
                                                         00003400
                                                         00003410
                                                         00003420
                                                         00003430
                                                         00003450
                                                         00003460
                                                         00003470
                                                         00003430
                                                         0000349U
                                                         00003500
                                                         000035PO
                                                         00003530
                                                         00003540
                                                         00003S50
                                                         00003560
                                                         00003570
                                                         00003580
                                                         000035QO
                                                         00003600
                                                         00003610
                                                         00003620
                                                         00003630
                                                         00003640
                                                         00003650
                                                         00003670
                                                         00003680
                                                         00003690
                                                         00003700
                                                         00003710
                                                         00003720
                                                         00003730
                                                         00003740
                                                         0000375C
                                                         00003760
                                                         00003770
                                                         0000378C
                                                         00003790
                                                         0000380C
                                                         0000381C
                                                         000038PC
                                                         00003B4C
                                                         0000385C
                                                         0000386C
                                                         0000337C
                                                         0000388T
                                                         0000389C
                                                         0000390C

-------
EXHIBIT A-3.  LISTING OF SUBROUTINE  DIFSUB  (Continued)
                                                        260
     DO 510 I =  1,N
510  Y(J,I) = Y(J,I)  +  A(J)*ERROR(I)
520  KFLAG = +1
     HNEW = H
     IF (IDOUP.LE.l)  GO  TO  550
     IDOUR = IDOUB -  1
     IF (IDOUB.GT.l)  GO  TO  700
     DO 530 I =  1,N
530  SAVE(10»I)  = FRROR(I)
     GO TO 700
540  KFLAG = KFLAG -  2
     T = TOLD
     IF (H.LE.(HMIN«1.00001)) GO  TO  740
     IF (KFLAG.LE.-5) GO  TO 720
550  PR2 = (D/E) *-*ENQ2*l c2
     PR3 = l.E+20
     IF ((NQ.GE.MAXDER).OR.(KFLAG.LE.-1))  GO TO 570
     D = 0.0
     DO 560 I =  1,N
560  D = D + UF.RROP(I)  - SAVE (1 0 » I) ) /YMAX (I) ) **2
     PR3 = (D/EUP)»*ENQ3<*1.4
570  PR1 = l.E+20
     IF (NQ.LE.l) GO  TO  590
     D = 0.0
     DO 5BO I =  1»M
580  D = D + (Y(K*I)/YMAX(I))»«2
     PR1 = (D/EDWN)**ENQ1*1.3
590  CONTINUE
     IF (PR2oLE.PR3)  GO  TO  650
     IF (PR3.LT.PR1)  GO  TO  660
600  R = 1 „ 0/AMAX1 (PR1;1.E-4)
     HfrWQ = NO - 1
610  IDOUB = 10
     IF ((KFLAG.EQ.l).AND.(R.LT.(1.1)))  GO TO  700
     IF (NEWQ.LE.NO)  GO  TO  630
     DO 620 I =  1»N
620  Y(NEWQ+1,I) = ERROR(I)*A(K)/FLOAT(K)
630  K = NEWQ +  1
     IF ( KFLAG.EG. 1 )  GO  TO 670
     RACUM = RACUM*R
     IRET1 = 3
     GO TO 750
640  IDOUB = K
     IF (NEWQ.EQ.NQ)  GO  TO  250
     NQ = NFWQ
     GO TO 170
650  IF (PR2.GT.PR1)  GO  TO  600
     NEWQ = NO
     R = 1.0/AMAX1(PR2.1.E-4)
     GO TO 610
660  R = 1.0/AMAX1(PR3.1.E-4)
     NEWQ = NO + 1
     GO TO 610
670  IRET = 2
     R = DMIN1(R»HMAX/DA8S(H))
     H=H*R
                                                          0000391.0
                                                          000039PO
                                                          0000^930
                                                          00003940
                                                          000039SO
                                                          00003960
                                                          00003970
                                                          00003980
                                                          00003990
                                                          0000^000
                                                          00004010
                                                          00004020
                                                          00004030
                                                          00004040
                                                          000040?0
                                                          00004060
                                                          00004070
                                                          000040HO
                                                          00004090
                                                          00004100
                                                          00004110
                                                          00004120
                                                          000041 TO
                                                          00004140
                                                          00004150
                                                          00004160
                                                          00004170
                                                          0 0 0 0 4 1 B 0
                                                          00004190
                                                          00004200
                                                          00004210
                                                          00004220
                                                          00004230
                                                          00004240
                                                          0000425H
                                                          00004260
                                                          00004270
                                                          000042*0
                                                          000042^0
                                                          0000430C
                                                          00004310
                                                          0000432C
                                                          0000433C
                                                          00004341
                                                          000043SC
                                                          0000436r
                                                          0000437T
                                                          000043«f
                                                          0000439f
                                                          0000440C
                                                          000044K
                                                          0000442C
                                                          0 0 0 0 4 4 3 C
                                                          0000444C
                                                          0000445f
                                                          0000446C

-------
              •EXHIBIT A-3.  LISTING OF SUBROUTINE DIFSUB  (Concluded)
                                                                       261
     HNEW = H
     IF (NQ.EO,
               NEWQ) GO TO  630
6BO
690

700
710
730
740
     NO
     GO
     Rl
     DO
     Rl
     DO
         = NFWQ
         TO 170
         = 1.0
         690 J =
         = R1»R
         690 I =
                2»K
                1,N
                1,N
      IDOUB = K
      DO 710 I =
      YMAXCI) = DMAX1 (YMAXd) *DABS(Yd»I) ) )
      JSTART = NO
      RETURN
 720  IF (NO.EQ.l) GO TO 780
      CALL DIFFUN(T,Y?SDOT1)
      R = H/HOLD
      DO 730 I = 1,N
      Yd,I) = SAVE(1,I)
      SAVE(2.I) = HOLD*SDOtld>
      Y(2»I) = SA\
      NQ = 1
      KFLAG = 1
      GO TO 170
      KFLAG = -1
      HNEW = H
      JSTART = NQ
      RETURN
     *«•*««••» •& t
C*  THIS SECTION SCALES ALL VARIABLES CONNECTED  WITH H AND RETURNS
C*  TO THE ENTERING SECTION.
   H
 750  RACUM = DMAX1(0ABS(HMIN/HOLD).RACUM)
      RACUM" = DM INI (RACUM,DABS(UMAX/HOLD))
      Rl = 1.0
      DO 760 J = 2«K
      Rl = R1*RACUM
      DO 760 I = 1»N
 760
770
780
     H = HOLD*RACUM
     DO 770 I = 1»N
     Yd. I)  = SAVEd.I)
    'GO TO d30«250,640) 9IRET1
     KFLAG = -4
     GO TO 470
     END
00004470
00004480
00004490
00004500
00004510
000045?0
00004530
00004540
00004550
00004560
00004570
000045PO
00004590
00004600
00004610
000046PO
00004630
00004640
00004650
00004660
00004670
00004680
00004690
00004700
00004710
000047?0
00004730
00004740

00004760
00004770

00004790
00004800
00004810
000048?0
00004R30
00004840
00004850
00004360
00004870
00004880
00004890
00004900
00004910
000049PO

-------
                                                  262
 EXHIBIT A-4.  LISTING OF SUBROUTINE DIFFUN
C SUBROUTINE
DIFFUN
C THIS SUBROUTINE CALCULATES THE RATE OF CHANGE OF DIFFERENTIAL AND
C STEADY-STATE SPECIES CONCENTRATIONS -- CALLED BY DIFSUB.
C                        '
C
C SYMBOL DESCRIPTIONS —
C
C COEFF    NUMBER OF PARTICLES? ONE PER PRODUCT SPECIES PER REACTION
C J        DO-LOOP INDICES OR LOCAL POINTERS
C JFLAG    INDICATES SPECIES HAS BEEN SEPARATED FROM SDEN CALCULATION
C K        DO-LOOP INDICES OP LOCAL POINTERS
C KCOF     COEFFICIENT POINTERS, ONE PER REACTION PRODUCT PER SPECIES
C KPRD     PRODUCT POINTERS, ONE PER PRODUCT SPECIES PER REACTION
C KRCT     REACTANT POINTERS* ONE PER PEACTANT SPECIES PER REACTION
C KRXN     REACTION POINTERS? ONE PER REACTION PER SPECIES
C L        DO-LOOP INDICES OR LOCAL POINTERS
C M        DO-LOOP INDICES OR LOCAL POINTERS
C MAXPRD   MAXIMUM NUMBER OF PRODUCTS
c MAXPCT   MAXIMUM NUMBER OF REACTANTS
C N        DO-LOOP INDICES OR LOCAL POINTERS
C MCNV     NUMBER OF CONVERGED STEADY-STATES
C NDIF     NUMBER OF DIFFERENTIAL SPECIES
C NOUT     THE FORTRAN OUTPUT UNIT NUMBER (NORMALLY 6)
C NRXN     NUMBER OF REACTIONS
C NS       LOCAL POINTER TO STEADY-STATE SPECIES
C NSTS     NUMBER OF STEADY-STATE SPECIES
C NTRY     NUMBER OF ITERATION ATTEMPTS FOR STEADY-STATE CONVERGENCE
C 0        DEGRADATION RATE, /MIN
C R        REACTION RATES. SEC? ONE PER REACTION
C RATE     LOCAL REPRESENTATION OF R, THE REACTION RATE
C RK       REACTION RATE CONSTANTS* PPM-MIN* ONE PER REACTION
C SDEN     DENOMINATOR IN STEADY-STATE CALCULATION? /MIN
C SNUM     NUMERATOR IN STEADY-STATE CALCULATION, PPM/MIN
C STEST    TEST VALUE FOR STEADY-STATE CONVERGENCE-. PPM
C T        CURRENT REACTION TIME; SEC? DOUBLE PRECISION?
C              FOR AND FROM DIFSUB
C TOL      CONVERGENCE TOLERANCE ON STEADY-STATE ITERATION? PPM
C Y        SPECIES CONCENTRATIONS? 8 PER SPECIES? DOUBLE PRECISION?
C              FOR AND FROM DIFSUB
C YAX      SPECIES CONCENTRATIONS. PPM, ONE PER SPECIES
C YCALC    LOCAL REPRESENTATION OF YDOT? THE RATE OF CHANGE
c YDOT     RATES OF CHANGE OF SPECIES CONCENTRATION,  PPM/MIN. ONE PER
C              DIFFERENTIAL SPECIES? DOUBLE PRECISION? FOR DIFSUB
C YIN      SPECIES INFLOW RATES? PPM/MIN? ONE PER "SPECIES
C
C BEGINNING OF PROGRAM.
C
C ENTRY POINT
C
      SUBROUTINE DIFFUNCT? Y, YDOT)
C
C DECLARE INPUTS FROM DIFSUR TO BE DOUBLE PRECISION WITH DIMENSIONS
C
                                                     0000001
                                                     000000?
                                                     0000003
                                                     0000004
                                                     oonooo<=.
                                                     0000006
                                                     0000007
                                                     OOOOOOR
                                                     0 0 0 0 0 0 Q
                                                     0000010
                                                     0000011
                                                     000001?
                                                     0000013
                                                     0000014
                                                     000001?
                                                     0000016
                                                     000001T
                                                     000001H
                                                     000001^)
                                                     OOOOOPO
                                                     00000?!
                                                     00000??
                                                     0000023
                                                     00000?4
                                                     00000?S
                                                     00000?6
                                                     00000?T
                                                     000002A
                                                     00000?.^
                                                     00000.30
                                                     0000031
                                                     000003?
                                                     0000033
                                                     0000034
                                                     0000035
                                                     0000036
                                                     0000037
                                                     000003^
                                                     0000039
                                                     0000040
                                                     0000041
                                                     000004?
                                                     0000043
                                                     0000044
                                                     0000045
                                                     000004A
                                                     0000047
                                                     000004B
                                                     0000049
                                                     0000050
                                                     0000051
                                                     000005?
                                                     0000053
                                                     0000054

-------
               EXHIBIT A-4.  LISTING OF SUBROUTINE DIFFUN (Continued)
                                                                       263
      DOUBLE  PRECISION T, Y» YDOT
      DIMENSION Y(8*40)» YDOT(40)
C
C DEFINE VARIABLES AND DIMENSIONS OF COMMON STORAGE  WITH  MODKIN
C
      COMMON  RK(99), R(99), YAX(50), YIN(50)? COEFF(3,9<5)
      COMMON  KRCT(4,99)« KPRD(3,99), KRXN(99t50)»  KCOF(99*50)
      COMMON  Q, TOL, NRXN« NDIF, 'NSTS
C
C DEFINE MISCELLANEOUS DATA VALUES
C
      DATA NTRY /25/, MAXRCT /4/9 MAXPRD /3/, NOUT /6/» NWARN  /O/
C
C MOVE DIFFERENTIAL CONCENTRATIONS TO LOCAL ARRAY
C
      DO 110  J = 1,NDIF
      YAX(J)  = Y(1?J)
110   CONTINUE
C
C SET ITERATION LOOP AND CALCULATE REACTION RATES
C
      DO 260  N = 1?NTRY
      DO 140  L = 1?NRXN
      RATE =  RK(L>
      DO 120  K = 1,MAXPCT
      J = KRCT(K,L)
      IF (J 0EQo 0) GO TO 130
      RATE =  RATE * YAX(J)
120   CONTINUE
130   R(L)  =  RATE
140   CONTINUE
C
C SET CONVERGENCE COUNTER AND BEGIN STEADY-STATE CALCULATION  LOOP
C
      NCNV =  0
      IF (NSTS .LE. 0) GO TO 255
      DO 250  M = 1»NSTS
      SDEN =  0
      SNUM =0.0
      NS = NDIF + M
      STEST = YAX(NS)
C
C IDENTIFY STEADY STATE SPECIES  IN REACTION
C
      DO 230  L = 1»NRXN
      J = KCOF(L»NS)
      K = KRXN(L?NS)
C
C SKIP OVER LUMPED MECHANISM REPLACEMENT SPECIES
C
      IF (K .GT. NRXN) GO TO 230
      IF (J)  205* 235* 203
C
C CALCULATE NUMERATOR OF STEADY  STATE EQUATION
C
203   SNUM =  SNUM * R(K) « COEFF(J,K)
0000055
000005^
0000057
000005H
0000059
OOOOOiSO
0000061
000006?
000006?
0000064
0000065
0000066
0000067
000006*
000006^
0000070-
0000071
000007?
0000073
OOOOOT&
0000075
0000076
0000077
0 0 0 0 0 7.«
000007Q
0 0 0 0 0 K 0
0000081
OOOOOR?
OOOOOP3
OOOOOB^-
OOOOOB5
0000086
0000087
OOOOOBH
0 0 0 0 0 fi 9
0000090
0000091
000009?
0000093
0000094
0000095
OOOOOQ6
0000097
0000091*
0000099
0000100
0000101
0000102
0000103
0000104
0000105
0000106
0000107
000010*
0000109
0000110

-------
                FXHIBIT A-4.  LISTING OF SUBROUTINE DIFFUN (Continued)
                                                                       264
      GO  TO  230                                                           0000111
 c                                                                         o o o o 11;
 C  START REACTION  PATE  CALCULATION AND SET SPECIES FLAG                   000011;
 C                                                                         0000114
 205   RATE = RK(K)                                                        0000111-
      JFLAG  = 0                                                           OOOOllf-
      DO  210 NR = 1,MAXRCT                                                0000117
      J = KRCT(NR,K)                                                      000011-'
      IF  (J  .EG.  0)  GO TO  220                                             000011C
      IF  (J  .NE.  WS) GO TO  208                                           000012C
 C                                                                         0000121
 C  CALCULATE  RATE* SKIPPING  FIRST OCCURRENCE OF SPECIES IN REACTION       000012?
 c                                                                         000012:
      IF  (JFLAG .EQ, 1)  GO  TO  208                                        0000124
      JFLAG  = 1                                                           00001?c
      GO  TO  210                                                           OOOOlSf:
 208   RATE = RATE *  YAX(J)                                                0000127
 210   CONTINUE                                                           000012*
 C                                                                         000012S
 C  CALCULATE  DENOMINATOR OF  STEADY STATE EQUATION                         0000130
 C                                                                         0000131
 220   SDEN = SDEN +  RATE                                                 0000132
 230   CONTINUE                                                           0000133
•C                                                                         0000134
 C  TEST VALUES FOR ZERO --  SKIP CONVERGENCE TEST IF SO                    000013^
 C                                                                         000013*=
 235   IF  (SDEN .LE.  0.0) GO TO 240                                       0000137
      IF  (SNUM .LE.  0.0) GO TO 240                                       000013^
 C                                                                         000013^
 C  CALCULATE  STEADY-STATE CONCENTRATION AND CHECK FOP CONVERGENCE         0000140
 C                                                                         0000141
      STEST  = SNUM  / SDEN                                                 0000142
      IF  (ABS((STEST - YAX(NS))  / STEST)  .GT.  TOL) GO TO 245             0000143
 C                                                                         0000144
 C  UPDATE  CONVERGENCE COUNTER AND SPECIES  CONCENTRATION                   0000145
 C                                                                         0000146
 240   NCNV = NCNV «•  1                                                     0000147
 245   YAX(WS)  = STEST                                                     000014P
 250   CONTINUE                                                           000014Q
 C                                                                         0000150
 C  TEST FOR CONVERGENCE OF  ALL  STEADY-STATES — WRITE MESSAGE IF FAILED   0000151
 C                                                                         0000152
 255   IF  (NCNV «,EQ*  NSTS)  GO TO  300                                       0000153
 260   CONTINUE                                                           0000154
      WRITE  (NOUT»1031>  NTRY                                             0000155
 1031  FORMAT (« STEADY STATE FAILED  TO CONVERGE IN '» I3»                0000156
     S.     •  ITERATIONS.')                                                0000157
 C                                                                         000015H
 C  INCREMENT  WARNING  COUNTER AND  STOP IF TOO MANY                         0000159
 C                                                                         0000160
      NWARN  = NWARN  +  1                                                   0000161
      IF  (NWARN .GT. NTRY)  STOP                                           000016?
 C                                                                         0000163
 C  CALCULATE  RATE  OF  CHANGE  OF  CONCENTRATION FOR DIFFERENTIAL SPECIES     0000164
 C                                                                         0000165
 300   DO  330  M =  1»NDIF                                  '                 0000166

-------
EXHIBIT A-4.  LISTING OF SUBROUTINE DIFFUN (Concluded)
                                                         265
      YCALC = 0.0
      DO 310 L = 1«NPXN
      J = KCOF(L«M)
      K = KRXN(L»M)
C
C SKIP OVER LUMPED MECHANISM  REPLACEMENT SPECIES
C
      IF (K .GT. NRXN) GO  TO  310
      IF (J) 305» 320* 307
305   YCALC = YCALC - R(K)
      GO TO 310
      YCALC = YCALC * R(K)
      CONTINUE
      YDOT(M) = YCALC *  Q
      CONTINUE
                                                           0000167
                                                           OOOni6-J
307
310
3?0
330
c
C END OF ROUTINE — RETURN  TO  CALLER
C
      RETURN
      END
 COEFF(J»K)

(YlN(M)  - YAX(M))
                                                           0000170
                                                           0000171
                                                           000017^
                                                           000017"
                                                           0000174
                                                           0000175
                                                           000017^
                                                           0000177
                                                           000017f-
                                                           000017s
                                                           00001«0
                                                           00001^1
                                                           OOOOIR;
                                                           00001H2
                                                           0000184
                                                           000018C
                                                           OOOOlflf

-------
                   EXHIBIT A-5.  LISTING OF SUBROUTINE MATINV
                                                                       266
67
  10
  15
  20
  30
  40
  45
  50
  60
  70
  80
  85
  90
  95
 100
 105

 110
 130
 HO
 150
 160
 170
 200
 260
 ?70
 310
 320
 330
 340
 350
 380
 390
 400
 420
 430
 450
 550
 600
 610
 620
 630
 640
 650
 660
 670
SUBROUTINE MATINV(PSAVE,N->MM,J1)
DIMENSION A(40,40),  INDEX(40»?)»
DIMENSION PSAVF(1600)
EQUIVALENCE  (IROW,JROW),  (ICOLUM,JCOLUM)
KK = 0
DO 67 I = 1,N
DO 67 J = 1,N
KK = KK «• 1
A(J,I) = PSAVE(KK)
OETERM=1.0
DO 20 J=1,M
IPIVOT(J)=0
DO 550 1 = 1,N
AMAX=0.0
DO 105 J=1,N
IF (IPIVOT(J)-l)
   100 K=1,N
   (IPIVOT(K)-l)  80,
                                      PIVOT(40),  IPIVOT(40)
                                                (AMAX,  T,  SWAP)
                     105? 60
                     100, 740
        (AMAX)-ABS  (A(J9KM) 85*  100,  100
                  60»
DO
IF
IF  (ABS
IROW=J
ICOLUM=K
AMAX=A(J,K)
CONTINUE
CONTINUE
IFCAMAX  .EQ.  0.)  GO  TO 760
IPIVOT(ICOLUM)=IPIVOT(ICOLUM)+1
IF  (IROW-ICOLUM)  140, 260, 140
DETERM=-DETERM
DO  200  L=1,N
SWAP=A(IROW.L)
A (IROW,L)=A(ICOLUM,L)
A(ICOLUM,L)=SWAP
INDEX(I91)=IROW
INDEX(I,2)=ICOLUM
P I VOT(I)=A(ICOLUM,ICOLUM)
DETERM=DETERM*PIVOT(I)
A (ICOLUM, ICOLUM )=KO
DO  350  L=  1,N
A(ICOLUM,L)=A(ICOLUM,L)/PIVOT(I)
DO  550  L1=1,N
'IF(LI-ICOLUM)  400*  550»  400
T=A(L1,ICOLUM)
A(L1»ICOLUM)=0.0
DO  450  L=1,N
CONTINUE
DO 710 I=1,N
L=N+1-I
IF (INDEX(L,1)-INDEX(L92))
JROW=INOEX(L,l)
JCOLUM=INDFX(L,2)
DO 705 K=1,M
SWAP=A(K.JROW)
A(K9JROW)=A(K,JCOLUM)
                            630,  710» 630
                                                                   00000010
                                                                   000000?0
                                                                   ooonoo TO
                                                                   OOOOOOAO
                                                                   ooooooso
                                                                   OOOOOOnO
                                                                   000 (J 0070
                                                                   000000^0
                                                                   OOOOOOQO
                                                                   00000100
                                                                   000001 in
                                                                   000001?u
                                                                   00000130
                                                                   00000140
                                                                   000001SO
000001^0
0 0 0 0 0 1 « 0
000001^0
00000200
OOOOOR10
'000002? 0
00000230
OOOOOPAO
000002^0
OOOOOP60
00000270
000002HO
000002^0
00000300
00000310
000003PO
00000330
00000340
00000350
00000360
00000370
00000 3 flO
00000390
00000400
00000410
00000420
00000430
00000440
00000450
000004AO
00000470
000004PO
000004QO
0000050C
00000510
ooooo5?n
0000053C
0000054C

-------
                                                                         267
               EXHIBIT A-5.  LISTING OF SUBROUTINE MATINV (Concluded)


 700 A(K,JCOLUM)=SWAP
 705 CONTINUE                                                             00000550
     .11 = 1                                                               00000560
                                                                          00000570
 710 CONTINUE
740 'GO TO 780                                                            000005«0
 760 DETERM = 0.                                                          00000590
     Jl - -1                                                              00000600
780  KK = 0                                                               00000610
     DO 68 I =  1,N                                                        000006PO
     00 68 J =  1,N                                                        00000630
     KK = KK *  1                                                          000006AQ
68   PSAVE(KK). =  A(J,I,
     RETURN                                                               00000660
     'E                                                                   00000670
                                                                          00000630

-------
                                                                    268
               EXHIBIT A-6.  LISTING OF SUBROUTINE PEDERV
SUBROUTINE  PEDERV(T»Y«PSAVE»N)                                      0000001C
DOURLE PRECISION  T, Y                                               OOOOOOPC
DIMENSION Y(8,40)«  PSAVE(1600)                                      OOOOOO^f
RETURN                                                               OOOOOO^C
                                                                     0000005C

-------
                    EXHIBIT A-7.  LISTING OF SUBROUTINE PLOT
                                                                           269
C SUBROUTINE *****•» PLOT *»»*«•»
C
C THIS SUHROUTINF READS THE PLOT CARDS  AND  PLOTS  THE  RESULTS AS PART
C OF THE PRINTED OUTPUT -- IT DOES NOT  DRIVE  A  PLOTTER.
C
C SYMBOL DESCRIPTIONS —
C
C CGRID    THE LENGTH OF THE VERTICAL AXIS, PPM
C CriIGH    HIGHEST CONCENTRATION VALUE? PPM
C CLOW     LOWEST CONCENTRATION VALUE,  PPM
C CSPAN    CONCENTRATION NORMALIZATION  FACTOR
C DATA     CONCENTRATION DATA POINTS? PPM,  UP TO  80
C J        DO-LOOP INDICES OR LOCAL POINTERS
C JBLANK   A HOLLERITH WORD OF FOUR BLANK CHARACTERS
C JCONC    CONCENTRATION LABELS
C JFACT    CONVERSION FACTOR FOR LABEL
C JGRID    THE PLOTTING GRID
C JSTAR    THE CHARACTER «*»
C JSYMB    SYMBOL TO BE USED FOR PLOTTING SAVED POINTS
C JVERT    VERTICAL LEGEND
C K        DO-LOOP INDICES OR LOCAL POINTERS
C KCON     CONCENTRATION COORDINATE ON  GRID
C KTIM     TIME COORDINATE ON GRID        .. ., ..-.;•,
C L        DO-LOOP INDICES OR LOCAL POINTERS  " "*
C M        DO-LOOP INDICES OR LOCAL POINTERS
C MAXCON   LIMIT ON NUMBER OF VERTICAL  POINTS
C MAXPNT   MAXIMUM NUMBER OF SAVED TIME AND CONCENTRATION  POINTS
C MAXTIM   LIMIT ON NUMBER OF HORIZONTAL POINTS
C N        DO-LOOP INDICES OR LOCAL POINTERS
C NAME     SPECIES NAMES, ONE PER SPECIES
C NDAT     NUMBER OF CONCENTRATION DATA POINTS
C NIN      TH-E FORTRAN INPUT UNIT (NORMALLY 5)
C NOUT     THE FORTRAN OUTPUT UNIT NUMBER (NORMALLY 6)
C NPNT     NUMBER OF SAVED TIMES AND CONCENTRATIONS
C NTEST    SPECIES NAME FOR TESTING
C NTIT     USER-INPUT TITLE FOR PRINTOUT? 3 FOUR-CHARACTER WORDS
C NTOT     TOTAL NUMBER OF SPECIES
C SAVCON   SPECIES CONCENTRATIONS? PPM? ONE PER SPECIES  AT 80  TIMFS
C SAVTIM   TIMES THAT CONCENTRATIONS ARE SAVED? MIN,  UP  TO 80  VALUES
C TGRI.D    THE LENGTH OF THE HORIZONTAL AXIS? MIN
C THIGH    HIGHEST TIME VALUE? MIN
C TIME     TIMFS AT WHICH CONCENTRATIONS ARE  INPUT? MIN? UP TO BO
C TLOW     LOWEST TIME VALUE? MIN
C TPRINT   TIMES FOR PRINTOUT ON HORIZONTAL AXIS?  MIN
C TSPAN    TIME NORMALIZATION FACTOR
C
C BEGINNING OF PROGRAM.
C
C ENTRY POINT
      SUBROUTINE PLOTtNTIT, NPNT, NTOT? NAME? SAVTIM? SAVCON)
C
C SET DIMENSIONS OF INCOMING ARRAYS
C
      DIMENSION SAVCON(50?80)? SAVTIM(80),  NTIT(3)» NAME(50)
oooooon
000000?!
00000031
00000041
00000051
OOOOOOM
00000071
00000081
00000091
00000101
oooooi n
00000121
00000131
0 0 0 0 0 1 * i
00000151
0 0 0 0 0 1 f- 1
00000171
oooooia<
00000191
0000020(
000002K
0000022(
0000023(
0000024C
0000025'
00000261
0000027(
00000?P(
0000029(
0000030(
0000031 '
0000032(
0000034(
0 0 0 0 0 3 5 (
0000036(
0000037(
0 0 0 0 0 3 B (
00000391
0000040(
0000041C
0000042C
00000431
0000044(
0000045f
0000046C
00000471
0000048C
0000050C
0000051C
0000052(
0000053C
0000054C

-------
             EXHIBIT A-7-  LISTING OF SUBROUTINE PLOT (Continued)
                                                                      270
SET DIMENSIONS OF LOCAL  ARRAYS

    DIMENSION JVEPT(52,2>,  JCONCC5),  TIMEC80),
   .DIMENSION JGRID<121,52)»  TPRINT(9)

DEFINE THE VERTICAL AXIS VIA  DATA  STATEMENTS
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
    DATA
DEFINE THE VERTICAL LABEL
                                       DATA(80:
JGRIDd ;1) /1H-/, JGRID (1,2) /1H I/, JGRID (1,3) /1HI/
JGRIDd »4) /1H| /, JGRIDd ,5)/lHI/,JGRID(1.6)/lHI/
JGRIDd, 7) /I H| /, JGRID (1,8) /I HI /» JGRID (1,9) /1H|/
JGRIDd ?10)/1H
JGRI"D(1,13)/1H
JGRID ( 1, 16) /1H
JGRIDd, 19 )/lH
JGRIDd, 2?)/lH
JGRIDd, 25) /1H
JGRIO (1,28) /1H
JGRIDd »31) /1H
JGRIDd, 34J/1H
JGRID(1;37)/1H
JGRID (l»4n) /1H-
JGRIDd ,43) /1H
JGRID  N'OUT  /6/,  JBLANK
         MAXTIM /121/,  MAXCON  /52/,
    DATA JSTAR /lH*/» TGRID  /1HO./,
                           /4H     /
                            MAXPNT /80/
                            CGRID  /52./
SET LOOP FOR ALL SPECIES  AND  CLEAR  GRID




P40
250
C
C READ
C
DO 360 N =
DO 250 K =
DO 240 J =
JGRIDU, K)
CONTINUE
CONTINUE

1,
1,
2,
=



PLOT CONTROL


NTOT
MAXCON
MAXTIM
JBLANK



CARD

    READ (NIN54,END=900)  NTEST,  NDAT,
   S,     CLOW, CHIGH, TLOW,  THIGH

TEST FOR END OF PLOTTING
                              JSYMB,  JFACT, JCONC,
                                                                         OOOOOBSr
                                                                         OOOOOS6C
                                                                         0000057C
                                                                         0 0 0 0 0 5 P f.
                                                                         0000059f
                                                                         ooooo^or
                                                                         0000061C
                                                                         000006PC
                                                                         0000063f
                                                                         0000064f
                                                                         0000065C
                                                                         0000066C
                                                                         0000067C
                                                              0000069f
                                                              0000070f
                                                              000007H
                                                              0 0 0 0 0 7 2 (
                                                              0000073f
                                                              ooono74(
                                                              ooooo7c;f
                                                              0 0 0 0 0 7 6 f.
                                                              0000077f
                                                              0000079C
                                                              OOOOOROt
                                                              0 0 0 0 0 R 1 f.
                                                              OOOOORPC
                                                              0 0 0 0 0 R 3 r
OOOOOR5C
oooooe^c
OOOOOB7f
OOOOORRr
OOOOOR9(
OOOOOQOf
000009K
000009?f
0000093f
000009M
000009S(
                                                                         0 0 0 0 0 9 7 (
                                                                         000009Ri
                                                                         000009Q(
                                                                         OOOOlOOf
                                                                         000010K
                                                                         ooooin?(
                                                                         oooni03(
                                                                         00001041
                                                                         OOOOIOM
                                                                         0000106(
                                                                         00001071
                                                                         000010R1
                                                                         0000109(
                                                                         00001101

-------
               EXHIBIT A-7.  LISTING OF SUBROUTINE PLOT (Continued)
                                                                       271
      IF (NTEST .EQ. JBLANK) GO TO 800
C
C TEST NUMBER OF DATA POINTS AND READ DATA
C
      IF (NDAT .IE. 0) GO TO 308
      IF (NDAT .LE. MAXPNT) GO TO 305
      WRITE (NOUT»10?0) MAXPNT
      GO TO 900
C
C READ DATA POINTS
C
305   READ (NIN»5)  (TIME(J), DATA(J)» J  =  1»NDAT)
C
C SET NORMALIZATION FACTORS AND VERTICAL CONCENTRATION  LABELS
C
308   CSPAN = CGRID /  (CHIGH - CLOW)
      TSPAN = TGPID /  (THIGH - TLOW)
      JVERTtl,?) = JCONC(5)
      JVERT(14?2)  = JCONC(4)
      JVERT(27»2)  = JCONC(3)
      JVERT(40»2>  = JCONC(2)
C
C SET HORIZONTAL TIME LABELS
C
      DO 310 J = 1,9
      TPRINT(J) = FLOATtJ - 1) / 8. *  (THIGH  -  TLOW)  +  TLOW
310   CONTINUE
C
C TEST FOR CORRECT SPECIES NAME
C
      DO 320 L = 1«NTOT
      IF (NTEST .EQ. NAME(D) GO TO 325
320   CONTINUE
      WRITE (NOUT?1021) NTEST
      GO TO 360
C
C IF THERE ARE DATA POINTS, GET THEIR COORDINATES
C
325   IF (NDAT .LE. 0) GO TO 335
      DO 330 J = 1»NDAT
      KTIM = IFIX((TIME(J) - TLOW) * TSPAN  +  1.5)
     "KCON = IFIX((DATA(J) - CLOW) * CSPAN  -  1.5)
      KCON = MAXCON - KCON
C
C CHECK FOR BEING WITHIN GRID? THEN PLACE  ON  GRID
C
      IF (KTIM .LT. 2) GO TO 330
      IF (KCON .LT. 1) GO TO 330
      IF (KTIM .GT. MAXTIM) GO TO 330
      IF (KCON.GT. MAXCON) GO TO 330
      JGRID(KTIM»KCOM) = JSTAR
330   CONTINUE
C
C IF THERE ARE CALCULATED POINTS, GET THEIR COORDINATES
C
335   IF (NPNT .LE. 0) GO TO 345
0000111C
0000112C
ooonii3C
00001141'
OOOOll^f
0000117*:
000011«(
OOOOll^C
0000120f
000012H
0000122C
0000123T
000012?'
0 0 0 0 1 ? 6 f
0000127C
000012PL
0000 IP^f.
0 0 0 0 1 3 0 f.
0000131C
0000132C
0000133'".
000 01 34 C
000013SC
0 0 0 0 1 3 6 C
0 0 0 0 1 3 7 r
0000139C
0000140 c:
ooooi4ir
0000142C
0000143C
0000 144 r,
000 01 45 C
0000146C
0000147C
0 0 0 0 1 4 P C
0000149C
0000150C
0000151C
0000152C
0000 153C
0000154C
0000155C
0000156f
0000157C
000015BC
0000159C
0000160f
0000161C
000016?C
0000163C
000016fC
0000165C
000 01 66 C

-------
               EXHIBIT A-7.  LISTING OF SUBROUTINE PLOT (Concluded)
                                                                       272
      DO 340 J = 1,NPNT
      KTIM = IFIX((SAVTIM(J) -  TLOW)  *
      KCON = IFIX((SAVCON(L,J)  -  CLOW)
      KCON = MAXCON - KCON
                                        TSPAN + 1.5)
                                        *  CSPAN - 1.5)
  CHECK FOR BEING WITHIN GRID, THEN  PLACE  ON  GRID
      IF (KTIM .LT. 2) GO TO 340
      IF (KCON .LT. 1) GO TO 340
      IF (KTIM .GT. MAXTIM) GO TO
      IF (KCON ,GT. MAXCON) GO TO
                         JSYM8
                                   340
                                   340
340
C
C SKIP
C
345
      JGRID(KTIM.KCON) =
      CONTINUE
       A PAGE? THEN PRINT THE VERTICAL  AXIS  AND  GRID

      WRITE (NOUT;1014)
      DO 350 K = 1,MAXCON
      WRITE (NOUT,1015) JVERT(K,1), JVERT(K,2>»
     &     (JGRir>( J,K) , J =  1,MAXTIM)
350   CONTINUE
C
C PRINT THF HORIZONTAL AXIS  AND LABELS
C
      WRITE
      WRITE
            (NOUT, 1016) JCONC(l)
            (NOUT.1017) TPRINT
      WRITE (NOUT.1018) NTIT, NAME(L), JFACf
      CONTINUE
360
C
C END OF SUBROUTINE -- RETURN TO CALLER
C
800   RETURN
900   STOP
C
C LIST OF FORMAT STATEMENTS
C
4
5
 1014
1015
 1016
 1017
 1018
     FORMAT (A4, IX, 12, IX. Al»  IX, 6(A4?
     FORMAT (8F10.0)
     FORMAT (1H1)
     FORMATdX, 2A4, 121A1)
     FORMAT (5X. A4, 1H+. B(15H	——	
     FORMAT (F12.2-. 8F15.2, /, 62X,  14HTIME
     FORMAT (27X* 11HFIGURE   .  , 3A4, 12H.
    8,     5X»  PRHCONCENTRATION SCALE FACTOR
1020 FORMAT (33H PROGRAM CANNOT  HANDLE MORE
    fii     2RH  PLOT POINTS — JOB  ABORTED.)
1021 FORMAT (14H1SPECIES NAME ,  A4,  21H NOT
    &     23H  SKIPPING TO NEXT PLOT.)
     END
                                                  4F10.0)
— I ) )
 (MINUTES).
  SPECIES:
:  9  A4)
 THAN  ,  14,
                                                         /)
                                                          A4,
                                              IN  SPECIES  LIST.
00001
0 0 0 0 1 ft R (
0000169(
0000171K
0000171C
00001 7P(
0000173f
0000174C
0 0 0 0 1 7 is f
0000177C
n000178(
OOOOlBOf
0 0 0 0 1 8 1 1
00001«?(
0 0 0 0 1 « ? (
00001«5f
00001flf>(
ooooiRrr
OOOOlRHt
00001fi9(
o o o o i q o (
0000191C
0000 19? f
0 0 0 0 1 9 3 (
0000194C
00001 95 (
000019AC
0000197C
0000198C
0000199C
OOOOPOOt
000020K
000020?f
0000203f
0000204(
000020S(
0000?06(
0000207C
OOOOPORf
0000209(
0000210(
000021K
0000212C
00002131
0000214(
000021S!

-------
                                                   273
EXHIBIT A-8.   SAMPLE MODKIN INPUT
SAMPLE DECK
5.0
N02
0
03
03
!M03
N03
N205
N205
NO
HN02
HNI02
OH
H02
H02
H202
ALD
ALD
R02
OH
RC03
RC03
RO
RO
RO
R02
R02
0
0
0
NO
HN02
H02
R02
N03
N205
OLEF
OLEF
OLEF
OLEF
PROP
ETHY
OLEF
PROP
ETHY
OLEF
PROP
ETHY
N02
NO
03
N205
H02
R02
RC03
HN02
375,0

02 M
NO
N02
NO
NO?

H20
N02 H20
HN02

N02
NO
H02


OH
NO
NO
NO
N02
02
NO
N02
R02
H02
NO
N02
N02
HN03
HN03
502
S02
S02
S02
0
03
OH
2
0
0
2
03
03
2
OH
OH
0.08
0.27
0.022





38 3 13
0.0001 0
1NO
103
1N02
1N03
2N02
1N205
1N02
2HN03
2HN02
1NO
10H
1HN03
10H
1H202
20H
0.63R02
0.63PC03
IRQ
1HN02
1R02
1PAN
1H02
1RN02
1RN03
2RO
IRQ
1N02
1NO
1N03
1HN02
1H20
1S03
1S03
1S03
1S03
1R02
1RC03
1R02

1R02
IRQ?

1RC03
1RC03

1R02
1R02








4 4- 2 4 5 1
.00001 5.0
10
1M
102
102


1N03


1N02 1H20
1NO

1N02
102

1.37H02
0.37H02 1H20
1N02

1N02 1C02

1ALD



10H

102

1N02
2N02
10H
1RO
1N02
2N02
0.5RC03 0«5H02
1RO 1ALD
1ALD

0.5RC03 0.5H02
ORC03 1H02

IRQ 1ALD
1RO 1ALD

1ALD
1ALD








0.001 0.00835
2c66E-l
2.00E-5
2.08E+1
4865E-2
lc50E+4
4.50E+3
2.70E+1
l.OOE-5
2.10E-ft
4.50EOO
1.30E-2
1.50E+4
7«OOE*2
5.30E+3
1.06E-3
2.50E-3
2.30E+4
9.10E*2
1.20E+4
9.10E+2
l.OOE+2
2.40E-2
2.50Ef2
4.90E+2



1.38E+4



4.5000E-1
6.0000E-1
1.5000E+4
4.0000E-1
1 .9776E*3
0.64E-2
0.70E+4

6.80E+3
7.72E+2

1.60E-2
4.00E-3

2.50E+4
2..50E+3









-------
                                                              274
EXHIBIT A-8.   SAMPLE  MODKIN  INPUT  (Concluded)
H202
ALD
S02
OLEF
HN03
0
N03
OH
RO
PAN
RNO?
RN03
503
ETHY
PROP
M
02
H20
C02
S02


N02


NO


ETHY


PROP


N02



NO


03



S02



PAN


OLEF




0.10
0.334
1.90
l.OE-9
l.OE-10
5.0E-10
0.0
0.19
1.71
1000000.0
210000.0
20000.0
lOOOOcO
5
0
231
7
0
157
7
0
157
7
a
177
7
0
177
9 0 10*0
8
117
297
7 0 10+0
8
117
10 0 10+0
11
.118
298
10 0 10+0
11
108
305
7 0 10+0
30
225
16 0 10+0
16
66
116
237

0.291
0*328

0.035
0.040

0.35
0.36

0.184
0.197

1.656
1.773
0.00 0.15
0.09
0.26
0.090
0.00 0.15
0.26
0.00
0.00 0.15
0.025
0.54
0.45
0.00 0.15
0.353
0.136
0.241
0.00 0.15
0.000
0.211
0.00 0.50
1.R5
1.16
0.71
0.69

49


29
187

29
187

35
271

35
271
0.30 0.45
29
157

0.30 0.45
29
157
0.30 0.45
32
158
298
0.30 0.45
11
173
305
0.30 0.45
101
266
1.00 1.50
31
71
132
261

0.241


0.039
0.043

0.36
0.39

0.193
0.200

1.737
1.800
0.60
0.27
0.12

0.60
0.16
0.010
0.60
0.036
0.46
0.43
0«60
0.384
0.291
0.415
0.60
0.054
0.220
2.00
1.78
1.09
0.69
0.70

108


64
297

64
297

91
330

91
330
0.0
44
187

O.'O
44
187
0.0
47
188

0.0
47
231

0.0
162
311
0.0
42
86
151
301

0.396


0.040
0.034

0.36
0.31

0.201
0.204

1.809
1.836
0.6
0.40
0.14

0.60
0.05
0.00
0.6
0.314
0.41

0.6
0.285
0.173

0.6
0.109
0.213
2.0
1.63
0.88
0.70
0.68

173


117


117


121


121

0.0
64
228

0.0
64

0.0
67
229

0.0
108
231

0.0
185

0.0
51
102
191
323

0.409


0.048


0.43


0.202


1.818

400.0
0.33
0.12

400.0
0.00

400.0
0.496
0.41

400.0
0.186
0.080

400.0
0.197

400.0
1.56
0.72
0.69
0.65

-------
                                                                        275






                  EXHIBIT A-9.  SAMPLE MODKIN OUTPUT—SELECTED PAGES





                    MODULAR KINETICS  RUN  NO.  SAMPLE DECK
TOTAL NUMBER OF REACTIONS =  38




NUMBER OF LUMPED REACTIONS =   3




NUMBER OF DIFFERENTIAL SPECIES =   13




NUMBER OF STEADY STATE SPECIES =    4




NUMBER OF UNCOUPLED SPECIES =   4




NUMBER OF REPLACEMENT SPECIES =    2 '




NUMBER OF INERT OR CONSTANT SPECIES =    4




NUMBER OF FLOWING SPECIES =   5




REACTION RATE PRINT REQUEST FLAG  =   1




TIME INCREMENT =    5.000E 00 MINUTES




ENDING TIME =    3.750E  02 MINUTFS




STARTING STEP SIZE =     l.OOOE-04  MTNUTES




MINIMUM STEP SIZE =    l.OOOE-05  MINUTES




MAXIMUM STEP SIZE =    5.000E 00  MINUTES




CONVERGENCE TOLERANCE =    l.OOOE-03




DILUTION RATE =    8.350E-03 MINUTES<-1>

-------
                                                                         276
            EXHIBIT A-9.  SAMPLE MODKIN OUTPUT—SELECTED PAGES (Continued)
    P.  CONST.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
2.660E-01
2..000E-05
2.080E 01
4.650E-02
1.500E 04
4.500E 03
2.700E 01
l.OOOE-05
2.100E-06
4.500E 00
1.300E-02
1.500E 04
7.000E 02
5.300E 03
1.060E-03
2.500E-03
2.300E .04
9.100E 02
1.200E 04
9.100E 02
l.OOOE 02
2o400E-02
2.500E 02
4.900E 02
0.0
OoO
0.0
1.380E 04
0.0
0.0
0.0
4.500E-01
6.000E-01
1.500E 04
4.000E-01
1.978E 03
6.400E-03
7.000E 03
      LIST OF REACTIONS

REACTANTS            PRODUCTS
                   H20

02
NO
N02
NO
N02

H20
N02
HN02

N02
NO
H02


OH
NO
NO
NO
N02
02
NO
N02
RO?
H02
NO
N02
N02
HN03
HN03
S02
S02
S02
S02
0
03
OH
N02
0
03
03
N03
N03
N205
N205
NO
HN02
HN02
OH
H02
H02
H202
ALO
ALD
R02
OH
RC03
RC03
RO
RO
RO
R02
R02
0
0
0
NO
HN02
H02
R02
N03
N205
OLEF
OLEF
OLEF
=
=
=
=
=
t:
=
=
=
=
=
=
=
=
z
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
—
=
1
1
1
1
2
1
1
2
2
1
1
1
1
1
2
0
0
1
1
1
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
«
e
9
e
«
a
©
e
e
o
o
o
e
a
»
a
0
e
o
e
0
0
e
a
»
o
«
a
o
e
o
o
e
a
e
*
a
c-
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
A3
63
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
NO
03
N02
N03
N02
N205
N02
HN03
HN02
NO
OH
HN03
OH
H202
OH
R02
RC03
RO
HN02
R02
PAN
H02
RN02
RN03
RO
PO
N02
NO
N03
HN02
H20
503
S03
S03
S03
R02
RC03
R02
1
1
1
1


1


1
1

1
1

1
0
1

1

1



1

1

1
2
1
1
1
2
0
1
1
.00
.00
»00
.00


.00


.00
.00

.00
.00

.37
.37
.00

.00

.00



.00

.00

.00
.00
.00
.00
.00
.00
.50
.00
.00
0
M
02
02


N03


N02
NO

N02
02

H02
H02
N02

N02

ALD



OH

02

N02
N02
OH
RO
N02
N02
RC03
RO
ALD









1.00 H20






1.00 H20


1.00 C02















0.50 H02
1.00 ALD

THE FOLLOWING SET OF  2 REACTIONS CORRESPONDS  TO  REACTION NUMBER  36
39  6.800E 03
40  7-720E 02
   0    PROP =  1.00 R02    0.50  RC03   0.50  H02
   0    ETHY =  1.00 R02    0.0   RC03   1.00  H02
THE FOLLOWING SET OF  2 REACTIONS CORRESPONDS  TO  REACTION NUMBER  37
41  1.600E-02
42  4.000E-03
   03   PROP =  1.00 RC03   1.00  RO     1.00  ALD
   03   ETHY =  1.00 RC03   1.00  RO     1.00  ALD
THE FOLLOWING SET OF  2 REACTIONS CORRESPONDS  TO  REACTION NUMBER  38
43  2.500E  04
44  2.500E  03. '
   OH   PROP =  1.00 R02    1.00  ALD
   OH   ETHY =  loOO R02    1.00  ALD'

-------
                                                                      277
             EXHIBIT A-9.  SAMPLE MODKIN OUTPUT-SELECTED PAGES (Continued)
                    INITIAL SPECIES CONCENTRATIONS


SPECIES   VALUE   SPECIES   VALUE   SPECIES    VALUE    SPECIES    VALUE

DIFFERENTIAL(PPM)

  NO?   8.000E-0?   NO    2.700E-01   03     2.200E-02    N205   0.0
  HO?   0.0
  H202  0.0
  HM03  0.0
R02   0,0
RC03  0.0
                                    HN02   0.0
ALD   l.OOOE-01   S02   3.340E-01   OLEF   1.900E  00
STEADY STATE(PPM)

  0     l.OOOE-09   N03   loOOOE-10   OH     5oOOOE-10    RO     0.0

UNCOUPLED(PPM)

  PAN   0.0         RN02  0.0         RN03   0.0          S03    0»0

REPLACEMENT(PPM)

  ETHY  1.900E-01   PROP  1.710E 00
INERT/CONSTANT(PPM)

  M     l.OOOE 06   02
      2.100E 05   H20   2.000E 04   C02    l.OOOE  04
                    TIME =    5.021E  00 MINUTES


SPECIES   VALUE   SPECIES   VALUE   SPECIES    VALUE    SPECIES    VALUE

DIFFERENTIAL(PPM)
  N02   1.856E-01   NO
  HO?   1.367E-04   R02
  H202  1.937E-04   ALD   1.904E-01
  HN03  2.405E-03
      1.505E-01   03    1.436E-02   N205  5.113E-07
      9.789E-05   RC03  80061E-06   HN02  1.142E-02
                  S02   3.319E-01   OLEF  1.773E  00
STEADY STATE(PPM)

  0     1.172E-08   N03   1.706E-08   OH     2.960E-07    RO     2.670E-06

UNCOUPLED(PPM)

  PAN   2.988E-04   RN02  5.147E-04   RN03   7.154E-04    S03   3.069E-04

REPLACEMENT(PPM)

  ETHY  1.816E-01   PROP  1.5S91E  00

-------
                                                                         278
            EXHIBIT A-9.  SAMPLE MODKIN OUTPUT-SELECTED PAGES  (Continued)
              REACTION RATES  (SORTED  INTO  DECREASING SIZE)
  NO.
RATE
NO.
RATE
NO.
                                       RATE
                                        NO.
                                     RATE
                                        NO.
                               RATE
1
18
1?
24
23
32
8
27
4.94E-02
1.34E-02
8.24E-04
2.43E-04
l.OOE-04
2.04E-05
1.02E-07
0.0
2
38
10
21
14
33
35
26
4.92E-02
1.19E-02
5.87E-04
1.50E-04
9.91E-05
le95E~05
6.79E-08
0.0
3
17
19
11
34
6
31
25
4.49E-02
1.30E-03
5.34E-04
1.48E-04
8.50E-05
1.43E-05
0.0
0.0
13
9
16
36
5
7
30

1
1
4
1
3
1
0

.44E-02
.17E-03
.76E-04
.28E-04
.85E-05
.38E-05
.0

22
20
37
4
28
15
29

1.
1.
3.
1.
3.
2.
0.

35E-02
10E-03
76E-04
24E-04
OOL-05
05E-07
0

                    TIME =






SPECIES   VALUE   SPECIES



DIFFERENTIAL(PPM)
                       1.007E 01  MINUTES
                     VALUE
                    SPECIES
                        VALUE
                      SPECIES
                           VALUE
N02
K02
H202
HN03
2.753E-01
4.046E-04
2.104E-03
7.896E-03
NO
R02
ALD
5.053E-02
2.890E-04
3.073E-01
03
RC03
S02
5.492E-02
3.710E-05
3.285E-01
N205
HN02
OLEF
5.508E-06
1.311E-02
1.634E 00
STEADY STATE(PPM)
        1.738E-08   N03   1.230E-07
                               OH
                            2.982E-07
UNCOUPLED(PPM)
  PAN
2..737E-03   RN02   8.319E-04    RN03  2.355E-03
                                  RO
                                        S03
                               2.827E-06
                                        2.064E-03
REPLACEMENT(PPM)
  ETHY  1.733E-01   PROP  1.A61E  00




              REACTION RATES  (SORTED  INTO  DECREASING SIZE)
  NO.
RATE
NO.
RATE
NO,
RATE
                                        NO.
                                     RATE
                                                               NO.
                                                              RATE
1
18
12
4
36
28
8
27
7.32E-02
1.33E-02
1.23E-03
7.03E-04
1.75E-04
6.60E-05
1.10E-06
0.0
' 2
38
21
34
11
32
35
2ft
7.30E-02
1.10E-02
1.02E-03
6«06E-04
1.70E-04
5.98E-05
7.24E-07
0.0
3
17
14
9
6
33
31
25
5.77E-02
2.11E
8.68E
5.84E
1.52E
5.70E
0.0
0.0
-03
-04
-04
-04
-05


13
20
10
24
7
23
30

K43E-02
1.71E-03
7.74E-04
3.81E-04
1.49E-04
3.57E-05
OoO

22
37
16
19
5
15
29

1
1
7
1
9
2
0

.43E-02
.3PE-03
.68E-04
.81E-04
.32E-05
.23E-06
.0


-------
                                                                        279
            EXHIBIT A-9.  SAMPLE MODKIN OUTPUT-SELECTED PAGES (Continued)

INCOMING S02  CONCENTRATION CHANGED TO   3.960E-01  AT    1.103E 02 WIN.
                    TIME =
                      1.160E 02 MINUTES
SPECIES   VALUE   SPECIES   VALUE   SPECIES    VALUE    SPECIES   VALUE

DIFFERENTIALtPPM.)
N02
H02
H202
HN03
1.451E-01
5.998E-04
1.795E-01
3.199E-02
NO
R02
ALD
7o451E~03
4.193E-04
6.646E-01
03
RC03
S02
2.239E-01
1.155E-04
lo532E-01
N205
HN02
OLEF
1.497E-05
2.416E-03
3.383E-01
STEADY STATE(PPM)

  0     9.179E-09

UNCOUPLED(PPM)

  PAN   1.826E-01

REPLACEMENT(PPM)
            N03
         6.252E-07
               OH
            1.454E-07
                                                 RO
                        7.696E-07
            RN02  6.014E-04   RN03   1.037E-02    S03    1.405E-01
  ETHY  6.209E-02   PROP  2.762E-01

              REACTION RATES  (SORTED  INTO DECREASING  SIZE)
  NO.
RATE
NO.
RATE
NO.
RATE
NO.
RATE
NO,
RATE
1
18
4
6
24
10
23
27
3.86E-02
2.84E-03
1.51E-03
4.08E-04
5.47E-05
2.63E-05
1.43E-06
0.0
2
17
34
7
9
2fi
35
26
3.86E-02
2.22E-03
1.44E-03
4.04E-04
4.54E-05
1.84E-05
9.17E-07
0.0
3
14
37
12
32
36
31
25
3.
1.
1.
3.
4.
1.
Oo
0.
47E-02
91E-03
05E-03
16E-04
14E-05
78E-05
0
0
22
21
38
15
33
19
30

3.88E-03
1.68E-03
1.03E-03
1.90E-04
3.86E-05
1.30E-05
0.0

13
16
20
5
11
8
29

3
1
7
6
3
2
0

.13E-03
.66E-03
.84E-04
.99E-05
.14E-05
.99E-06
.0

INCOMING N02  CONCENTRATION CHANGED TO   40flOOE-02  AT    1.185E 02 WIN.

INCOMING NO   CONCENTRATION CHANGED TO   4.300E-oi  AT    I.ISSE 02 MIN.
                    TIME =
                      1.204E 02 MINUTES
SPECIES   VALUE   SPECIES   VALUE   SPECIES    VALUE    SPECIES    VALUE

DIFFERENTIAL(PPM)
  N02   1.458E-01
  H02   5.840E-04
  H202   1.802E-01
  HN03   3.225E-02
            NO    7.718E-03   03    2.208E-01    N205   1.46RE-05
            R02   3.974E-04   RC03  1.112E-04    HN02   2.422E-03
            ALD   6.492E-01   S02   1.554E-01    OLEF   3.173E-01

-------
                                                                       280
            EXHIBIT A-9.  SAMPLE MODKIN OUTPUT—SELECTED PAGES (Continued)
STEADY STATE(PPM)

  0     9.225E-09

UNCOUPLED(PPM)
            N03
         6.102E-07
               OH
            1.514E-07
  PAN   1.831E-01   RN02  5.857E-04   RN03   1.023E-02

REPLACEMENT(PPM)

  ETHY  5.951E-02   PROP  2.57BE-01
                                                 RO
                                                 S03
              REACTION RATES  (SORTED INTO DECREASING  SIZE)
  NO.
RATE
NO,
RATE
NO,
RATE
                                               NO.
                                              RATE
                        7.428E-07
                                              1.420E-01
NO.
RATE
   1  3.88E-02
  18  2.79E-03
   4  1.50E-03
   6  4.00E-04
  24  5.31E-05
  10  2»64E-05
  23  1.43E-06
  27  0.0
          2  3.87E-02
         17  2.26E-03
         34  1.42E-03
          7  3.96E-04
          9  4.73E-05
         28  1.86E-05
         35  9.12E-07
         26  0»0
                3  3*54E-02
               14  lo81E-03
               38  1,OOE-03
               12  3.31E-04
               32  4.08E-05
               36  1.67E-05
               31  0»0
               25  0»0
22
16
37
IS
33
19
30
3.74E-03
1.62E-03
9.70E-04
1.91E-04
3.71E-05
1.40E-05
0.0
13
21
20
5
11
8
29
3.16E-03
1.62E-03
7o81E-04
7.06E-05
3.15E-05
2.93E-06
0.0
INCOMING ETHY CONCENTRATION CHANGED TO
INCOMING PROP CONCENTRATION CHANGED TO
                                 2.020E-01  AT
                                 1.818E  00  AT
                                        1.226E  02  MIN.
                                        1.226E  02  MIN.
                    TIME =
                      1.298E 02 MINUTES
SPECIES
  VALUE
 SPECIES
DIFFERENTIAL(PPM)
     VALUE   SPECIES
              VALUE
                SPECIES
                                                                 VALUE
N02
H02-
H202
HN03
1.514E-01
5.474E-04
1.801E-01
3.307E-02
NO
R02
ALD

8.386E-03
3.493E-04
6»146E-01

                                      03    2.122E-01
                                      RC03  1.003E-04
                                      S02   1.600E-01
                                                 N205   1.474E-05
                                                 HN02   2.508E-03
                                                 OLEF   2.763E-01
STEADY STATE(PPM)

  0     9.581E-09
            N03
         5.899E-07
               OH
            1.647E-07
                                                 RO
                                              6.859E-07

-------
                                                                        281
           EXHIBIT A-9.  SAMPLE MODKIN OUTPUT-SELECTED PAGES (Continued)
                    TIME =
                       3.760E  02 MINUTES
SPECIES   VALUE   SPECIES    VALUE'   SPECIES   VALUE   SPECIES    VALUE



DIFFERENTIAL(PPM)
N02
HO?
H202
HN03
1.518E-01
6.287E-06
3.019E-02
6.088E-02
NO
R02
ALD

1.044E-01
2.154E-06
5«619E-02

03
RC03
S02

K856E-02
1.025E-06
2.635E-01

N205
HN02
OLEF

5.930E-07
1.166E-02
1.002E-02

STEADY STATE(PPM)




  0     9.607E-09




UNCOUPLED(PPM)




  PAN   5«204E-02




REPLACEMENT(PPM)
            N03
2e370E-08
OH
1.367E-07
            RN02   1.897E-04    RN03  2.607E-03
RO
                               S03
4.048E-08
                         7.403E-02
  ETHY  5.277E-03   PROP   4.743E-03
  NO.
      REACTION RATES  (SORTED  INTO DECREASING SIZE)




RATE     NOo   RATE      NO.    RATE     NO.   RATE
                                                               MO,
                                           RATE
1
13
19
34
6
23
8
27
4.04E-02
4.62E-04
1.71E-04
9.37E-05
1.62E-05
1.06E-06
1.19E-07
0.0
2
12
11
5
1
32
35
26
4.
3.
1.
3.
1.
7.
6.
0.
03E-02
11E-04
52E-04
71E-OS
60E-05
49E-07
25E-08
0
3
18
16
15
21
36
31
25
4.03E-02
2.06E-04
1.40E-04
3.20E-05
1.57E-05
3.65E-07
0.0
0.0
9
22
4
28
24
33
30

6.66E-04
2.04E-04
1.31E-04
2.01E-05
3.01E-06
3.42E-07
0.0

10
17
20
38
37
14
29

6
1
9
1
1
2
0

.11E-04
.77E-04
.80E-05
.88E-05
.88E-06
.11E-07
.0

THIS RUN TERMINATED WITH KFLAG  =   1

-------
                                  EXHIBIT  A-9.  SAJ1PLE MODKIN  OUTPUT—SELECTED PAGES (Concluded)
   0.60-
   0.45
C
0
H
C
E
N
T
R
A  0.30
T
I
0
N

P
P
M
   0.15
0 0
        0                                                         . .   ."                                       0
         o                                                      :••,-.•  '                                  oo oo o
           o            ••   ...                  Vi  ••              ' .                        •»     o'oo oo
            o             .               ..-'.'          ;;  '•'                         oo oo  o
              oo           .;      ••-••••.                   '   .  o oo o ooo oo  oo
                 0                                             •• 0 00 0
                  0                                       0 00 0
                     00        »                     00 0 0
                        00                 0 000 00                *
                           00         00 00
                              00 00
    0.00*--
       0.0

                                      •I-
                                                     •I-
                                                                                   •I-
                                                                                                  •I-
                     50.00
                         100.00         150.00


                         . SAMPLE DECK .  SPECIES: S02
                                                                  200.00
                                                              TIME (MINUTES)
          250.00         300.00


CONCENTRATION SCALE FACTOR: 10+0
                                                                                                               350.00
                                                                                                                                           CO
                                                                                                                                           IX)

-------
                                                                           283
                              REFERENCES
 Bergstrom,  R. W. and R. Viskanta (1973), "Modeling of the Effects of Gaseous
     and  Particulate Pollutants in the Urban Atmosphere.   Part I.  Thermal
     Structure," J. Applied Meteor.. Vol. 12, p.  901.

 Blackadar,  A. K. (1962), "The Vertical Distribution of Wind and Turbulent
     Exchange in a Neutral Atmosphere," J.  Geo.  Res. , Vol. 67, p. 3095.

 Boris, J. P. and D. L. Book (1973), "Flux Corrected Transport.  1.  SHASTA,
     An Algorithm That Works," J. of Comp.  Phys.,  Vol.  11, pp. 38-69.

 Bufalini, M. (1971), Environ.  Sci.  Techno!., Vol.  5,  p.  685.

 Calvert,  J. G.  (1975), to appear in Environ. Letters.

 Castleman,  A. W. , Jr., R. E. Davis, H. R. Munkelwitz, I.  N.  Tang, W.  P.  Wood
     (1974), "Symposium on Chemical Kinetics Data  for the Lower and Upper
     Atmosphere,," Warrenton, Virginia (September  1974).

 Cheng, R. T., M. Corn, and J.  0. Frohliger T971), Atmos. Environ. , Vol.  5,  p.  987.

 Cox, R. A.  (1974), to appear in J.  Photochemistry.

 Cox, R. A., and S. A. Penkett (1972), J. Chem.  Soc.,  Faraday Transactions I,
     Part 9, Vol. 68, p.  1735.

 Crowley,  W. P.  (1968), "Numerical Advection Experiments," Monthly Weather
     Review, Vol. 96, pp. 1-11.

 Davenport,  A. G. (1965),  "The  Relationship of Wind Structures to Wind  Loading,"
     Proc.  Conf. on Wind Effects on Eknldings and  Structures, H.M.S.O.,  London.

 Davis, D. D., W. Payne, and L. Stief (1973), J. Amer. Chem.  Soc., Vol.  95,  p.  7614.

 Deacon, E.  L-.  (1949), "Vertical Diffusion in the  Lowest  Layers of the  Atmosphere,"
     Quart. J. Roy. Meteorol.  Soc. , Vol. 75, p. 323.

 DeMarrais,  G.  A. (1959),  "Wind Speed Profiles at  Brookhaven  National  Laboratory,"
     J. Applied Meteorol., Vol. 16, p. 181.

 Egan, B.  A.  and J.  R. Mahoney  (1972), "Numerical Modeling of Advection and
     Diffusion of Urban Area Source Pollutants," J.  of Applied Meteorol.,
     Vol. 11,  pp 312-322.

	(1971), "Application of a Numerical  Air Pollution Transport Model
     to DTspension in the Atmospheric Boundary Layer," J. of Applied Meteorol.,
     Vol. 11,  PP- 1023-1039.

-------
                                                                       284

Eschcnroeder, A.  Q., J.  R.  Martinez,  and R.  A.  Nordsieck  (1972),  "Evaluation
     of a Diffusion Model  for Photochemical  Smog  Simulation."   Report  CR-1-273,
     General Research Corporation,  Santa Barbara, California.

Fankhauser, 0. C. (1974),  "The Derivation of Consistent Fields  of Wind and
     Geopotential Height from Mesoscale Rawindsonde  Data,"  J. Applied  Meteorol.
     Vol. 13, p.  637.	

Foster, P. M. (1969), Atmos.  Environ.,  Vol.  3,  p. 157.

Fromm, J. E. (1969), "Practical  Investigation of  Corrective Difference
     Approximations of Reduced Dispersion,"  Physics  of  Fluids,  Supplement II,
     pp. II-3 - 11-12.

Gartrell, F. E.  et al.  (1963), J.  Ann.. Ind.  Hyg.  Assoc. ,  Vol. 24,  p. 113.

Gear, C.  W. (1971), "The Automatic  Integration  of Ordinary  Differential
     Equations,"  Comm.  ACM, Vol.  14,  p.  176.

Gerhard,  E. G. and H. F. Johnstone  (1955),  Ind. Eng.  Chem., Vol.  47, p. 972.

Harlow, F. H. and J. E.  Welch (1965), "Numerical  Calculation of Time Dependent
     Viscous Incompressible Flow of Fluids 'with Free Surface,"  Physics of
     Fluids, Vol. 3, pp. 2182-2189.

Hecht, T. A., J.  H. Seinfeld, and  M.  C.  Dodge (1974), "Further  Development  of
     a Generalized Mechanism for Photochemical  Smog," Environ.  Sci.Techno!.,
     Vol. 8, p.  327.

Hecht, T. A., P.  M. Roth,  and J.  H.  Seinfeld (1973),  "Mathematical  Simulation
     of Photochemical Reactions:   Model  Development,  Validation,  and Applica-
     tions," Report R73-28, Systems  Applications, Incorporated, San Rafael,
     California.

Hecht, T. A. (1972), "Further Validation of  a Generalized Mechanism Suitable
     for  Describing Atmospheric  Photochemical  Reactions," Report  72-SAI-26,
     Systems Applications,  Incorporated, San Rafael,  California.

           (1972), "Further Validation  of a  Generalized Mechanism Suitable  for
     Describing  Atmospheric  Photochemical  Reactions,".Appendix  B>  of  "Further
     Development and Validation  of a  Simulation  Model  for  Estimating  Ground
     Level  Concentrations  of Photochemical  Pollutants,"  Report  72-SAI-26,
     Systems Applications,  Incorporated, San  Rafael,  California.

Hecht, T.  A. and J.  H.  Seinfeld  (1972), Environ.  Sci.  Techno!.,  Vol.  6,  p. 47.

Hino, M.  (1968),"Computer  Experiment  on Smoke Diffusion  over  Complicated
     Topography,"  Atm.  Environ.,  Vol.  2, p. 541.

Hosier,  C.  R.  (1969),  "Vertical  Diffusivity from Radon Profiles,"  J.  Geophys.
     Res.,  Vol.  74,  p.  7018.

-------
                                                                         285

 Hotchkiss, R. S. and 0. S. Hirt (1972), "Particulate Transport in Highly
     Disturbed Three-Dimensional Flow FieVds," LA-DC-72-364, Los Alamos
     Laboratories, Los Alamos, New Mexico.

 Jerskey, T. N. et al.  (1975), "Analysis of the Data from the Three-Dimensional
     Gradient Study -  Final Report," Systems Applications, Incorporated, San
     Rafael, California.

 Jones, P. M., M.A.B. de Larrinaga and C. B. Wilson (1971), "The Urban Wind
     Velocity Profile," Atmos. Environ. , Vol. 5, p. 181.

 Johnstone, H. F., and  D. R. Coughanowr  (1958), Ind. Eng. Chem., Vol.  50, p. 1169.

 Johnstone, H. F., and A, J. Moll (1960), Ind. Eng. Chem., Vol. 52, p.  861.

 Jung, C. E., and T. G. Ryan (1958), Quart. J. Roy. Meteor. Soc. , Vol. 84,  p. 46.

 Keldysh, M. V. (1964), "On B. G. Galerkin's Method for the Solution of Boundary
     Value Problems,"  NASA Technical Translation, NASA TT F-195 (March 1964).

 Laikhtman, D. L. (1944), "Profile of Wind and Interchange in the Layer of  the
     Atmosphere Near the Ground," Bull. Acad. Sci. , U.S.S.R. , Geo.  and Geophys. ,
     Ser. 8, No. 1.

 Lissaman, P.B.S. (1973), "A Simple Unsteady Concentration Model  Explicitly
     Incorporating Ground Roughness and Hea~ Flux," paper presented at the
     66th Annual Meeting of the Air Pollution Control  Association,  June 24-
     28, Chicago, Illinois.

 Liu, M. !(., P. Mundkur, and M. A.  Yocke (1974),  "Assessment of the  Feasibility
     of Modeling Wind  Fields Relevant to the Spread of Brush Fires,"  Report
     R74-15, Systems Applications, Incorporated, San Rafael, California.

 Lloyd, A. C.  (1974), "Evaluated and Estimated Kinetic  Data for Gas  Phase
     Reactions of the Hydroperoxyl Radical," Int. J.  Chem. Kinetics,  Vol.  6, p. 169

.Mahoney, J. R. , and B. A.  Egan (1970),  "A Me'soscate Numerical Model  of Atmos-
     pheric Transport Phenomena in Urban Areas," paper presented at the Second
     International  Air Pollution Conference of the International  Union of  Air
     Pollution Prevention Association,  Raleigh,  North  Carolina.

 Matteson, M.  J., W. Stober, and H. Luther (1969), I &  E. C.  Fundamentals,
     Vol. 8,  p.  677.

 McMichael,  C.  L.,  and G.  W. Thomas (1973), "Reservoir  Simulation by Galerkin's
     Method,"  J.  of SPE,  pp.  123-138 (June 1973).

 Mikusinski, J. (1959),  Operational  Calculus (Pergamon Press, Warsaw, -Poland).

 O'Brien,  J.  J.  (1970), "Alternative Solutions to the Classical Vertical Velocity
     Problem," J.  Applied Meteor., Vol.  9, p. 197.

O'Neal,  H.  E., and  C.  Blumstein (1973),  Inter.  J. Chem.  Kinetics, Vol. V,  p. 397.

-------
                                                                         286

Pandolfo, J. p., M. A. Atwater, and G.  E.  Anderson (1971), "Prediction by
     Numerical Models of Transport and  Diffusion in an Urban Boundary Layer,"
     Final Report.  The Center for the  Environment and Han, Incorporated,
     Hartford, Connecticut.

Pindcr, G. F., and W. G.  Gray (1974), "The Finite Element Method in Surface
     and Subsurface Hydrology," Notes for a summer course on hydrology,
     Princeton University, Princeton, New Jersey.

Plate, E. J. (1971), "Aerodynamic Characteristics of Atmospheric Boundary
     Layers," U. S. Atomic Energy Commission, Oak Ridge,  Tennessee.

Price, H. S., R. S. Varga, and J. E.  Warren (1966), "Applications of Oscilla-
     tion Matrices to Diffusion Convection Equations," J.  Math-Phys. , Vol.  45,
     pp. 301-311.

Ragland, K.  W. (1973), "Multiple Box  Model for Dispersion of Air Pollutants
     from Area Sources,"  Atm. Environ. ,  Vol.  7, p.  1017.

Reynolds, S. D., M. K. Liu, T.  A. Hecht, P. M. Roth, and  J.  H.  Seinfeld (1973),
     "Further Development and Evaluation of a Simulation  Model  for Estimating
     Ground  Level  Concentrations of Photochemical Pollutants."  Final  Report
     and four appendices.   Systems Applications, Incorporated,  San Rafael,
     California.

Reynolds, S. D., P. M. Roth, and J.  H. Seinfeld (1973),  "Mathematical  Modeling
     of Photochemical Air Pollution - I.  Formulation of  the Model,"  Atmos.
     Environ., Vol. 7., pp. 1033-1061.

Reynolds, S. D.  (1973), "Numerical Integration of the Continuity Equations.
     Appendix D of Further Development and Validation of  a Simulation Model
     for Estimating Ground Level  Concentrations of Photochemical  Pollutants,"
     Report  R73-T7, Systems Applications,  Incorporated, San  Rafael, California.

Roberts, P.J.W., P. M. Roth, and C. L. Nelson (1971), "Contaminant Emissions
     in the  Los  Angeles Basin--Their  Sources, Rates, and  Distribution.   Appendix
     A of "Development of a Simulation Model  for Estimating  Ground Level  Con-
     centrations of Photochemical  Pollutants." Report 71-SAI-6,  Systems Appli-
     cations, Incorporated, San Rafael,  California.

Roth, P. M., S.  D.  Reynolds, and P.J.W.  Roberts (1971), "The Treatment of
     Meteorological Variables.   Appendix C of Development  of a  Simulation
     Model for Estimating Ground Level Concentrations of  Photochemical  Pollu-
     tants."  Report 71-SAI-17,  Systems  Applications, Incorporated, San
     Rafael, Cal i form a.

Seinfeld, J.  H.  (1973),  "Existing Needs  in the Experimental  and  Observational
     Study of Atmospheric  Chemical  Reactions," Report EPA-R4-73-031 ,  Systems
     Applications,  Incorporated,  San  Rafael,  California.

Seinfeld, J.  H., T.  A.  Hecht, and P.  M.  Roth  (1971),  "A Kinetic  Mechanism
     for Photochemical  Reactions,"  Report  71-SAI-9,  Systems  Applications,
     Incorporated,  San Rafael,  California.

-------
                                                                          287

Seinfeld, J. H., T. A. Mecht, and P.  M.  Roth (1971), "A Kinetic Mechanism
     for Atmospheric Photochemical  Reactions," Appendix B of "Development
     of a Simulation Model  for Estimating Ground Level  Concentration of
     Photochemical Pollutants,"  Report 71-SAI-9, Systems Applications,
     Incorporated, San Rafael, California.

Shellard, H. C. (1965), "The Estimation of Design Wind  Speeds," Proc.  Conf.
     on Wind Effects on Building and  Structures," H.M.S.O., London.

Shir, C. C., and L. J. Shieh (1973),  "A Generalized Urban Air Pollution
     Model and Its Application to the Study of SOo Distributions in  the
     St. Louis Metropolitan Area,"  Report No.  RJ1227 (#19588),  IBM Research
     Laboratory, San Jose,  California.

Sklarew, R.  C., A. J. Fabrick, and  J. E.  Prager (1971), "A Particle-In-Cell
     Method for Numerical  Solution  of the Atmospheric Diffusion Equation and
     Applications to Air Pollution  Problems,"  Final Report No.  3SR-844, EPA
     Contract 68-02-0006,  Systems,  Science, and Software, La Jolla,  California,

-Smith, J. P., and P. Urone  (1974),  Environ. Sci. Techno!. , Vol. 8, p.  742.


Trijonis, J. C., and K.  W.  Arledge  (1975),  "Utility of Reactivity Criteria in
     Organic Emission Control  Strategies  for Los Angeles," TRW Environmental
     Services, Redondo Beach,  California.

Urone, P., and W. H. Schroeder (1969), Enyiron. Sci. Te c h n o1. Vol. 3,  p. 436.

Whitney, D.  C. (1974), "The Automatic Construction of Chemical  Kinetics Mech-
     anisms and Solution of Their Governing Equations," Report  R73-51 , Systems
     Applications, Incorporated, San  Rafael, California.

Wilson, W. E., Jr., and A.  Levy (1969), "A Study on Sulfur Dioxide-in  Photo-
     chemical Smoq.  II.,"  Project S-ll, Battelle Memorial Institute,  Columbus
     Ohio.

Wilson, W. E., Jr., A. Levy, and D. B. Wimmer  (1972), J. Air PoTiut. Contr.
     Assn., Vol. 22, p. 27.

Wu, S. S. (1965), "A Study of Heat  Transfer Coefficients in the Lowest 400
     Meters of the Atmosphere,"  J.  Geoph.ys. Res. , Vol.  70, p. 1801.

Yanenko, N.  N. (1969), Method of Fractional Steps (Springer-Verlag,  New York).

Zienkiewicz, 0. C. (1971),  The Finite Element  Method in Engineering  Science
     (McGraw-Hill, London,  England)"

-------
                                                                                 288
                                   TECHNICAL REPORT DATA
                            (Please read fatiniciions on tlic reverse before completing)
 1. REPORT NO.         ~~~
   EPA-600/4-76-016b
2.
 4. TITLE  CONTINUED RESEARCH  IN  MESOSCALE AIR POLLUTION
        SIMULATION -MODELING.   VOLUME II.  Refinements
        in the Treatment of Chemistry,  Meteorology,
  	and Numerical Integration Procedures
                             6. PERFORMING ORGANIZATION CODE
                             3. RECIPIENT'S ACCESSION-NO.
                             5. REPORT DATE
                                May 1976
 '. AUTHOR(S)
  S.  D. REYNOLDS, J. P.  MEYER,  T.  A. HECHT,
  D.  C. WHITNEY, J. AMES,  AND M.  A. YOCKE
                             8. PERFORMING ORGANIZATION REPORT NO.

                               EF75-69
 9. PERFORMING ORG ^.NIZATION NAME AND ADDRESS
  SYSTEMS APPLICATION
  950 NORTHGATE DRIVE
  SAN RAFAEL, CALIFORNIA   94903
                              10. PROGRAM ELEMENT NO.

                               1AA009
                              11. CONTRACT/GRANT NO.

                               68-02-1237
 12. SPONSORING AGENCY NAME AND ADDRESS

  ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
  OFFICE OF RESEARCH AND DEVELOPMENT
  U.  S.  ENVIRONMENTAL  PROTECTION AGENCY
  .RESEARCH ..TRIANGLE PARK,  N.C.  27711	
                              13. TYPE OF RE PORT AND PERIOD COVERED
                               FINAL REPORT 6/74-6/75
                              14. SPONSORING AGENCY CODE
                               EPA-ORD
 15. SUPPLEMENTARY NOTES
 16. ABSTRACT
        This report describes  the refinement of a .mesoscale  photochemical air quality
  simulation model through  studies of selected chemical and  meteorological phenomena
  that contribute to air pollution.   The chemistry activities  focused on the design of
  an automatic computer program for evaluating kinetic mechanisms,  the improvement of
  a  photochemical mechanism for incorporation in mesoscale models,  and the development
  of a chemical mechanism for  describing S02 oxidation.  The meteorology studies ex-
  amined the sensitivity of the model to the inclusion of wind shear,  algorithms for
  deriving mass-consistent  wind fields, -and the treatment of turbulent diffusivities
  and elevated inversion layers.   Alternative numerical techniques  for solving the
  advection/diffusion equation in grid models are evaluated, including various finite
  difference, particle-in-cell, and finite element methods,  in an attempt to find a
  suitable methodology for  accurately calculating the horizontal  transport of pollu-
  tants.  Finally, the report  considers the problem of multiday model usage and pre-
  sents results from a two-day CO simulation for the Los Angeles  basin.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
                                              b.IDENTIFIERS/OPEN ENDED TERMS
                                           c.  cos AT I  Field/Group
  *Air Pollution
  *Photochemical Reactions
  *Reaction Kinetics
  *Numerical Analysis
  ^Mathematical Models
  ^Meteorological Data
                                             13B
                                             07E
                                             07D
                                             12A
                                             14B
                                             04B
 8. DISTRIBUTION STATEMENT

  RELEASE TO PUBLIC
                19. SECURITY CLASS (Tills Report)
                  UNCLASSIFIED
1. NO. OF PAGES
 287
                20. SECURITY CLASS (This page)

                  UNCLASSIFIED	
                                           22. PRICE
EPA Form 2220-1 (3-73)

-------