EPA-R4-73-012a
October 1972
Environmental Monitoring Series
-------
EPA-R4-73-012a
EVALUATION
OF A
DIFFUSION MODEL
FOR
PHOTOCHEMICAL SMOG SIMULATION
FINAL REPORT
by
A.Q. Eschenroeder, J.R. Martinez, andR.A. Nordsieck
General Research Corporation
P.O. Box 3587
Santa Barbara, California 93105
Contract No. 68-02-0336
Program Element No. 1A1009
EPA' Project Officer: Ralph C . Sklarew
Meteorology Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C- 20460
October 1972
-------
This report has been reviewed by the Envirpnmental Protection Agency
'approved for publication. Approval does not signify that the cpntents
necessarily reflect the views and policies, pf the Agency, npr dpe§
mention of trade names, or cpmmereial products constitute endprsenient
or recommendation for use.
ii
-------
CONTENTS
SECTION PAGE
OVERVIEW AND SUMMARY OF THE WORK 1
I STRUCTURE OF THE REPORT 9
2 ANALYSIS OF CHEMICAL KINETICS IN SMOG CHAMBER
EXPERIMENTS 11
2.1 Introduction 11
2.2 Chemical Mechanism 11
2.3 Methodology of Adjustments and Evaluations 17
2.4 Experimental Data Base 18
2.5 Results of Smog Chamber Validations 20
2.6 Adaptation of Chemical Model to Atmospheric
Modeling 43
3 MODEL METHODOLOGY IMPROVEMENTS 51
3.1 Perspectives on Model Updating 51
3.2 Advective Air Trajectories 51
3.3 Vertical and Horizontal Eddy Diffusion
Coefficients 58
3.4 Emission Flux Histories 65
4 TRANSVERSE DIFFUSION AND ITS EFFECT ON THE GRC
MODEL 66
4.1 Lateral Diffusion Between Neighboring
Streamtubes 66
4.2 Lateral Diffusion Effects in the Vicinity of
High-Flux Elevated Point Sources 72
5 VALIDATION STUDIES 79
5.1 Selection of Days for Model Tests 79
5.2 Program Conversion 82
iii
-------
CONTENTS (Cont.)
SECTION
5.3 Atmospheric Validation Tests
5.4 Techniques for Model Operation
5.5 Sources of Uncertainty Due to Solar Radiation
and Particulate Reactions
6 CONCLUDING REMARKS
APPENDIX A A VIEW OF FUTURE PROBLEMS IN AIR POLLUTION MODELING
REFERENCES
PAGE
84
148
152
162
167
207
iv
-------
ILLUSTRATIONS
NO. .PAGE
2.1 Experiment 329, Propylene/NO . Plot of Propylene, NO, NO 23
X Ł*
2.2 Experiment 329. Propylene/NO . Plot of Ozone and PAN 24
X
2.3 Experiment 253, Toluene/n-Butane/NO . Plot of NO and NO. 27
"X, &•
2.4 Experiment 253, Toluene/n-Butane/NO . Plot of Ozone and PAN 28
2.5 Experiment 253, Toluene/n-Bu€ane/NO . Plot of Toluene and
n-Butane x 29
2.6 Experiment 251, Toluene/n-Butane/NO . Plot of NO and NO 30
X Ł•
2.7 Experiment 271, Toluene/NO . Plot of Toluene, NO, and N0? 32
X ^
2.8 Experiment 271, Toluene/NO . Plot of Ozone and PAN 33
X
2.9 Experiment 231, Dilute Auto Exhaust (Controlled Vehicle).
Plot of NO and N02 38
2.10 Experiment 231, Dilute Auto Exhaust (Vehicle with Emission
Controls). Simulation of Reactive Hydrocarbon Decay 39
2.11 Experiment 222, Dilute Auto Exhaust (Uncontrolled Vehicle).
Plot of NO and N02 41
2.12 Experiment 222, Dilute Auto Exhaust (Uncontrolled Vehicle).
Plot of Ozone and PAN 42
2.13 Experiment 336. Dilution Model Compared with Measured
Ethane Concentration 47
2.14 Experiment 336. Effect of Dilution on Propylene Concentration.
(Curves Computed using a Single set of Rate Constants) 48
3.1 Air Trajectory in the Los Angeles Basin 53
3.2 Comparison of Ground Trajectory with Tetroon Trajectory 54
3.3 Comparison of Wind Speed Measurements at El Monte 56
-------
ILLUSTRATIONS (Cont.)
NO. PAGE
3.4 Comparison of Wind Direction Measurements at El Monte 57
3.5 Vertical Diffusivity Versus Wind Speed 60
3.6 Vertical Diffusivity Versus Vertical Temperature Gradient 61
3.7 Vertical Diffusivity Profiles 63
3.8 Lateral Separation of Simultaneously Released Tetroon Pairs
as a Function of Time Since Release 65
4.1 CO Flux Histories on Two Neighboring Trajectories 67
4,2 Eddy Diffusivity Profile for Neighboring Streamtube Analysis 67
4.3 CO Concentration Histories on Right-Hand (Commerce)
Trajectory 68
4.4 CO Concentration Histories on Left-Hand (Synthetic)
Trajectory 68
4.5 Superposed CO Flux Histories for 24 Trajectories 69
4.6 Bounding CO Flux Histories for Parallel Trajectory Analysis 70
4.7 CO Concentration Histories for Worst-Case Parallel Trajectories 70
4.8 Superposed NO Flux Histories for 24 Trajectories 72
4.9 Eddy Diffusivity Profile for Elevated Point Source/Lateral
Diffusion Analysis 73
4.10 Ground Concentration Effects of Elevated Point Sources:
10 kg/hr/km2 Background Flux 75
4.11 Ground Concentration Effects of Elevated Point Sources:
15 kg/hr/km2 Background Flux 76
4.12 Ground Concentration Effects of Elevated Point Sources:
20 kg/hr/km2 Background Flux 77
5.1 Air Quality and Meteorological Monitoring Network in the
Los Angeles Basin 86
VI
-------
ILLUSTRATIONS (Cont.)
NQ. . . . . _ ., _ .... . PAGE
5.2 Interpolated and Observed Oz&rie Concentration at a Monitor-
ing Station in the West San Gabriel Valley on September 29,
1969 93
5.3 Inteirpdiated and observed Ozorife Concentration at a Monitor-
ing Station in the West San Gabriel Valley on November 4,
1969 94
5.4 Observed Versus Commuted Carbon Monoxide Concentration 97
5.5 Observed Versus Computed Ozone Concentration 98
5.6 September 115 1969 Trajectory Starting at Commerce at 0530 100
5.7 Trajectory No. 1—Computed and Observed CO Concentrations 101
5.8 Trajectory No. 1—Computed and Observed NO and N02
Concentrations 101
5.9 Trajectory Nd. I—Computed and Observed Ozone Concentrations 101
5.10 September 11s 1969 Trajectdry Starting at Commerce at 0636 102
5.11 Trajectory No. 2—Computed and Observed CO Concentrations 103
5.12 Trajectory No. 2--Computed arid Observed NO and NO-
Concentrations 103
5.13 Trajectory No. 2--Cdmputed and Observed Ozone Concentratidns 103
5; 14 September 11, 1969 Trajectory Starting in Oowntdwri Los
Angeles at 0530 104
5.15 Trajectory Nd. 3—Computed arid Observed CO Concentrations 105
5;16 Trajectory No. 3--Cdmputed arid Observed NO and N0?
Concentrations
5.17 Trajectory No. 3—•Cdniputed arid Observed Ozone Cdriceritratidns 105
5.18 September 11, 1969 Trajectory Starting in tiovritowh Los
Angeles at 0630 l06
5il9 Trajeetdry Nd; 4—Gomputed arid Obsgifved CO Cdnceritratioii 107
vii
-------
ILLUSTRATIONS (Cont.)
NO.
5.20 Trajectory No. 4—Computed and Observed NO and N02
Concentrations 107
5.21 Trajectory No. 4—Computed and Observed Ozone Concentrations 107
5.22 September 29, 1969 Trajectory Starting at Commerce at 0530 108
5.23 Trajectory No. 5—Computed and Observed CO Concentrations 109
5.24 Trajectory No. 5—Computed and Observed NO and NO-
Concentrations 109
5.25 Trajectory No. 5—Computed and Observed Ozone Concentrations 109
5.26 September 29, 1969 Trajectory Starting at Commerce at 0630 110<
5.27 Trajectory No. 6—Computed and Observed CO Concentrations 111
5.28 Trajectory No. 6—Computed and Observed NO and NO-
Concentrations 111
5.29 Trajectory No. 6—Computed and Observed Ozone Concentrations 111
5..30 September 29, 1969 Trajectory Starting in Downtown Los
Angeles at 0530 112
5.31 Trajectory No. 7—Computed and Observed CO Concentrations 113
5.32 Trajectory No. 7—Computed and Observed NO and NO
Concentrations 113
5.33 Trajecotry No. 7—Computed and Observed Ozone Concentrations 113
5.34 September 29, 1969 Trajectory Starting Near the Coast at
0230 114
5.35 Trajectory No. 8—Computed and Observed CO Concentrations 115
5.36 Trajectory No. 8—Computed and Observed NO and NO
Concentrations 115
5.37 Trajectory No.' 8—Computed and Observed Ozone Concentrations 115
5.38 September 30, 1969 Trajectory Starting at Commerce at 0530 116
viii
-------
ILLUSTRATIONS (Cont.)
NO. PAGE
5.39 Trajectory No. 9—Computed and Observed CO Concentrations 117
5.40 Trajectory No. 9—Computed and Observed NO and N02
Concentrations 117
5.41 Trajectory No. 9—Computed and Observed Ozone Concentrations 117
5.42 September 30, 1969 Trajectory Starting at Commerce at 0630 118
5.43 Trajectory No. 10—Computed and Observed CO Concentrations 119
5.44 Trajectory No. 10—Computed1 and Observed NO and NO-
Concentrations 119
5.45 Trajectory No. 10—Computed and Observed Ozone Concentrations 119
5.46 September 30, 1969 Trajectory Starting in Downtown Los
Angeles at 0430 120
5.47 Trajectory No. 11-Computed and Observed CO Concentrations 121'
5.48 Trajectory No. 11—Computed and Observed NO and N02
Concentrations 121
5.49 Trajectory No. 11—Computed and Observed Ozone Concentrations 121
5.50 September 30, 1969 Trajectory Starting in Downtown Los
Angeles at 0530 122
5.51 Trajectory No. 12—Computed and Observed CO Concentrations 123
5.52 Trajectory No. 12—Computed and Observed NO and N0?
Concentrations 123
5.53 Trajectory No. 12—Computed and Observed Ozone Concentrations 123
5.54 October 29, 1969 Trajectory Starting in"Downtown Los Angeles
at 0530 124
5.55 Trajectory No. 13—Computed and Observed CO Concentrations 125
5.56 Trajectory No. 13—Computed and Observed NO and N0?
Concentrations 125
ix
-------
ILLUSTRATIONS (Cont.)
N0._ _ ....-- PAGE
5,57 Trajectory No, 13—Computed and Observed Ozone Concentrations 125
5.58 October 29, 1969 Trajectory Starting at 0630 in Downtown
Los Angeles 126
5.59 Trajectory No, 14—Computed and Observed CO Concentrations 127
5.60 Trajectory No. 14—Computed and Observed NO arid N02
Concentrations 127
5,61 Trajectory No. 14—Computed and Observed Ozone Concentrations 127
5.62 October 29, 1969 Trajectory Starting at Commerce at 0630 128
5,63 Trajeettiry No, 15—Computed and Observed CO Concentrations' 129
5.64 Trajectory No. 15—Computed and Observed NO and NO,,
Concentrations 129
5.65 Trajectory No. 15—Computed and Observed Ozone Concentrations 129
5.66 October 29, 1969 Trajectory Starting at El Monte at 0830 130
5.67 Trajectory No. 16—Computed and Observed CO Concentrations 131
5.68 Trajectory No. 16—Computed and Observed NO and NO- Concen-
trations 131
5,69 Trajectory No. 16—Computed and Observed Ozone Concentrations 13l
5*70 October 30, 1969 Trajectory Starting at Pasadena at 0530 132
"5*71 Trajectory No, 17—Computed arid Observed CO Concentrations 133
5.72 Trajectory No, 17—Computed arid Observed NO and ML
Concentrations 133
5,73 Trajectory No. 17—Computed and Observed Oztifle Concentrations 133
5.74 October 30i 1969 Trajectory Starting at Ccmrniefce at 0630 134
5.75 Trajectory No. 18—Comp'Uted and Observed CO Corieeritratioris 135
5,76 Trajectory No. 18—Computed and Observed NO a'nd NO'
Concentrations 135
-------
ILLUSTRATIONS (Cotit.)
N0,_ . __ PAGE
5,77 Trajectory No. IS—Computed and Observed Ozone Concentrations 135
5.78 October 30, 1969 Trajectory Starting at El Monte at 0630 136
5.79 Trajectory No* 19—Computed and Observed ClO Concentrations 137
5.80 Trajectory No. 19—Computed and Observed NO Arid NO*
Concentrations 137
5,81 Trajectory No* 19—Computed and Observed Ozone Concentrations 137
5,82 October 30, 1969 Trajectory Starting in Downtown-Los Angeles
at 0830 138
5,83 Trajectory No, 20—Computed and Observed CO Concentrations 1^9
5.8:4 Trajectory No, 20—^Computed and Observed NO and NO A
Concentrations 139
5.85 Trajectory No. 20—Computed and Observed Ozone Concentrations 139
5.86 November 4j 1969 Trajectory Starting at Commerce at 0530 140
5,87 Trajectory No, 21—Computed and Observed CO Concentrations 141
5.88 Trajectory No* 21--Computed and Observed NO and N0«
Concentrations 141
5,S9 Trajectory No. 21—Computed and Observed Ozone Concentrations 141
5.90 November' 4, 1969 Trajectory Starting at Commerce at 0630 142
5.91 Trajectory 22-Computed and Observed Co Concentrations 143
5.92- Trajectory No. 22—Computed and Observed NO and NO,
Concentrations 143
5,93 Trajectory No-; 22^-Computed and,Observed Ozone
Coneefifitfatlofis 143
5.94 N'ovember 4, 196'9 Trajectory Starting in Pasadena at 0530 144
5J.-9S Trajectory N§« 23--Cdmputed aiid Observed CO Concentration's 145
-------
ILLUSTRATIONS (Cont.)
NO.
PAGE
5.96 Trajectory No. 23—Computed and Observed N09 Concentrations 145
5.96 Trajectory No. 23—Computed and Observed NO and N02
Concentrations 1"
5.97 Trajectory No. 23—Computed and Observed Ozone
Concentrations
5.98 November 4, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 146
5.99 Trajectory No. 24—Computed and Observed CO Concentrations 147
5.100 Trajectory No. 24—Computed and Observed NO and N02
Concentrations 147
5.101 Trajectory No. 24—Computed and Observed Ozone Concentrations 147
5.102 NO,., Photolysis Rate Constant, k. , for September 11, 1969 154
5.103 N02 Photolysis Rate Constant, k , for September 29, 1969 155
5.104 N02 Photolysis Rate Constant, k_ , for September 30, 1969 156
5.105 N02 Photolysis Rate Constant, k , for October 29, 1969 157
5.106 N02 Photolysis Rate Constant, k , for October 30, 1969 158
5.107 N02 Photolysis Rate Constant, k , for November 4, 1969 159
A.I (NO + N02) - Concentration Ground Level Huntington Park 170
A.2 CO/NO Ratios for Huntington Park 1968 171
X
A.3a Chemiluminescent Measurements in New York - 1970 184
A.3b Chemiluminescent Measurements in New York - 1970 184
A.4 Quasiequilibrium Test for 1969 Ground Data at El Monte-
High NO Levels 187
A.5 Ozone Inaccuracies Needed to Explain the Departures from
Quasiequilibrium in Fig. A.4 188
xii
-------
ILLUSTRATIONS (Cont.)
NO. PAGE
A. 6 Quasiequilibrium in a Simulated.:Smog Chamber Experiment 189
A.7 LAPS Coordinate System 197
A.8 Cross Section of Depressed Six-Lane Freeway 199
A.9 Wind-Oriented Coordinate System 199
A.10 CO Concentration Profiles Normal to Roadway at Various
Wind Aspect Angles 200
A.11 Ozone and Nitric Oxide in an Air Mass Moving Over a Roadway 200
xiii
-------
xiv
-------
TABLES
PAGE
2,1
2.2
2,3
2.4
2.5
2.6
Basic Kinetic Model
Experimental Data
Rate Constants Used in Propylene Simulations
Rate Constants for Toluene/n-Butane Simulation
Rate Constants for Toluene Experiment 271
Initial Hydrocarbon Concentrations (PPM) for Auto Exhaust
16
19
21
31
35
Experiments 36
2.7 Rate Constants Used in Simulation of Experiment 231 , Dilute
Exhaust from a Vehicle with Emission Control 40
2.8 Rate Constants Used for Simulating Experiment 222 , Dilute
Exhaust from a Vehicle Without Emission Control 44
2.9 Mole-Weighted Reactivity of Atmospheric and Smog Chamber
Hydrocarbon Mixtures 49
4.1 "Worst Case" Bounding Error Fractions Due to Omission of
Lateral Diffusion in Urban Modeling Assuming Zero Initial
Concentration of Carbon Monoxide 71
4.2 "Worst Case" Bounding Error Fractions Due to Omission of
Lateral Diffusion in Urban Modeling, Assuming an Initial
Concentration of Carbpn Monoxide of 10 ppm 71
4.3 Maximum Fractional Contribution of an Elevated Point So.urce
at a Ground Location Two Miles from the Plume Centerline 78
5.1 Directory of Air Quality and Meteorological Monitoring Sta-
tions in the Los Angeles Basin 87
5.2 Trajectory Identification Table 89
5.3 Initial Concentrations Used in Atmospheric Simulations 90
xv
-------
TABLES (Cont.)
NO. PAGE
5.4 Correlation Coefficients for CO and Ozone 96
5.5 Regression Equations for CO and Ozone 99
5.6 Rate Constants Used in Atmospheric Modeling Studies 149
A.I Concentrations (ppm) and Gas Phase Rate Constant Assumed
for Comparative Analysis 178
A.2 Upper Limit of (Surface Rate/Gas Phase Rate) Ratio 180
A.3 Ford/New York Data (First 20 Minutes) 185
A.4 Air Quality Effects for 1974 Trajectory 195
xvi
-------
OVERVIEW AND SUMMARY OF THE WORK
For the past several years, General Research Corporation (GRC) has
been developing and refining a photochemical/diffusion model for the US
Environmental Protection Agency (EPA) and its predecessor agency. The
application of this model is the prediction of air quality in terms of
pollutant emission .patterns and meteorological features of a particular
airshed. Tracing from our first steps down to now, the efforts have
balanced the emphasis between fidelity in the air chemistry and realism
in the fluid dynamic transport. Unlike earlier static models based on
superposition of plumes, ours is based on time-dependent processes.
Therefore, one significant improvement has been treatment of unsteady
diffusion and another has been the finite-difference formulation to allow
for atmospheric transformation processes.
The work described in this report combines another round of model
improvements with a controlled evaluation first of chemistry alone, and
finally chemistry combined with diffusion. The evaluation has been done
in parallel with two other contractors, Pacific Environmental Services
and Systems Applications, Inc., who are pursuing similar tests on- their
models, each of which is somewhat different from ours. It should be
emphasized that these evaluation studies are carried out in parallel
with the only interaction between contractors being an exchange of
monthly progress letters and occasional informal meetings.
Before summarizing our findings in a point-by-point narrative, it
is helpful to digress here and review briefly just what the GRC photo-
chemical/diffusion model does and how. We take an initial state for an
airshed to be the spatial distributions of the concentrations of pollu-
tants of interest; e.g., the parts per million of carbon monoxide, ozone,
hydrocarbon, nitrogen dioxide, and nitric oxide. We must also specify
the boundary conditions that control how the system evolves from its
initial state. Strictly speaking, the boundary conditions are limited
to temporal and spatial emission source distributions of the various
pollutants just named. In an indirect fashion, the kinematics of airflow
-------
including both advection and diffusion are the fundamental boundary condi-
tions for our model, since we follow an air mass-center as it is guided
by the winds from place to place around the air basin. Where it goes will
influence what pollutant emissions it receives.
We also trace the upward spread of pollutants in the air mass.as
they are introduced at the ground by emission sources. The ongoing chemical
changes are simultaneously calculated. By stratifying the air vertically
in the computation, we determine pollutant concentration as a function of
height so that our output takes the form of concentration profiles in the
air mass as functions of time-from-initial-state (or, equivalently, loca-
tion in the air basin). To generate concentrations on a horizontal grid,
we need to compute many air mass trajectories and to interpolate concentra-
tions at prescribed time intervals.
Returning to the summary of the present evaluation study, let us now
examine the findings of the chemical calculations. The approach is to
formulate a functional list of reactions that describe phenomenologically
the main observable species in a laboratory smog chamber experiment. With
an eye toward atmospheric application, we work to minimize the computing
load by collapsing some of the reaction chains into a single rate-controlling
step with overall stoichiometry specified. Similarly, parallel reactions
involving analogous members of an organic species family are lumped into
a single composite step that involves a single class of generic reactants,
a composite rate constant, and a single class of generic products. Our
adopted ground rules required:
1. Initial determination of chain stoichiometry (which there-
after is held fixed),
2. Maintenance of reported rate constants within their measured
intervals (except where reasons exist to believe otherwise),
and
3. Adjustment of a minimum number of the unknown rate constants.
-------
Following this procedure, we obtained simplified mechanisms for
fourteen chemical systems (dilute hydrocarbon/nitric oxide mixtures in
air) undergoing photooxidation in smog chamber experiments. A single
prototype reaction mechanism involving twelve species and sixteen reac-
tion steps was used for all systems. Different hydrocarbons were charac-
terized by different rate constants in the hydrocarbon oxidation reactions.
The ratios of rate constants followed ratios of the hydrocarbon reactivity
reported elsewhere. This confirmed our earlier practice of scaling labor-
atory systems to uhe atmosphere according to these ratios.
In evaluating the photochemical kinetic model, we had to compensate
for certain aspects that are peculiar to the smog chamber. Surface-to-
volume ratios are much higher in the laboratory than in the atmosphere
and an appropriate reaction chain has to be included beginning with ozone
reacting with nitrogen dioxide and ending with nitric acid on the chamber
wall. Another artificial feature of the smog chamber is the dilution of
the reaction sample by removal of sizable samples for analysis and replace-
ment with "clean" air. In certain cases, this correction became so large
as to obscure the effects of some reactions; therefore, it is our recom-
mendation that all future smog chamber work utilize in situ measurements.
(The use of long-path infrared cells for many smog studies in the past
has demonstrated the feasibility of in situ measurements.)
A striking example of the need for gas-solid reactions is found in
the analysis of dilute automobile exhaust in the smog chamber results.
Despite its small surface area compared with that of the walls, the fine
suspended particulate matter in the exhaust forced the addition of an
(N09 + particles)-reaction in order to account for nitrogen loss from the
pollutant fraction in the gas phase. This was not necessary for the
synthetic hydrocarbon/NC
of particulate material.
synthetic hydrocarbon/NO mixture experiments that were essentially free
X
Moving from laboratory experiments to polluted atmospheres, we
added several improvements and carried out further cross-checks on the
-------
GRC photochemical diffusion model that was briefly outlined above. In
preparation for the eventual coding of the air trajectory calculation,
we established the logic for objective determination of advection paths
of air mass-centers. Certain intrinsic weaknesses in the data and in
the use of ground winds suggest that high degrees of refinement are unwar-
ranted. Very poor agreement was found between wind directions and wind
speeds measured at two stations set up in the same area. Likewise, very
poor agreement was noted between a computed ground track using station
data and the measured trajectory of a tetroon flying with the wind at
a few hundred meters altitude.
In the past, our vertical eddy diffusivity values were based on wind
speed, but a more detailed analysis of the data showed that vertical
temperature gradient was better than wind speed or Richardson number for
correlating measured diffusivities. Based on these findings, we adopted
five vertical profiles of eddy diffusivity, each characterized by a range
of vertical temperature gradients. Subsequent diffusion calculations for
nonreactive species required some downward adjustments of the diffusivity
values; however, they were still within the range of observational uncer-
tainty. Horizontal diffusivities were obtained from radar measurements
of tetroons as reported in the literature.
The pollution source program was updated by further automating the
generation of emission fluxes and by introducing the new numbers provided
by Systems Applications, Inc. The"bearings and speeds for one-hour tra-
jectory segments are fed into this auxiliary program. Given a geographical
and temporal starting point, the program generates a data deck that supplies
boundary conditions for the photochemical/diffusion model. Most of the
numerous revisions given to us were incorporated in the pollutant emissions
calculations.
A three-dimensional (vertical displacement, transverse displacement,
and streamwise displacement) time-dependent diffusion code was used to
—
A neutral bouyancy balloon that floats with the air along levels approxi-
mating constant ambient density.
-------
assess the importance of neglecting transverse horizontal diffusion in
the photochemical/diffusion model. The 3-D code's coordinate frame fol-
lows the air, but Gaussian spread lateral to the air motion is considered,
for each time step as well as vertical diffusion using flux/gradient rela-
tionships. Air parcels moving parallel to one another were assumed to
pass over emission fluxes differing widely from one another. For the
worst case, errors between 27% and 39% were noted for CO-increments over
a five-hour period; however, the errors in CO-concentration after five
hours would scale down to only 10% to 20% because of the addition of
initial concentrations (usually 5 to 10 ppm) for high air pollution con-
ditions. Another assessment of horizontal spread was made for an air
*
mass-center passing, at closest approach, one grid distance away from
a stack emitting oxides of nitrogen. A trajectory model would omit the
subsequent spread of the plume that, in reality, would raise ground con-
centration somewhere downwind. Calculations showed that moderate values
of stack emission and distributed ground-based source emissions give a
maximum error of 5% due to omission of the stack plume contribution.
The heart of our evaluation study is a series of tests of the simu-
lation model against real-world air quality measurements. If there is any
return on~our investment in development and refinement efforts, it must
show up as successful predictions of contaminant concentrations in the
atmosphere. Extensive groundwork in chemical and meteorological improve-
ment has been summarized above. Our test design.will now be outlined
and the results will be summarized.
Six days in the 1969 Los Angeles smog season (September-November)
were designated for the data base. In addition to having two instrumental
trailers in operation measuring detailed aerometric data, on those six
"Grid distance" refers to the cell size that specifies the spatial reso-
lution of the emission-source inputs to the model. At the present time,
this distance is 2 miles.
-------
days airborne studies were conducted yielding numerous detailed temper-
ature profiles. The profiles are needed to obtain the vertical eddy d±f-
fusivity of pollutants. Ordinarily, only one or two soundings are avail-
able each day from instrumented balloons which telemeter information back
to a station.
Three of the days are designated "hands-off" days and the other
three, "hands-on" days. The intent of the test design is to adjust para-
meters and to develop a set of operating rules for optimal model performance
based on "hands-on" data. Then the test proceeds without any further model
manipulation for the "hands-off" days to see how well the predictions are
made. Each test of the model involves taking measured initial contaminant
levels early in the morning (0230 to 0830 for our tests) and computing the
concentration histories through the morning and early afternoon hours to
exercise both the photochemical and meteorological parts of the model.
For "hands-on" operation, the diffusional parts of the model are examined
in the absence of chemistry by checking against carbon monoxide measurements.
This has the combined advantage of being a nearly-inert tracer and of being
essentially all derived from widely distributed (vehicular) sources.
Normalizing the diffusional parts of the simulation against carbon
monoxide highlights the chemical aspects in subsequent exercises of the
full model. For example, the peaking of nitric oxide levels during the
morning commute-rush reflects first, the inability of the air to dilute
the material; and later, the effects of dispersion and transformation to
nitrogen dioxide that combine to cause a decay as they overcome the weaken-
ing emission sources as cars leave the road. Without the advance tests
of diffusion, the interpretation of the nitric oxide tests would be ambi-
guous, at least, because of uncertainties in meteorological dispersion
superimposed on uncertainties in chemistry.
Despite these seemingly difficult obstacles that must be overcome,
the test results were consistently more accurate than were the exercises
-------
of previous versions of the model. The report presents the detailed data,
but the general findings are as follows:
1. The chemistry we used to model smog chamber experiments is
applicable to the atmosphere if the hydrocarbon rate con-
A
stants are appropriately chosen and the (NO,., + particulate)-
rate is decreased.
2. The diffusion in the vertical is described well by using a
set of profiles determined directly from inert gas disper-
sion data found in the literature providing that diffusivity
profiles for stable atmospheric conditions are uniformly
decreased by factors of two or three. This degree of uncer-
tainty is not unusual for vertical diffusivities.
3. The source emission inputs give results consistent with the
materials balances observed in the atmosphere with the con-
spicuous exception of nitric oxide. Its fluxes had to be
reduced by 75% (as in tests of previous versions of the
model ) in order to achieve observed loadings of NO 4- NO..
1 *•
Carbon monoxide (and previously hydrocarbon ) require no
adjustment. No physical mechanisms have been specifically
identified to explain this deficit; however, surface uptake
of NO appears to be a strong possibility that must be in-
vestigated in future field programs.
4. Curves of individual species fit the interpolated measure-
*A
ments rather well during the peaking phases; however, both
A
Automobile exhaust runs in the smog chamber were modeled by stratifying
all the reactive hydrocarbons into two groups based on speed of oxida-
tion rate. Good atmospheric results were obtained using rates derived
for the slower of the two groups of hydrocarbon.
A*
Since the air parcel trajectories generally cut between monitoring
stations rather than going right over them, we must use some inverse-
distance-weighted averages to interpolate among the station measurements.
-------
carbon monoxide and nitrogen dioxide are overpredicted at
the ends of trajectories where their levels are relatively
low. Inadequate accounting for the mixing is likely respon-
sible for these discrepancies. Furthermore, in the case of
nitrogen dioxide, heterogeneities can cause large errors
since the chemistry in the model is treated homogeneously.
5. Ozone, one of the main pollutants, is relatively well pre-
dicted in its time-phasing for net production and its level.
This success is especially fortunate because ozone is the
subject of much control planning activity and is a very
sensitive indicator of validity for photochemical models.
In an appendix to the report, we indicate two areas where future
research is needed to build confidence in applying photochemical smog
models. One deals with gas-solid interactions and the other, with the
interference of finite mixing rates with the reaction kinetics. Gas-
solid interactions include aerosol reactions and adsorption to urban
surfaces. These processes may well be responsible for the difficulty
in achieving a nitrogen balance in morning air samples. Surface uptake
of pollutants will assume growing importance in the analysis of large-
scale urban/rural air pollution. The turbulence interference phenomenon
occurs when chemistry proceeds rapidly compared with mixing. Experimental
evidence is analyzed to demonstrate that the magnitude of this effect
could lead to significant modeling errors. Both of these new problem
areas must receive more attention in field studies before the modeling
methods can be further advanced to deal with them.
-------
1 STRUCTURE OF THE REPORT
This report describes the rationale behind and the evaluation of
various advances in the existing GRC photochemical/diffusion model. The
model's objectives and methodology have been discussed in detail in the
preceding Overview and Summary; therefore, we will emphasize the near-
term goals and activities without repeating the background discussions.
Briefly, our immediate purposes are, first, to make changes that
should improve the chemistry and physics used in the GRC photochemical/
diffusion model and, second, to subject the updated model to controlled
evaluations using measured data. A subsidiary objective is to convert
the program to an IBM system and create manuals to afford US Environmental
Protection Agency (EPA) personnel the opportunity to operate the model.
Improving the model's physical and chemical content necessitates some
refinement and some innovation. Refinements in source inventories are
available from other recent work in the field.
A brief preview of the remaining sections in this document is
given below.
Section 2. Chemical kinetic improvements are introduced in three
ways: (1) updating input values to incorporate newly measured rate con-
stants, (2) adding or deleting reactions based on recent findings, and
(3) exercising the kinetics submodel for smog chamber conditions over a
wider span of systems and mixture ratios than that used previously.
Section 3. The meteorological innovations are based on better
choices of diffusion coefficients and systematized (but still manual)
wind field analysis.
Section 4. The introduction of chemical and meteorological improve-
ments is followed by controlled evaluations of modeling assumptions adopted
previously, i.e., 'the neglect of crosswind diffusion and the treatment of
-------
large point sources in the framework of a source model laid out on a two-
mile grid. Of particular concern is the omission of plumes from off-
trajectory point sources in inputs to the moving contrpl volume. These
plumes are left out if the mass center of the air parcel never actually
transects the grid square containing the source. This plume error and
lateral diffusion error from area source nonuniformities are assessed
*
using the three dimensional LAPS code developed at GRG.
Section 5, Model tests for six days were undertaken first for dif-
fusion of CO and subsequently for major species undergoing diffusion com-
bined with photochemistry. Although the objective was four six-hour tra-
jectories per day, some modifications had to be incorporated in the form
of tradeoffs between numbers and lengths of trajectories. (An auxiliary
study qf transportation control abatement strategies was performed using
three of the trajectories. The results of this task are reported in a
separate volume.)
*
Appendix A, See. A.3.2.
10
-------
2 ANALYSIS OF CHEMICAL KINETI,GS IN SMQG CHAMBER EXPERIMENTS
2.i INTRODUCTION
The mathematical model of smog photochemistry described belgw is a
lumped-paramete.r model which is a key element of a larger model of polluted
atmospheres. In developing and validating such a chemical model, the fol-
lowing objectives have guided our approach;
1. Reproduction of the essential features of smog chamber
experiments
2. Capability to simulate experiments using a variety of hydro=
carbons and hydrocarbon/NO mixtures
X
3. Ease of adaptation to atmospheric modeling
4. Retaining physical plausibility by using rate constant values
which are in agreement with experimental data
5. Maintaining computational simplicity by using lumped
parameters
6. Explicit consideration of surface reaction effects that may
be peculiar to smog chambers
In the following sections, we shall deal with the validation of the
model using smog chamber data. First, we describe the kine.tie model itself
in some detail. The methodology used for validation is reported next.
Then the smog chamber data and the results are described. Finally, we
discuss criteria used in adapting the chemical model for simulating
atmospheric chemistry.
2,2 CHEMICAL MECHANISM
2.2.1 Reaction Steps
The basic mechanism is composed of sixteen reactions. Some of the
reactions are elementary and in gome cases a set of elementary reactions
11
-------
has been reduced to a single step, hence the lumped-parameter nature of
the model. The reactions included in the model are shown below.
Following the inorganic cycle
hv + NO -> NO + 0
1 implies (2.1)
hv + N02 -> NO + 03
+ 0 + M -* 03 + M
NO
we have the hydrocarbon oxidation chain initiators
0 + HC
OH + HC + (b2)R02 ^2'4)
0 + HC + (b)R0 (2.5)
Reactions (2.3)-(2.5) are lumped reactions which represent the oxidation
9 — Ł\
chains which have been postulated to occur. HC denotes a generic
hydrocarbon and RO is an organic radical. The b's denote branching
factors which account for the fact that the oxidation chain produces a
multiplicity of radicals. In reaction (2.4) we- have treated the radical
7 R
attacking HC as OH because of its likely dominance.
^
Because the reaction 0 + 02 + M ->• 0. + M is known to be very fast,
the two reactions shown in (2.1) can be combined into a single reaction
whose net product is NO + 0 , i.e., hv + N02 ->- NO + 0 . This is
equivalent to assuming 0-atom quasistationarity.
12
-------
The conversion of NO to NO occurs via the chain-carrying
reaction
RO + NO -»• NO + (y)OH (2.6)
where (y) is a yield factor which represents that fraction of the con-
version which returns OH to the system. The yield factor is less than
one because all R's are not H .
Note that reactions (2.4) and (2.6) form essentially a closed loop
early in the reaction and stability requirements call for by < 1 + a ,
where a is a positive function of other rate constants and concentra-
tions which is small compared to 1 at early times and gradually increases
throughout the reaction. Imposition of this constraint prevents RO
runaway during the early part of the reaction.
Chain-termination steps consist of the lumped reaction
R02 + N02 •* PAN* (2.7)
and the elementary reactions
OH + NO -> HONO (2.8)
OH + NO -* HNO (2.9)
The photodissociation of HONO has been suggested as a possible
7 8
source of OH-radical '
hv + HONO -> OH + NO (2.10)
*
PAN denotes peroxyacetylnitrate.
13
-------
Formation of HONO is assisted by the reaction
NO + NO, + H?0 '-> 2HONO (2.11)
which is likely to proceed in two steps
NO + NO -> N203
NO + HO -> 2HONO
4
as suggested by Altshuller and Bufalini. The N^ reacts with 1^0
so rapidly that the two reactions 'can be lufllped into reaction (2.11).
The late-time behavior of 0 and NO is best reproduced when the fol-
lowing reaction is included
N02 + 03 + N03 + 02 (2.12)
Nitrogen imbalances in smog chambers have prompted various investi-
gators to attempt to track down the fate of the nitrogen compounds. Gay
Q
and Bufalini report that a large fraction of the nitrogen loss can be
accounted for by nitrate formation on the walls of the chamber. Follow-"
ing their suggestions as well as those of Dodge, the reactions shown
below have been included in the mechanism.
•3 9 O fc '(2.; 13)
2HN03 (2.15)
It should be noted that up to how :no :HNO has been observed in the gas
9
phase in smog chambers. this prompted us t-6 compare fch'e -relative effi-
ciency of reactions (2.9) and (2.11)-. In oaf simulations, tfee result was
14
-------
that for 50% relative humidity reaction (2.15) is about six times faster
than reaction (2i9), so oo.e would expect to find significantly more HNO
ori the walls of the chamber than in the gas phase. As a final commentj
we note that the significance of reactions (2.13)-(2.15) for atmospheric
modeling is not clear at this time,
Aerosol was observed in the chamber experiments with dilute auto
st. The disappearance of NO from the system differed signifies
from that observed in the other experiments in which no aerosol was pre-
sent. This promp
tion of the form
exhaust. The disappearance of NO from the system differed significantly
triments
sent. This prompted the suggestion by Dodge that a first-order reac-
NO + particulates -> products (2.16)
be added to the system to account for the observed effects. Inclusion of
this reaction has improved the simulation of the auto-exhaust experiments.
Also, this reaction may be helpful in atmospheric modeling since aerosol
is observed in the real world.
Previous versions of our kinetic model also contained the reaction
RO + NO -> PAN (2.17)
12
This reaction has been suggested by Hanst as likely to be important in
the formation of PAN. Such likelihood was supported'by our earlier model-
•4),
7,14,15
13
ing work. However, the increase of rate constants for reactions (2.4),
(2.7) and (2.8) which was required by recently available measurements
caused reaction (2.17) to become unimportant in our model and thus it has
been dropped.
Table 2.1 shows a list of all the reactions in our current kinetic
model along with rate constants for which measurements exist. The rate
constants obtained by repeated trials from the simulations are reported
in the section on results of the adjustments and tests.
15
-------
TABLE 2.1
BASIC KINETIC MODEL
Reaction
1.
la.
2.
3.
4.
5.
6.
7-
8.
9.
10.
11.
12.
13.
14.
15.
16.
hv + N00 -> NO + 0
i
o + q2 + M -> o3 + M
NO + 03 + N02 + 02
0 + HC -> (b1)R02
OH + HC ^ (b2)R02
0 + HC -> (b )RO
RO + NO -> NO + (y)OH
RO + NO ->• PAN
OH + NO+ HONO
OH + N00-> HNO_
2 3
hv + HONO ^ OH + NO
H.O
7
NO + NO + 2HONO
N°2 + °3 "" N°3 + °2
N03 + N02 -> N205
N205 -> N03 + N02
N2°5 + H2° "*" 2HN°3
NO + particulates ->- products
Experimental Rate
Constant Values^
2.67(-l)min~1
-2 -1
1.32(-5)ppm min
2..2(+l) to- 4.4(+l)ppm min
1.5(+3)ppm min
3.0(+3)ppm~ min~
5(-2) to 1.25(-l)ppro 1min~1
4.5(+3)ppm min
7
2.5(-3)ppm min
16
17
14
15
18
19
18
20
t
The number in parentheses denotes the power of ten by which the coef-
ficient must be multiplied, e.g., 2.67(-l) = 2.67 x 10"1 .
k
Experimental values for these rate constants are often known for parti-
cular hydrocarbons and will be reported in Sec. 2.5.
For the validation process, k^5 was converted to a pseudo-first-order
rate constant by lumping water vapor content of air at 50% relative
humidity into k]^ since the smog chamber experiments were conducted
at this level of humidity. The resulting rate constant is 60.5 min"1.
16
-------
2.2.2 Quasistationarity Assumptions
Several of the species included in our kinetic model can be assumed
to be in a quasistationary state with respect to the other species.
Apart from the computational advantages of this assumption, quasistation-
arity can be justified on physical grounds by examining the relative
rates of the various reactions involved. Such a check has revealed that
it is likely that 0-atom, RC>2 , OH , NO , and NO are in a quasi-
stationary state. 0-atom quasistationarity can be justified on the basis
that the removal of 0-atom by the reaction 0 + 0 + M -> .0 + M is known
to be very fast. We have tested quasistationarity assumptions for RO ,
OH , NO , and NO by solving parallel cases with and without station-
arity. The results of the tests showed that assuming stationarity has a
negligible effect on the computed concentration of all the species. A
similar test for HONO yielded negative results, thus HONO has been
retained as an active species.
2.3 METHODOLOGY OF ADJUSTMENTS AND EVALUATIONS
Two kinds of adjustable parameters are available to us: rate con-
stants and branching factors. However, the number of free parameters is
limited by the fact that several elementary rate constants have been mea-
sured. Our approach is to keep the values of the measured rate constants
within the range of experimental uncertainties. The unknown values of
nonelementary, i.e., lumped, rate constants are then estimated from com-
parisons with analogous reactions, if they exist., and during the simula-
tion process itself. Indeed, the object of the simulation is to deter-
mine the values of these unknown constants. In cases where rate-control-
ling processes can be identified, the rate constants are sometimes avail-
able. These are also confined to ranges of measurement wherever possible.
The branching factors of the model are determined by the NO and
;cay prior to the NO^ peak &
branching factor is estimated from
HC decay prior to the N07 peak and before the ozone buildup. The
17
-------
d[NQ]/dt fy
b2 ~ d[HC]/dt (2
In Eq. (2.18), it is implied that most of the HE decay is due tb the
reaction OH + HC ->• (b )RO . This is justified by virtue df the fact
that reaction (2.3) is slow compared to (2.4) and that early in the reac-
tion the ozone is essentially zero and thus (2.5) plays no role in HC
decay. Since NO -> NO conversion occurs via reaction (2.6), the rela-
tion shown in (2.18) can be used because R02 is in stationary state.
Thus, using the available data, we obtained linear least-squares
fits of the NO and HC to estimate their respective decay rates. Then
from Eq. (2.18), we estimated b prior to any adjustment of rate con-
stants. We then chose a value of b^ from among the various values
obtained for each experiment set. Subsequently, we set b.. = \> . The
value of b was obtained by modeling the late-time behavior of the sys-
tem. Once set, the values of the branching factors remain constant
throughout the simulation.
2.4 EXPERIMENTAL DATA BASE
Experiments on different mixtures of the following four systems in
air were used in the validation process:
1. Propylene/NO
X.
2. Toluene/n-Butane/NO
X
3. Toluene/NO
x
4. Dilute auto exhaust/NO
x
These four groups comprised a total of fourteen experiments which were
used in testing the kinetic model. The number of experiments in each
group was four, three, five, and two, respectively. Table 2.2 shows a
detailed breakdown of the experimental data used in our simulations. The
experiment numbers shown on the table are the same ones used by EPA in
their laboratory procedures.
18
-------
TABLE 2.2
EXPERIMENTAL DATA
Initial Concentrations,
ppm
Average Fractional
Dilution Rate
Group
Propylene
Toluene/
n-Butane
Toluene
Auto
Exhaust
Exp . No .
321
325
329
336
251
253
257
250
258
263
271
305
**
CUE 222 ,
***
CE 231
NO
1.23
.30
.29
1.14
1.11
.53
.27
1.17
.35
.54
.32
1.26
1.94
2.73
NO 2
.09
.04
.01
.04
.11
.11
.07
.08
.04
.06
.04
.06
.10
.23
HC
.275
.45
.24
.61
1.60/3.02
1.43/3.42
1.29/2.97
1.53
2.88
1.71
1.20
3,14
1.06/1.151"
.39/.201"
-4 -1 *
(x 10 min )
7.52
7.33
7,22
7.03
6.17
6.22
4.90
5.78
4.88
5.80
6.05
5.76
5.65
5.98
**
ft**
t
To obtain average volumetric dilution rate multiply by the chamber
volume (335 ft^) (see Sec. 2.6.1 for a detailed discussion of dilution
effects in smog chambers).
Exhaust from a vehicle without exhaust hydrocarbon and CO emission
controls.
Exhaust from a vehicle with exhaust hydrocarbon and CO emission controls,
The multiple hydrocarbon mixture has been aggregated into two classes:
high and low reactivity hydrocarbon. The fraction shown here gives
the ppm of each class in the order high-reactivity/low-reactivity.
19
-------
2.5 RESULTS OF SMOG CHAMBER VALIDATIONS
In this section, we discuss the results of the validation tests
which were performed on our chemical model. The goal of these validation
tests was to simulate smog chamber experiments performed in EPA labora-
tories. The experimental data used for comparison with the model's
results has been described in Sec. 2.4. The outcome of the tests is pre-
sented by means of graphs that are representative of the group of tests
performed. In addition, a table of the rate constants used in the vali-
dation tests is presented for several experiments. Finally, the branch-
ing factors for the various simulations are also given.
2.5.1 Propylene Experiments
The simulation of the propylene experiments was carried out by-
varying a single rate constant within the framework of a basic set of
parameters. Thus, after determining the branching factors and setting
values for the other rate constants, it was sufficient to adjust k (the
OH attack on HC) to reproduce the experimental results obtained for the
various propylene/NO mixtures. Two of the experiments (nos. 321 and 329)
X4 -1 -1
required k, = 6 x 10 ppm min . Experiments number 32i> and 336
4 4
required k = 2 x 10 and k, = 3 x 10 , respectively. Table 2.3
shows the rate constants used in the four cases. The branching factors
were determined to be b^ = b2 = 4 and b = 1 . The yield factor for
OH in reaction (2.6) is set to 0.25. It should be noted that Stedman7
measured the rate constant of the hydroxyl-propylene reaction and
obtained k = 2.5 x 10 ppm min . (If this experimental value of k,
H 4
were used for all the simulations, the reaction would be slowed down in
experiments 321, 329, and 336, and it would be speeded up in 325.) Thus,
our model values for k4 agree well with the measured quantity, which
is a hundred times greater than a previous estimate of Westberg^ of
k4 = 244 ppm min . In our previous modeling work, we had lowered
this estimate to 80 ppm min . To preserve the good agreement between
data and simulation, we had to raise kg from 1500 pprn'^in"1 to 105 ,
and k? from 6 to 600. (It will be recalled that k was increased
20
-------
TABLE 2.3
RATE CONSTANTS USED IN PROPYLENE SIMULATIONS
Reaction Number
1
la
2
3
4 (experiment 321 and 329)
4:(experiment 325)
4 (experiment 336)
5
6
7
8
9
10
11
12
13
14
15
16
Rate Constant
2.67(-l)min~1
1.32(-5)ppm~2min~1
2.67(4-1)
6.1(+3)**
6.0(4-4)
2.0(4-4)
3.0(4-4)
9.27(-3)**
1.0(4-5)
6.0(4-2)
3.0(4-3)
1.0(-3)min
5.0(-3)
4.5(4-3)
-1
6.05(+l)min
0 min'1
Units in ppm~ min unless otherwise specified.
*
Measured rate constant from Ref. 2.
21
-------
from 10 to 1500 and k from 30 to 3000 in order to be consistent with
1 / "1 ^
experimental values. ' ) The success of this procedure is especially
significant, since it shows that a workable set of rate constants is far
from unique. However, this is not surprising in view of the highly non-
linear character of the system.
_3
Referring to Tables 2.1 and 2.3, we note that k^ = 5 * 10 in
Table 2.3 and this is a factor of 10 lower than the lower bound of the
experimental value shown on Table 2.1. This had to be done in order not
to impair greatly the late-time behavior of N02 and 03 . The same
effect was encountered in modeling the other experiments with various
types of hydrocarbons and HC/NO mixtures. In the present model, reac-
tion (2.12), NO + 0 -> NO + 0 , is rate-controlling late in the reac-
tion. Furthermore, the computed concentrations of N0_ and 0 are very
sensitive to small changes in k . Thus this parameter is very impor-
tant and the availability of a very accurate value would be a boon for
the modeler.
Figures 2.1 and 2.2 illustrate results obtained in the simulation
of experiment 329. Note that the data are closely approximated by the
model. The ozone plot shown in Fig. 2.2 exhibits lower concentrations
than are indicated by the data late in the reaction. This discrepancy
is due to dilution effects in the model and cannot be resolved by plausible
adjustments in rate constants. Omission of dilution terms from the model
results in ozone concentrations which provide a better fit to the data.
However, omitting dilution terms also results in markedly poorer fits
for the hydrocarbon and NO . Thus it seems that the inclusion of dilu-
tion both benefits and hinders the results. One possible explanation of
this apparent paradox is that the continuous-dilution approximation- used
in the model is not accurate enough to simulate those cases where the
dilution rate becomes large compared with the chemical rate. (See Sec.
2.6.1 for a detailed discussion of dilution effects in smog chambers.)
22
-------
n.oo
!n 'NO
•O HC
A NO-
EXPERIMENTAL
DATA
HC
NO
4.00
e.oo
P. 00
16.00 20.00 2H.OO
TIME :CX10 M
i t~
28-00 38.00
Figure 2.1. Experiment 329, Propylene/NO Plot of Propylene,
NO, N00
-------
s
a
a:
o
§
c_>
8
MODEL RESULTS
D OZONE
O PAN
EXPERIMENTAL j + OZONE
DATA I X PAN
0.00 H.OO 8.00 12.00
16.00 20.00 Ł4.00
TIME (X10 ']
28.00 38.00 36.00
Figure 2.2. Experiment 329, Propyjlene/N0x> Plot of Ozone and PAN
24
-------
2.5.2 Toluene/n-Butane Data
The simulations of this binary hydrocarbon mixture were very success-
ful. Most remarkable is the fact that we were able to reproduce accurately
all three experiments with a single set of parameters. This contrasts
with the propylene tests where k, had to be. varied. It should be noted
that two sets of hydrocarbon reactions were used in these simulations.
Some remarks are in order about the rate constants for reactions
(2.3)-(2.5). Published rate constant data were found for both toluene
9 fi
and n-butane only for reaction (2.3). ' For reaction (2.4), a measured
f\ O T
rate constant could be located ' only for n-butane. The rate constant
for toluene in reaction (2.4) was estimated from the literature value of
k, for propylene using the relative reactivity of the two hydrocarbons.
This initial estimate of k. was then adjusted until agreement between
computed results and chamber data was obtained.
No literature data for toluene or n-butane ozonolysis was found.
Nevertheless, the rate constants for the ozonolysis of these hydrocarbons
are expected to be low, as indicated by the low values given"'fbr aromatics
such as xylene. For example, Niki, in a scale of relative rate constants
(with propylene as unity), gives xylene ozonolysis an upper bound of 0.2
compared with relative rates ranging from 2 to 62 for the olefins. Reac-
tivity scales are not useful for obtaining estimates for k since hydro-
carbon reactivity and k do not correlate well for other than olefinic
* 6
compounds, as has been pointed out by Niki, et al. For purposes of the
simulation, we chose some initial low value for the ozonolysis rate con-
stant and subsequently adjusted it to obtain good agreement with chamber
-4 -1 -1
measurements. The resulting values remained low, 3 * 10 ppm min
for toluene and 10 ppm min for n-butane. These values should be
compared with 9.27 x 10~ ppm" min which is used for propylene.
Based on relative reactivity considerations, the OH + HC rate
constants for toluene and n-butane were maintained at a 2:1 ratio with a
25
-------
4
rate constant for toluene of 2 x 10 . Although this rate constant is
higher than the one used in the pure toluene experiments, this result is
not surprising inasmuch as we have synergistic effects to account for.
Figures 2.3-2.5 illustrate a simulation of experiment 253. We can
see that NO , both hydrocarbons, and 0 are reproduced very well
X ~*
indeed. In contrast with the pure toluene runs (see Fig. 2.8), the ozone
buildup has both the correct time phasing and magnitude. Similar results
were obtained for the other two experiments in the set. However, as
shown in Fig. 2.6, the NO buildup occurs late for experiment 251, al-
though the shape of the curve is similar to that shown in the data.
Table 2.4 shows a list of the parameters used for these cases. The
branching factors used are b = b = 2 and b = 1 . The yield factor
for OH in reaction (2.6) was set to 0.5.
2.5.3 Toluene Experiments
The results of the simulations of the toluene experiments were
fair. The main problem encountered was that too much ozone was produced
in the simulation. Moreover, it seems that no plausible adjustments qf
the rate constants remedied the situation. The reproduction of toluene
and NO histories was generally good, but the NO tended to linger
late in the reaction. Figures 2.7 and 2.8 illustrate one such simulation,
in this case for experiment 271. Note that although the NO decay is
well modeled, the simulated 0 buildup is about 40 minutes too late.
Since the interaction between NO and 0 is so strong, efforts aimed
at speeding up the 0 buildup are bound to impair the modeling of NO
and vice versa. Finally, typical OH + toluene rate constants used in
these simulations were 10 ppm min~ . This compares with the range
44 A
2 x 10 to 6 x 10 used for propylene and with 2 x l(T used in the
toluene/n-butane case.
26
-------
MODEL JO NO
RESULTS |A :N02
EXPERIMENTAL I + N0
DATA ] * NO,
:H-'QQ fl-OQ 12 -ID 16-OT ;2Q,'QO 2H-00 '28.^00 3g.:QQ
TIME aro !1i
Figmre 2.3. Experlmemt 253, Toluene/.n-Butane/NO . Plot of NO and HO,
36 .,00
-------
NO
00
EXPERIMENTAL I + OZONE
DATA I v PAN
g
T T
16.00 KJ.OO 2H.QO
TIME (X10 M
28.QO
32-00
36.00
0.00 M.OD 8.00 12-00
Figure 2.4. Experiment 253, Toluene/n-Butane/NO . Plot of Ozone and PAN
-------
MODEL In TOLUENE
RESULTS Q n-BUTANE
EXPERIMENTAL I * 'TOLUENE
DATA X n-BUTANE
Q.DO
M.QO
8.00
12.00
1 T
16-00 20-00 , 2M.OO
TIME (X1Q ')
28-00
32.00
36.00
Figure 2.5. Experiment 253, Toluene/n-Butane/NO . Plot of Toluene
and n-Butane
-------
to
O
0.00
MODEL I n NO
RESULTS j A N02
EXPERIMENTAL! x N0?
DATA
\ + NO
H.QEt §'.00 Ig'-OO 16.00 20-00 , 2M-00 28.00' ^
TIME TXIQ M
Figure 2.6. Experiment 251, Toluene/n,-Butane/NO . Plot of NO and NO
x
36.00
-------
Tolu
.ene
n-Butane
TABLE 2.4
RATE CONSTANTS FOR TOLUENE/n-BUTANE SIMULATION
Reaction Number
1
la
2
3
4
5
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Rate Constant Used in Validation
2.67(-l)min~1
1.32(-5)ppm min
2.67(+l)
1.69(+2)**
2.0(+4)
3.0(-4)
1.0(+4)
2.0(+2)
1.0(-3)min
5.0(-3)
4.5(+3)
-1
6.05(+1)min"
0 min
Units are ppm min unless otherwise specified.
Measured values obtained from Refs. 2 and 6, respectively.
31
-------
to
NJ
(D
} n
NO
MODEL I o HC
RESULTS |
( A N02
EXPERIMENTAL) " ""2
DATA
X NO,
i
HI
NO
0.00 M.OO 8.00 12.00 16.00 80.00 2M.OO
TIME (X10 ')
28.00
38.00
36.00
Figure 2.7. Experiment 271, Tolue.ne/N0 . Plot of Toluene, NO, and N0
-------
EXPERIMENTAL + OZONE
DATA
Q. 00
4.00
16.00 GO.00 2M.OO
TIME (X10 >J
SB. 00
38.00
36.00
Figure 2.8. Experiment 271, Toluene/NO . Plot of Ozone and PAN
X
-------
The rate constant data used to simulate experiment 271 are given
in Table 2.5. No additional data are shown for the other toluene experi-
ments because the results obtained do not warrant it; See Sec. 2.5.2 for
a discussion of the rate constant data for toluene reactions. The branch-
ing factors used are- b^ = b2 = 2 and b^ = 1 . The yield factor for
OH is 0.5.
2.5.4 Auto Exhaust Validations
The experiments with dilute auto exhaust introduced the additional
complication of having to deal with a multiplicity of hydrocarbons. Also>
particulate matter was observed in these experiments, but was not observed
in the previous experiments. The auto exhaust data also exhibited
NO ->- NO conversion with relatively small amounts of hydrocarbon having
reacted. To account for NO disappearance in the presence of aerosol
11 X
it was suggested that reaction (2.16) be added to the model, the rate
constant to be determined by adjustments during the simulation. Because
of the multiple hydrocarbon mixture, the branching factors were increased
to reflect the increased length of the hydrocarbon oxidation chains with
its concomitant increase in organic radicals. Finally, since it is not
feasible to consider each hydrocarbon individually, the mixture was
aggregated into three types: nonreactive and low- and high-reactivity
hydrocarbons. Hydrocarbons considered nonreactive were ignored in the
simulation. Thus the modeling runs were conducted using the two reactive
hydrocarbon groups. The initial hydrocarbon concentrations were obtained
from an analysis by Dodge and are reproduced in Table 2.6.
The initial rate constants for each group were obtained by mole-
weighted averages of the rate constants for typical members of each grbupi
Thus for group I, we used n-butane and toluene to represent the paraffins
and aromatics, respectively. For group II, ethylene and propylene were
used. Again, these rate constants were adjusted during the simulation.
34
-------
TABLE 2.5
RATE CONSTANTS FOR TOLUENE EXPERIMENT 271
-•-
Reaction Number Rate Constant Used in Validation
1 2.67(-l)min l
la 1.32(-5)ppm~2min~1
2 2.67(+l)
**
3 1.69(+2)
5 5.0(-4)
7 2.0(+2)
8 1.5(+3)
9 3.0(+3)
10
11 5.0(-2)
12 5.0(-3)
13 4.5(+3)
14
15 6
16 0 min
* — 1 — 1
Units are ppm min unless otherwise specified.
%
Measured constant obtained from Ref. 2.
35
-------
TABLE 2.6
INITIAL HYDROCARBON CONCENTRATIONS (PPM) FOR AUTO EXHAUST EXPERIMENTS
Experiment
Number
Group
I. Low-Reactivity
II. High-Reactivity
Hydrocarbon Type
C. + paraffins
Aromatics (excluding benzene)
Ethylene
Olefins (excluding ethylene)
222
0.68
0.47
0.48
0.58
231
0.11
0,09
0.22
0.17
It should be noted that rather large amounts of CO were present
in the mixtures described above. The data provided for these experiments
show that for Exp. 222, the concentration of CO was 53 ppm, and for
Exp. 231, it was 12 ppm. Dodge has pointed out that CO may be partly
responsible for the oxidation of NO and that this would explain, at
least in part, why the NO ->• N0_ conversion occurs so rapidly even
though very little hydrocarbon has reacted. The CO effect could come
about via the following steps:
OH + CO -> CO + H
(2.19)
H + 0 + M -> HO + M
HO + NO + OH + NO
H0? + NO -*• HNO + o.
(2.20)
(2.21)
(2.22)
Reaction (2.21) is analogous to reaction (2.6) and would be the step
responsible for part of the NO -> N02 conversion. Our model does not
include reactions (2.19)-(2.22), of course, and so if CO reactions are
indeed significant, we will have to compensate for them by other means
36
-------
such as increasing branching factors and k . We note, however, that
previous work with CO (Refs. 4, 5, 8, 22-24), appears to indicate that
CO concentrations of the order of 100 ppm are required before CO can
be considered to play a significant role in the oxidation of NO . Thus
CO is probably not important in experiment 231, but may be a significant
factor in experiment 222.
Figure 2.9 shows the results for NO and NO obtained for the
controlled exhaust case, experiment 231. No ozone results are shown
because this experiment produced very small amounts of ozone and the
simulation behaved accordingly. This is due to the relatively large con-
centrations of NO that exist throughout the experiment. As the graph
shows, the NO is well modeled but the NO buildup in the model is
not as fast as the data would indicate. It is puzzling, however, that
the data appear to show a rapid buildup of NO. even though NO decays
very little in the interval 0-80 min. Nevertheless, we note that the
NO achieves its correct magnitude late in the reaction. The effect of
reaction (2.16) is to take NO out of the system to simulate the NO
Z. X
decay. The NO plot shown in Fig. 2.9 illustrates that the NO removs
is well modeled. Note that the correspondence between computed NO and
the (smoothed) NO data provided by EPA (denoted by the asterisks)
X
is very close. It should be noted, however, that in this simulation 20%
of the difference between initial and final NO concentration is
x,
accounted for by dilution.
Figure 2.10 shows a plot of the computed reactive hydrocarbon for
experiment 231. The intent here is not to reproduce the hydrocarbon
data, but rather to show that relatively little hydrocarbon has reacted.
The experimental data show a final concentration of 0.3 ppm and the model
yields 0.26 ppm at 360 minutes. As was the case with NO , dilution
accounts for 20% of the concentration change from initial to final value.
Thus the main characteristics of this experiment (N0x disappearance,
hardly any ozone production, and a slow reactive-hydrocarbon decay) are
reproduced well by the model.
37
-------
to
CO
Q.QO
MODEL
RESULTS
D NO
A N0?
O N0u
EXPERIMENTAL .) x
DATA I
NO
N02
NO,
M.QO
8-00
12-00
1 - T
16.00 20-00 2M-00
TIME 1X10 ')
88.00
36.00
Figure 2.9. Experiment 231, Dilute Auto Exhaust (Controlled Vehicle).
Plot of NO and N0n
-------
0.6
Ł
IX
0.4
0.2
FINAL EXPERIMENTAL
HC CONCENTRATION
I
80
160 240
TIME, min
\.
320
Figure 2.10. Experiment 231, Dilute Auto Exhaust (Vehicle with Emission
Controls). Simulation of Reactive Hydrocarbon Decay
Table 2.7 contains the rate constants used in the simulation. Note
that the ratio k, /k. is 2.5, a plausible value in view of the relative
4a 4
reactivity of the components of each hydrocarbon group. Note also that
the value of k, is the same as that used for n-butane in the binary
4 ,
mixture experiments, whereas k. = 2.5 x 10 lies in the range used for
propylene. It is also interesting to note that the value of k is only
three times greater than the dilution "rate constant" of 5.98 x 10
Finally, the branching factors were increased considerably from previous
values, with b = b = 8 and b = 1 for both hydrocarbon groups.
This was to be expected in view of the increased length of the chains.
The OH yield factor was set to y = 1/8 .
Figures 2.11 and 2.12 show the results obtained for experiment 222.
It can be seen that the time phasing of the N02 peak is off, the peak
occurring about 50 minutes late. This of course affects the concentra-
tion-time curve for NO and for this reason, it has not been plotted.
x
39
-------
TABLE 2.7
RATE CONSTANTS USED IN SIMULATION OF EXPERIMENT 231
DILUTE EXHAUST FROM A VEHICLE WITH EMISSION CONTROL
Reaction Number
Rate Constant
Low-Reactivity j
Hydrocarbon
High- Reactivity
Hydrocarbon
1
la
2
3
4
5
3a
4a
5a
6
7
8
9
10
11
12
13
14
15
16
2.67(-l)min l
1.32(-5)ppm~2min~1
2
8
1
2
2
2
7
1
2
.67(41)
.39(41)
^0(44)
•8(-4)
.82(43)
.5(44)
•56(-3)
.0(45)
.0(42)
1.5(43)
3.0(43)
1.0
1
C-3)min~1
0(-3)
5.0(-3)
4.5(43)
1.4(
^Dmin'1
6, 05 (41) rain'1
2.0(-3)min~1
* -1 -1
Units are ppm min unless otherwise specified.
40
-------
MODEL RESULTS
EXPERIMENTAL
DATA
Q NO
A N02
+ NO
X NO,
1 T
o.oo M.OO s.QO 12.00 le.oo eo.oo SM.QO
TIME 1X10 ')
S8.QO
f-B-
3S.OO 36.00
Figure 2.11. Experiment 222, Dilute Auto Exhaust (Uncontrolled Vehicle).
Plot of NO and NO-
41
-------
o
cu
cc
a:
-------
However, the NO levels occurring late in the experiment agree well with
the measured values. Furthermore, the NO peak has the correct magni-
tude. Equally important is the reproduction of the ozone data. Figure
2.12 shows that the time phasing and the magnitude approximate well the
experimental data. Although it is not shown, the final concentration of
reactive hydrocarbon is 0.8.6 ppm compared with 0.7 ppm for the data.
Table 2.8 contains all the rate constants used for simulating experi-
ment 222. Note that k /k. = 2 and that k.. ., has the same value used
4a 4 16
in experiment 231. The values of the branching factors are b.. = b = 5
and b- = b = 10 for low- and high-reactivity hydrocarbons, respectively.
The branching factor b is i
yield factor is equal to 0.1.
The branching factor b is unity for both hydrocarbon groups. The OH
2.6 ADAPTATION OF CHEMICAL MODEL TO ATMOSPHERIC MODELING
The parameters of the chemical system obtained from the smog chamber
simulation tests must be modified when we move from the smog chamber to
the atmosphere. What one hopes to obtain from smog chamber experiments
is a qualitative agreement between laboratory and atmospheric observables.
Modeling the smog chamber experiments then gives us an indication that
the physical mechanism proposed for modeling these observables contains
the most important features of the highly complex phenomena which take
place in reality. The evaluation process using smog chamber data also
gives us an understanding of the model's sensitivity to various parameters.
Having obtained this information about the model using laboratory
data, the quantitative link with atmospheric observables must come from
attempts to mo-del the atmospheric processes themselves. In order to do
this, we must identify those features of the chemical model which are most
likely to be heavily influenced by smog chamber conditions. We must also
find out the degree of correlation which exists between the chamber and
atmospheric mixtures.
43
-------
TABLE 2.8
RATE CONSTANTS USED FOR SIMULATING EXPERIMENT 222
DILUTE EXHAUST FROM A VEHICLE WITHOUT EMISSION CONTROL
Reaction Number
Rate Constant
Low-Reactivity
Hydrocarbon
High-Reactivity
Hydrocarbon
1
la
2
3
4
5
3a
4a
5a
6
7
8
9
10
11
12
13
14
15
16
2.67(-l)min
-1
1.32(-5)ppm~2min~1
2.67(4-1)
7.76(4-1)
1.5(4-4)
2.6C-4)
3.45(4-3)
3.0(4-4)
8.54C-3)
1.0(4-5)
2.0(4-2)
1.5(4-3)
3.0(4-3)
1.0(-3)min
1.0(-3)
5.0(-3)
4.5(4-3)
-1
1. 4(4-1) min"1
6. 05(4-1) min"1
2.0(-3)min~1
* -1 -1
Units are ppm min unless otherwise specified.
44
-------
2.6.1 Wall Effects and Dilution Effects in Smog Chambers
Two factors characteristic of smog chambers which are most likely
to influence the parameters as well as the nature of the chemical model
are dilution effects due to sampling in the smog chamber and wall effects.
Omitting dilution from the model has the effect of requiring unrealisti-
cally high values for some of the adjustable rate constants. This is
especially significant with k.-adjustments needed to obtain satisfactory
simulations.
Wall effects influence the nature of the model. In our particular
case, three reactions, (2.13)-(2.15), have been added to try to account
for the nitrogen imbalance which is presumably due to NO reacting on
the chamber walls. Whether these reactions will play any role under
atmospheric conditions is not known, but intuitively one would expect
them not to be significant. Hopefully, the evaluation process under
atmospheric conditions will yield an answer to this question.
The concerns we expressed earlier regarding the importance of wall
reactions as an NO sink were confirmed by comparison of absolute reac-
X.
tion rates throughout the simulation. The reaction of NO with water
on the wall to form nitric acid dominated the NO removal as would be
9 X
expected from prior experimental findings. Specifically, this reaction
rate exceeded the gas phase production of nitric acid by about a factor
of 6. In the course of checking out sensitivity of the system to the
chain breaking reactions, we individually varied reaction rate constants
in reactions (2.12), (2.13), (2.14), and (2.15). A measure of the sen-
sitivity was the influence on nitrogen dioxide decay at late time. The
greatest sensitivity of all was exhibited with respect to variation in
k.. . A threefold increase in this rate constant, for example, resulted
in a fourfold decrease of end-point N0? concentration in the simulation.
The next reaction in the sequence, that between nitrogen dioxide and
nitrogen trioxide, had a lesser effect. Perhaps counter to intuition,
an increase in k-, by a factor of three actually resulted in a slight
45
-------
(less than 10%) increase in end-point nitrogen dioxide concentration.
Rather large changes in the wall reaction rate k were introduced,
but these had virtually no influence on the system.
Samples were periodically withdrawn during each smog chamber experi-
ment to analyze the composition of the gas. "Clean" air replaced the
sample in each case. In our evaluation process, we have found that this
dilution can play a significant role in smog chamber experiments. Thus
dilution was accounted for in all of the simulations discussed previously.
Given an average volumetric dilution rate for an experiment, the
first step was to convert this rate to a first-order "rate constant" by
dividing by the chamber volume. Then we subtract a factor due to dilu-
tion from each chemical rate equation as shown below:
=R-6c (2-23)
where c. = ith species concentration
i
R. = chemical rate for ith species
i
6 = average fractional dilution = average volumetric dilution
rate/chamber volume
To test the validity of our approach for simulating dilution effects,
we used the ethane concentration data in the propylene experiments, since
ethane can be considered to be essentially unreactive. The predicted and
observed ethane concentrations for one such experiment, experiment 336,
are shown in Fig. 2.13. It can be seen that the two sets of data agree
closely. The maximum error is 5%. Additional checks with other experi-
ments produced similar results.
46
-------
EXPERIMENTAL DATA
COMPUTED
CONCENTRATION
80
160 240
TIME, min
320
400
Figure 2.13. Experiment 336. Dilution Model Compared with Measured
Ethane Concentration
The effects of dilution on the reactive species can be seen in Fig.
2-. 14 which contains a plot of propylene concentration with and without
dilution; The effect on other species was similar but the degree of
influence varies for each species. Thus for nitric oxide, the dilution
effect is generally small due to the rapid decay of NO by chemical
reaction, ±.e., R^ » 6c in Eq. (2.23). On the other hand, for
ozone, dilution has been observed to play a large role late in the
reaction*
Dilution can also account for some of the discrepancy found in
the nitrogen balance in smog chambers. Our simulations indicate that
dilution effects can account for up to 25% of the nitrogen loss for those
experiments used in these model validations •.
From the above remarks » it is clear that dilution can have a sig-
nificant impact oh the whole reaction. Furthermore, because of system
47
-------
0.6
0.4
0.2
DILUTION
WITH DILUTION
80
160 240
TIME , min
320
400
Figure 2.14. Experiment 336. Effect of Dilution on Propylene Concentration
(Curves Computed using a Single set of Rate Constants)
nonlinearities, simple scaling cannot be used to compensate for dilution
effects. Thus caution must be exercised when attributing changes in
species concentration to chemical causes alone. Therefore, it is impera-
tive that dilution data be included when experiments are reported in the
literature.
As a final comment, we note that the approach described above
assumes a uniformly distributed dilution which implies a continuous
sampling process. This is, of course, not the case, since sampling is
done at irregular intervals and no dilution takes place between sampling
points. Whether approximating the discontinuous sampling process by a
continuous one has any significant effects on the simulation remains to
be determined.
48
-------
2.6.2 Smog Chamber vs Atmospheric Mixtures
The reactivity of the mixture is one measure that can be used as
a guide to estimate the magnitude of the modifications in the rate con-
stants of the hydrocarbon reactions. We have followed this approach in
e\i
previous work. Table 2.9 shows the mole-weighted reactivity of the
hydrocarbon mixtures in the atmosphere and in the smog chamber experi-
ments previously described. The measure of reactivity is based on a
25
hydrocarbon-consumption scale due to Altshuller. The reactivity of
the Los Angeles atmosphere was obtained from our previous study of Los
n r
Angeles atmospheric reaction data.
It is obvious from the table that the dilute auto exhaust approxi-
mates the reactivity of the Los Angeles atmosphere. So we expect these
experiments to yield the most useful rate constant data for the model.
Our previous approach of using a chemical model validated only for pro-
pylene and dividing its hydrocarbon rate constants by three for atmos-
pheric modeling purposes is, of course, a result of the reactivity
TABLE 2.9
MOLE-WEIGHTED REACTIVITY OF ATMOSPHERIC
AND SMOG CHAMBER HYDROCARBON MIXTURES
A
Mixture Reactivity
Los Angeles Atmosphere 6+2
Propylene 17
Toluene 3
Toluene/n-Butane 1-6
Auto Exhaust—With HC and CO exhaust 6.8
Auto Exhaust—Without HC and CO exhaust 6
* , 25
Based on a hydrocarbon consumption scale.
49
-------
relationship shown in Table 2.9. Finally, we note that the reason for
the controlled vehicle exhaust to be slightly more reactive than the
uncontrolled vehicle exhaust is due to the fact that the propylene con-
tent of the former is a larger fraction of the total hydrocarbon mix than
is the case for the uncontrolled exhaust.
50
-------
3 MODEL METHODO-LOGY IMPROVEMENTS
3.1 PERSPECTIVES ON MODEL UPDATING
27
Initially, our model treated a constant-thickness air layer with
a very compact lumped kinetic scheme. Generalization of the model intro-
duced advection and more realistic chemistry, but ground-fixed coordinate
systems led to unacceptable numerical diffusion errors and the chemistry
O /
still needed more chain-termination steps. Subsequent work used scaling
r\ r
parameters from actual air data for hydrocarbon reactivity adjustments
and for nitric oxide emission adjustments. These adjustments have been
necessary in past work to convert theoretical estimates to the values
that better approximate what actually occurs in the atmosphere. Further
model development saw the replacement of the Eulerian coordinate system
by the semi-Lagrangian system now used. Substantial improvements in computing
efficiency and accuracy were also made by introducing Fade approximant
numerical techniques.
Still, some lingering questions remained unanswered., The present
work addresses these questions. Chemical refinements were described in
the previous section. These studies confirmed the validity of o.ur earlier
approaches in converting smog chamber findings to the atmosphere. More-
over, the main structure of the model must necessarily deal with the
advection, diffusion, and the array of sources as they all interact in
air masses moving through regions where air pollution is to be predicted.
The following subsections deal individually with the current round of
improvements in modeling these phenomena.
3.2 ADVECTIVE AIR TRAJECTORIES
The speed and direction of the air mass-center are determined, as
O /
before, by taking weighted averages of wind speeds and directions from
neighboring measurement stations. It is useful to. examine the theoretical
basis for the weighting and its detailed application in the current phases
of our work.
51
-------
Reciprocal distance weighting of wind station readings is used as
a theoretical basis of interpolation. The rationale for reciprocal-
distance over reciprocal-distance-square stems from the nature of fluid
dynamic singularities in plane flows.
Because of the comparisons of horizontal and vertical scales in the
atmospheric boundary layer problem, the flow may be regarded more nearly
planar than three-dimensional. Clearly, three-dimensional effects occur
at convergences and do generate vertical velocities, but horizontal
advection seems to be the most important character of the lower atmospheric
flow fields we are treating. This being the case, it can be conceived
that any velocity field can be generated by the cumulative effect of plane
singularities such as sources, sinks, and vortex elements. Classical
fluid dynamics shows that the dependence of the influence of each one of
these singularities upon distance is reciprocal in the distance from the
flow element. Thus, if the flow at a field point is assumed to be the
superposition of flows described at neighboring points, it follows that
the relative weight given to each neighboring point should be proportional
to its reciprocal distance from the field point.
The output of hand calculations of air trajectories takes the form
of schedules describing hourly magnitude and direction of successive
trajectory segments. For each node point in the trajectory, air-quality
weighting factors are derived in addition to wind weighting factors.
These were applied to concentration data from the two or three nearest-
neighboring air monitoring stations (which do not generally correspond
to the neighboring wind measurement stations). This gives an estimated
hourly history of air quality along the trajectory, thereby affording a
much larger data base than merely the end point concentrations. The net
effect of the expanded base is to impose stricter standards on validation
compared with end point receptor-oriented calculations.
52
-------
Figure 3.1 shows two typical Los Angeles morning trajectories.
(Stations having hydrocarbon data were chosen as origins for the trajec-
tory to minimize uncertainties in initial conditions.) In this case,
the Air Pollution Control District (APCD) Downtown station has hourly
total hydrocarbon and methane concentrations. The computed crossing
of trajectories may not have occurred had there been data which would
allow the use of a shorter interval size in the calculation. The sampling
station located at the City of Commerce has gas chromatographic data
commencing in the early morning hours. The validation runs for these
trajectories can be checked against interpolated air quality data from
nearby monitoring stations.
Figure 3.2 illustrates a study comparing the surface wind trajectory
calculations with winds aloft tracked by an ESSA tetroon. In this example,
LEGEND:
DOLA - DOWNTOWN LOS ANGELES
COM
ELM
POMA
WHTR
WNTT
RLA
RVA
VER
COMMERCE
EL MONTE
POMONA
WHITTIER
WALNUT
RANCHO LOS AMI 60S
-RIVERA
VERNON
1330
POMA
WNTT
OCT. 29, 1969
Figure 3.1. Air Trajectory in the Los Angeles Basin
53
-------
28
•TETROON (MEASURED)
•GROUND (CALCULATED)
MISH
SEPTEMBER 30, 1969
BURBANK
COMMERCE
DOWNTOWN LOS ANGELES
HOLLYWOOD
LACA —LA CANADA
LOS ANGELES INT'L AIRPORT
LENNOX
MISSION HILLS
PASADENA
VENICE
WEST LOS ANGELES
Figure 3.2. Comparison of Ground Trajectory with Tetroon Trajectory
the solid line is the tetroon path and the dashed line is the calculated
surface trajectory. Short dashed arrows along the tetroon trajectory show
one-hour segments computed from wind station data. Except for the initial
segment, the computed surface winds have a rightward heading from the
tetroon path (looking in the direction of tetroon motion). The result is
the surface trajectory heading north through Cahuenga Pass, over Burbank
and into the Verdugo Hills, while the tetroon travels due west in an
offshore direction until it veers inland over Santa Monica late in the
morning; therefore, it appears that these differences between surface and
28
elevated winds can be especially distinct. Although Angell, et al.,
analyze such disparities in their paper, their comparisons exhibited a
higher degree of correspondence between paths than ours. The difficulties
in neglecting height-dependence of advection are obvious to the modeler;
however, neither theoretical nor empirical corrections are presently
available. The means for handling this effect may come out of boundary
layer meteorological researches.
54
-------
Even at a single location, variations in wind speeds and directions
appear to ciccur due to localized effects. Figures 3.3 and 3.4 show plots
of Los Angeles Air Pollution Control District data versus Scott Research
Laboratories data. Both sets of data were reportedly taken at El Monte,
California. Because of topographic setting and instrumental differences,
the simultaneous readings show wide variances. This may be due to hourly
averages of the Scott data versus instantaneous readings of the Los
Angeles APCD (LAAPCD) station. In evaluating any air pollution simulation
model, this sort of deviation must be considered, as well as that described
in the preceding paragraph. Confronted with these conflicts, we attempted
to resolve the differences by taking arithmetic means.
An objective technique for approximating continuous wind fields
from discrete data will be needed in the near future. In the course of
carrying out the hand calculations, we have developed the following
logical structure on which such a technique might be based:
1. Forward differencing is the zero order approximation; that
is, each segment is laid out based on the hourly averaged
wind that is interpolated from station data surrounding its
origin-point.
2. Higher approximations will require interpolation of wind data
to the midpoint of the zeroth order segment and then refining
the magnitude and direction of the segment in the first
approximation.
3. Reiteration of the segment calculation may be coded with
either an iteration counter or a convergence criterion to
terminate the succession of approximations for the segment.
(Discreteness of wind data may prevent convergence if
proximity tests intervene in the sequence and alter the
selection of wind stations utilized for interpolation).
4. Beyond some maximum distance, wind stations must be rejected
even if it means reducing the number of input stations.
55
-------
12,
10
E
et
• •
I • •
_L
I
I
468
SCOTT WIND SPEED DATA, mph
10
12
Figure 3.3. Comparison of Wind Speed Measurements at El Monte
(Mornings in 1969)
56
-------
500.-
400
CD
Ol
«=c
Q
z
o
t—I
I—
CJ
300
Q
Q
200
Q
C_5
Q-
100
TOO
200
300
400
SCOTT WIND DIRECTION DATA, deg
500
Figure 3.4. Comparison of Wind Direction Measurements at El Monte
(Mornings in 1969)
57
-------
5. Closer than some minimum distance, a single station's wind
vector should be used directly.
6. If all station distances from the region of interest exceed
a certain threshold (input) for a certain numBer of segments
(also input) the computation should be terminated with an
exit flag.
7. Numerical experiments with directional weighting should be
conducted for both trajectory generation and for inter-
polated air quality at a trajectory point; for example, the
air quality should be more sensitive to where it has been
than to where it is going. For improving the zero-order wind
motion approximation, direction and magnitude of a trajectory
segment will be more sensitive to station readings near its
future path than those it is moving away from.
8. Numerical experiments should be conducted on the relative
-1 • -2
merits of r -weighting versus r -weighting. This is done
most directly by using a wind station as an unknown point.
Then the three nearest neighbors (obeying the distance and
barrier selection laws) are weighted by each power and the
merit of each is evaluated by some measure such as least sum
of square residuals between predicted and measured value's i
3.3 VERTICAL AND HORIZONTAL EDDY DIFFUSION COEFFICIENTS
In the evaluation phase of developing the GRC photochemical/diffusion
model, greater emphasis is placed on the use of temperature profiles for
0 "7
the test days. This led us to reappraise our earlier formulation of the
eddy diffusivity profile by incorporating more of the measured data that
have been reported in the literature.
58
-------
Up to now, the eddy diffusivity profile has been assumed to be
ft
trapezoidal from the "ground" Up to the mixing depth (previously set at
29
the inversion base). Following Estoque> the ramp portions of the profile
extends to a height of 50 meters and nonvanishing Values are set at the
bottom and top of the vertical mesh. Assigning the flat portion of the
profile of value of K , the vertical diffusivity, we used the formula
Zoo
2
K^ * 8300 (u + 500) cm /seconds where u is the wind speed in cm/second.
OO
The approximation was based on a fit of scattered data from various sources
77
cited in our 1969 paper.
30
A key element in the review of; this approach is the use of Hosier's
vertical diffusivity data (for a 0 to 90 m height interval) to determine
a useful method of calculation that can be based on available data.
Figures 3.5 and 3.6 show the vertical diffusivity plotted versus wind
speed and vertical temperature gradient. The widely scattered values in
either Case are discouraging to the theorist who wishes to use finite
difference diffusion terms. Note on Fig. 3.5 that the wind velocity
dependence previously chosen for K is inconsequential compared to the
z
two-ofder-of-magnitude scatter of the data. In contrast, an examination
of Fig. 3.6 reveals a relatively systematic dependence on vertical tem-
perature gradient. The velocity formula plotted on Fig. 3.5 gives an
estimate of vertical diffusivity lying largely in the neutral stability
range. This may be expected to approximate conditions .averaged over space
and time in the marine layer overlaid by an inversion layer.
On the other hand, temperature gradient inputs are apparently far
more influential than wind speed inputs in determining vertical diffusivity
values. As an improvement on our earlier approach, we have reconstructed
Because of the assumption of uniform horizontal velocity profiles, we
take OUE "ground" elevation to be something like mean rooftop height
where the air sampler inlets were located on the Scott stations. This
is equivalent to setting z - 0 at elevations an order of magnitude
or ewo greater than roughness heights.
59
-------
10 V
CSJ
o
10
I— I
05
+ 500)
I
J
0 100 200 300 400 500
WIND SPEED, cm/s
600 700
Figure 3.5. Vertical Diffusivity Versus Wind Speed (Ref.. 30)
60
-------
icr
• *•!
CXI
05
100 < u < 700 cm/s
CURVE FAIRED TO DATA
UNSTABLE
VERY \
UNSTABLE-I*1*-}
-2
0 1
(AT/AZ), °c/ioo m
Figure 3.6. Vertical Diffusivity Versus Vertical Temperature Gradient
(After Ref. 30)
61
-------
the diffusivity profiles to represent the broad stability categories
delineated along the bottom of Fig. 3.6. Dropping the dependence on
wind speed and introducing the dependence on temperature gradient, we
might expect a larger dynamic range of diffusivity values throughout any
given simulation run. It is difficult to justify any more elaboration
on the diffusivity calculation on the basis of the body of theory that
depends on fluxes and stresses. The guesses necessary to arrive at a
Richardson number or a Monin-Obukhoff length parameter cannot be seriously
expected to yield better results than the correlation shown in Fig. 3.6
(especially in view of the poor correlations with Richardson number cited
by Hosier).
Having a guide to the choice of the plateau value of vertical diffu"
sivity, we now turn our attention to assigning a ground value for diffu-
-2
sivity. Surface shear stresses typically vary from 1 to 10 dynes-cm
*
leading to friction velocity u values of 20 to 90 cm/second. For
*
neutral stability conditions K = ku z assuming that the vertical diffu-
z
sivity approximately equals the eddy viscosity. Typically, the coefficient
k takes on values about 0.4 if u - 50 cm/second. Assuming the "ground"
plane to be at about 5 meters, we get K(0) ~ 10 cm /second. Considering
the larger roughness for urban areas as compared with the surfaces in
32
Table 3.1 on p. 72 of Pasquill's book, this falls just above the range
of K values listed there. For very stable conditions as defined on
Fig. 3.6, the value of K(0) is likely to be far less than 10 cm /second.
Translating the ground values and the elevated values of vertical
diffusivity (Fig. 3.6) into profiles, we arrive at the data shown on
Fig. 3.7. The choice of 50 m for the knee height illustrates a nominal
value. Because of uncertainties, this height is taken to be the first
mesh point above the ground in a five-point mesh. For mesh ranges
typically varying from 100 to 300 m, the knee will be 25. m to 75 m above
the ground. Details of K-profile shape at the top of the mes.h have little
62
-------
2A
CI5
2x10
KNEE HEIGHT A ± 50 m
ACCORDING TO ESTOQUE
29
IN MODEL CALCULATIONS A ^ 30
TO 70 m, DEPENDING ON MESH
INTERVAL SIZE
4x10
VERTICAL DIFFUSIVITY, cm^/s
CO
Ł
in
tvj
oi
I
2x10'
IN URBAN, AREAS 0-F VERY TALL BUILDING HEIGHTS, ESTOQUE'S ESTIMATE MAY BŁ
TOO LOW; HOWEVER, THIS EXCEEDS ROUGHNESS HEIGHTS IN NEARLY ALL AREAS
OVER REGIONAL EXTENT.
Figure 3.7- Vertical Diffusivity Profiles (Fig. 3.6 Defines Stability
Categories)
63
-------
influence for the space and time scales of our modeling calculations;
therefore, for photochemical/diffusion validations the profiles are
taken to be constant all the way to the top.
Reference to Fig. 3.6 shows how the vertical diffusivity values are
assigned. Data on the graph are for 100 cm/second to 700 cm/second wind
speed. As mentioned previously, a comparison of Figs. 3.5 and 3,6 shows
a tighter grouping of points on the temperature gradient plot than there
is on the velocity plot; hence, the use of this correlation in place of
the velocity formula. Hybridization of these profiles may be a good
approach if the vertical mesh is taken to be higher than the inversion
base.
Horizontal diffusivities, K , were needed for the transverse
diffusion tests of the model that are reported elsewhere. The choice of
/: n
a typical value, 5 x 10 cm /s, was based on the Los Angeles tetroon data
33
of Angell and Pack. One can infer a Fickian diffusion coefficient from
the horizontal separation of tetroons flying a few hundred meters above
the ground. The values grow with time so that some time scale charac-
terizing the problem must be selected. It will be noted from Fig. 3.8
that the value of K selected represents a travel time of about one hour.
Figure 3.8 shows lateral separation of simultaneously released tetroon
pairs as a function of time since release for flights in morning (IFA)
and afternoon (IFF) at Idaho Falls, and flights at Atlantic City (ACY)
and Los Angeles (LAX) . The heavy line represents the mean of these
weighted according to number of cases. Also entered are lateral eddy
diffusivity (K ) isopleths based on Fickian theory, and dissipation
isopleths («) based on similarity theory (data reproduced from Angell
and Pack).33
64
-------
30 100
TRAVEL TIME, min
300
Figure 3.8. Lateral Separation of Simultaneously Released Tetroon Pairs
as a Function of Time Since Release (after Ref. 33)
3.4 EMISSION FLUX HISTORIES
For the purposes of production computations, we produced a code
that traced air trajectories using a 2-mile interval grid for numerical
computations and source inventory compiled by Roberts, Roth, and Nelson,
The inputs are wind speeds and headings along line segments of a trajec-
tory. The output is the flux of each pollutant into the air parcel for
the time/space-history traced out by the parcel. This code, which was
developed by R. Nordsieck prior to the present phase of the work, produces
output on punched cards suitable for use as inputs to the GRC photochemical/
diffusion model.
The primary revisions made in the emissions model during this study
involved- (1) the introduction of additional spatial and temporal variations
in freeway traffic emissions resulting from vehicle emission rate vfria-
tions with average speed, and (2) inclusion of a correction factor for
surface street emissions to account for the non-uniform temporal distri-
bution of vehicle cold-start emissions.
65
-------
4 TRANSVERSE DIFFUSION AND ITS EFFECT ON THE GRC MODEL
Two analyses have been conducted to assess the errors incurred in
the model calculations due to the omission of transverse diffusion in the
semi-Lagrangian formulation. In the first, we have assessed the effects
of lateral exchange between adjacent stream tubes. In the second analysis,
we have quantified the errors which can result when our Lagrangian control
volume passes near an elevated point source, but fails to sweep over it,
thus ignoring its contribution that spreads laterally into the control
volume. Both of these investigations were performed using our L_ocal Air
Pollution ^simulation (LAPS) code (see Appendix A, Sec. A.3.2) which
incorporates lateral diffusion but is limited to quasi-equilibrium chemistry.
4.1 LATERAL DIFFUSION BETWEEN NEIGHBORING STREAMTUBES
To examine the effect of lateral exchange between "parallel" trajec-
tories, we located a relatively straight trajectory originating at the
Commerce station and moving north to Burbank. We then synthesized a
*
parallel trajectory two miles to the west by simply moving the start
point two miles west and assuming an identical wind history. Figure 4.1
shows histories of CO surface flux obtained for these trajectories from
our source emissions program. (The fluxes are normalized with respect to
air density to yield the somewhat unusual units of meters per minute.)
Assuming a 200-m inversion height, and the slightly unstable diffusivity
profile shown in Fig. 4.2, the LAPS model was used to simulate atmospheric
CO concentrations within two parallel area-source strips 3 km wide.
The lateral cell size was 500 meters, providing six cells over each source
strip. Outside boundaries of the horizontal mesh were assumed fully
reflecting, simulating a semi-infinite area source on each side. Initial
concentrations of CO were set to zero everywhere. Figures 4.3 and 4.4
show the concentration histories obtained along two symmetrically located
*
Two miles was selected because the minimum resolution which can be
expected in our model is set by the two-mile aggregation imposed by the
source model.34
**
The two-mile distance was approximated by 3 km in this test case for
convenience in the use of metric units.
66
-------
RIGHT SIDE TRAJECTORY
• LEFT SIDE TRAJECTORY
TIME, hours
Figure 4.1. CO Flux Histories on Two Neighboring Trajectories
200
INVERSION BASE
o
3.
100
VERTICAL EDDY DIFFUSIVITY, 104 cm2/s
Figure 4.2. Eddy Diffusivity Profile for Neighboring Streamtube Analysis
67
-------
8r
5 6
o
co
Q-
D-
o
o
o 2
i_i ^
C_>
z
o
RIGHT SIDE WITH
LATERAL DIFFUSION
RIGHT SIDE, NO EXCHANGE
FULL REFLECTION AT OUTER BOUNDARIES
I J I
012345
TIME, hours
Figure 4.3. CO Concentration Histories on Right-Hand (Commerce) Trajectory
^, 6
LEFT SIDE, NO EXCHANGE
LEFT SIDE WITH
LATERAL DIFFUSION
FULL REFLECTION AT OUTER BOUNDARIES
TIME, hours
Figure 4.4. CO Concentration Histories on Left-Hand (Synthetic) Trajectory
68
-------
parallel paths 3.5 km (about 2 mi) apart with and without lateral diffusion
/• rt
(using a lateral eddy diffusivity of 5 x 10 era. /s) as discussed in Sec. 3.3.
The quantitative differences seem to be small.
To get an idea of what a worst case might be like for the parallel
streamtube analysis, we superposed the CO flux histories from 24 test
trajectories to indicate the real-world bounds on CO flux histories in
the Los Angeles Basin (Fig. 4.5). It was decided that the envelope shown
in Fig. 4.6, representing a constant ratio of 3:1 between maximum and
minimum, would provide a very acceptable bounding case for two trajectories
only two miles apart. Using the simulation parameters described above,
the concentration histories shown in Fig. 4.7 were obtained for the worst
case comparison.
10.00
LOCAL TIME
Figure 4.5. Superposed CO Flux Histories for 24 Trajectories
69
-------
8
e 6
"E
° 4
X
U_
o
0
Figure 4.6. Bo
-
___
_ J
•=<
L__ HIGH FLUX
1 | LOW FLUX
1 1
5 7 9 11 13
TIME, hours
unding CO Flux Histories for Parallel Trajectory Analysis
FULL REFLECTION AT OUTER BOUNDARIES
3 4
TIME, hours
Figure 4.7. CO Concentration Histories for Worst-Case Parallel
Trajectories
70
-------
Table 4.1 summarizes worst case errors that may be encountered under
the assumption of zero initial concentration. The fractional error will
be lower if the initial concentration is greater than zero. In effect,
Table 4.1 shows the worst case errors for increments of concentration
rather than for concentration itself. Table 4.2 shows the worst case
errors that could be found assuming an initial concentration of 10 ppm
of CO, which is a typical early-morning value.
TABLE 4.1
"WORST CASE" BOUNDING ERROR FRACTIONS DUE TO
OMISSION OF LATERAL DIFFUSION IN URBAN MODELING
ASSUMING ZERO INITIAL CONCENTRATION OF CARBON MONOXIDE
Time
5 hr
8 hr
High Flux
0.28
0.36
Low Flux
0.39
0.44
TABLE 4.2
"WORST CASE" BOUNDING ERROR FRACTIONS DUE TO OMISSION
OF LATERAL DIFFUSION IN URBAN MODELING, ASSUMING
AN INITIAL CONCENTRATION OF CARBON MONOXIDE OF 10 PPM
Time High Flux Low Flux
5 hr 0.12 0.14
8 hr 0.17 0.19
71
-------
4.2 LATERAL DIFFUSION EFFECTS IN THE VICINITY OF HIGH-FLUX
POINT SOURCES
A quantitative assessment of the pollutant contributions of high-
flux elevated point sources to neighboring trajectories was performed in
a parametric fashion, by determining the ranges of real-world surface
fluxes and point source fluxes encountered in the Los Angeles Basin and
examining the results of superposing various combinations.
Power plant stacks were chosen as typical high flux elevated pollu-
tant sources, and hence, in order to set background fluxes, we needed to
establish the range of surface NO fluxes encountered in the Basin. Pro-
ceeding as with the CO fluxes, Fig. 4.8 shows the NO flux histories
obtained for our 24 test trajectories. The heavy lines indicate the three
fluxes selected fr.om the range of NO fluxes encountered. They are equiva-
lent to 10, 15 and 20 kg/hr/km . The source model of Science Applications,
13.00 14.00
Figure 4.8. Superposed NO Flux Histories for 24 Trajectories
72
-------
34
Inc. shows a distribution of power plant NO fluxes ranging up to
532 kg/hr (1172 Ib/hr) . From this distribution, we chose nominal stack
NO fluxes of 60, 200, and 600 kg/hr as including roughly 33%, 67%, and
100% of the power plants on a relative occurrence basis. An effective
point source height of 100 meters was used to include the height of an
average stack plus the additional plume rise due to jet momentum and
buoyancy effects.
The meteorological conditions simulated for these cases included
an inversion base height of 183 meters with neutral (2 * 10 cm /s)
3 2
diffusivity below the inversion and very stable (5 x 10 cm /s) above.
The resulting profile of vertical eddy dif fusivities is shown in Fig. 4.9.
6 ?
As before, the lateral diffusivity was set at 5 x 10 cm /s. A light wind
of 1 m/s (about 2 knots) was assumed.
o
CŁ
03
O
CO
300,-
200
100
INVERSION BASE
Figure 4.9.
0123
VERTICAL EDDY DIFFUSIVITY, 104' cm2/s
Eddy Diffusivity Profile for Elevated Point Source/Lateral
Diffusion Analysis
73
-------
Since the object of these tests was to assess the effect of lateral
diffusivity on trajectories missing large point sources, and the basic
resolution of our source model is two miles, we have chosen to simulate
the average situation in which the stack is centered in a neighboring
two-mlle-wide path and we are reading concentrations two miles distant
at the center of another two-mile swath. Accordingly, Figs. 4.10, 4.11,
and 4.12 show NO concentrations at the ground, two miles from the plume
centerline for each of the three background NO fluxes combined with stack
NO fluxes of 0, 60, 200, and 600 kg/hr. Below these curves, Figs. 4.10,
4.11, and 4.12 plot the corresponding percent error incurred if lateral
diffusion were ignored and only the background flux is accounted for.
Table 4.3 summarizes the maximum fractional errors associated with each
combination of background and stack fluxes. Even under the worst condi-
tions, these error percentages are much smaller than NO-flux uncertainties.
74
-------
20
15
o
10
0 5
a
o
STACK FLUX, kg/hr
600
200
60
0
NO SURFACE FLUX = 10 kg/hr/knT
RECEPTOR 2 mi FROM PLUME CENTERLINE
EFFECTIVE STACK HEIGHT = 100 m
WIND SPEED = 1 m/s
NEUTRAL STABILITY
«6
NJ
4
01
LU
2§ r-
20
15
10
STACK FLUX kg/hr
600
20
60
TIME, min
80
100
120
Figure 4.10. Srouiid Coneeiitratiori fiffeets Of Elevated Po'int Sources:"
2
lO.kg/hE/km. Background Flux
75
-------
.0-
o_
30
25
20
LU
O
o
o
o
o;
STACK FLUX, kg/hr §
NO SURFACE FLUX 15 kg/hr/km"
RECEPTOR 2 mi FROM PLUME CENTERLINE
EFFECTIVE STACK HEIGHT = 100 m
WIND SPEED = 1 m/s
NEUTRAL STABILITY
o
o;
o
o;
STACK FLUX kg/hr
600
60 80
TIME, min
100
120
Figure 4.11. Ground Concentration Effects of Elevated Point Sources:
2
15 kg/hr/km Background Flux
76
-------
a.
O-
o
o
o
•z.
o
o
QŁ
CS
STACK FLUX, kg/hr
600
200
60
NO SURFACE FLUX = 20 kg/hr/km^
RECEPTOR 2 mi FROM PLUME CENTERLINE
EFFECTIVE STACK HEIGHT = 100 m
WIND SPEED = 1 m/s
NEUTRAL STABILITY
cc.
o
O
Of
10
STACK FLUX kg/hr
,600
20
40 60 80
TIME, min
120
Figure 4.12. Ground Concentration Effects of Elevated Point Sources:
20 kg/hr/km2 Background Flux
77
-------
TABLE 4.3
MAXIMUM FRACTIONAL CONTRIBUTION OF AN ELEVATED POINT
SOURCE AT A GROUND LOCATION TWO MILES FROM THE PLUME CENTEPvLINE
Effective Stack Height =? 100 m
Neutral Stability, 2 knot wind
Average
Surface
Flux,
kg/hr/km2
10
15
20
Stack Flux, kg/hr
60
0.022
0.015
0.011
200
0.072
0.048
0.036
600
0.217
0.145
0,109
78
-------
5 VALIDATION STUDIES
5.1 SELECTION OF DA.YS FOR MODEL TESTS
The culmination of the improvements in chemistry and meteorology is
a set of controlled retests of the GRC Photochemical/Diffusion model and
a conversion of the code to IBM 360-compatible form. The selection of
days was originally predicated on a consistent basis with the design of
35
the Scott Research Laboratories Los Angeles Basin Program; namely, that
morning air movements from Commerce to El Monte would be studied. The
design philosophy was that primary pollutants (reactants) would be carefully
measured at Commerce with special emphasis on gas chromatographic resolution
of hydrocarbon samples. Advection and reaction of the air mass would then
occur with the composition of the secondary pollutants (products) indicated
by (hopefully) downwind measurements taken at El Monte. To implement this
philosophy, we conducted a systematic search for days having morning air
trajectories that nearly connected the two stations. The original program
design did not consider the use of Los Angeles County Air Pollution Control
District station data because they are difficult to obtain and they are not
as detailed as the Scott Research Laboratories station data.
According to the field program objectives, our rationale for a
requested selection of 9/4/69, 9/15/69, 9/27/69, 10/16/69, 10/29/69, and
11/4/69 was based on air movement calculations. To get these dates, we
calculated trajectories that could be used to validate the transport/
diffusion, mo.dule (and ultimately the photochemical/diffusion model). The
calculations used the 1969 Scott Research Laboratories data which were
collected at Commerce and El Monte in the Los Angeles Basin. We performed
a search for trajectories which originated at Commerce in the morning and
arrived at El Monte later in the day. The se'arch was accomplished using
a special trajectory-generation program designed to compute a wind trajec-
tory between two stations given wind speed and direction data at each
station. The program calculates inverse distance-weighted averages of
79
-------
wind speed and direction between the stations and uses these quantities
to generate the trajectory. The output of the program consists of the
following trajectory descriptors:
1. Geographical coordinates of the air parcel
2. Time, wind speed, and direction of the air parcel at
every point in the trajectory
3. Distance of the air parcel from each of the two reference
stations
A computer plot of the trajectory can be obtained on an optional basis.
The program has the capability to detect and compensate for anomalies
in the data such ,as missing data points. Moreover, the program warns the
user about the existence and nature of the data defects and of the actions
taken to overcome the deficiencies. This allows the analyst to assess
the reliability of a trajectory.
We considered an air parcel to have "arrived" at El Monte if it
passed within one mile of El Monte. The search examined all the trajec-
tories originating at Commerce from 0600 to 0900 for the 72 days (Aug. 28
to Nov. 7) of data available at both Commerce and El Monte. A total of
1368 possible trajectories over the 72 days were computer-analyzed. Only
87 trajectories spread over 12 days were found to satisfy the miss-distance
criterion of one mile or less. Additional evaluation of the wind and
aerometric data eliminated five of these days, thus leaving only seven
possible days with usable trajectories. The seven days are listed below
in order of number of trajectories 'available on each day:
80
-------
Date
Oct. 1.6
Nov. 4
Sept. 4
Oct. 29
*
Sept. 6
Sept. 15
Sept. 27
No. of
Trajectories
18
9
9
8
6
3
3
Peak CO (ppm)
(Commerce)
7.9
25.2
12.2
18.4
10.2
10.0
5.0
Peak 03 (pphm)
(El Monte)
10.0
12.0
18.1
19.0
14.6
5.4
24.5
A
Missing data
For Sept. 6, El Monte is missing NO , HC, and CO data. For Commerce, we
X
found that Oct. 29 is missing the NO measurements for the interval 0600-
0730, which is precisely the interval when the trajectories originate.
The El Monte data could have been supplemented by LAAPCD Azusa data,
except for the hydrocarbon measurements. Supplementing the Commerce data
may not have been possible, since no LAAPCD station is located nearby.
The program objectives were then redirected and the original two-
station concept was discarded. The newly adopted selection criteria
stressed availability of airborne temperature data and high peak oxidant.
All but two of the requested dates were discarded in favor of the follow-
ing set of dates: 9/11/69, 9/29/69, 9/30/69, 10/29/69, 10/30/69, and
11/4/69.
The chemical data for Sept, 29 are satisfactory and the point of
closest approach to El Monte is less than a mile. However, the time of
closest approach occurs around 1100. Thus for a start time of 0600, we
have only a five-hour travel time. The data for September 11 are excel-
lent, but for a 0600 Commerce start, the trajectory misses El Monte by
about 6 miles and would require interpolation to estimate the final con-
centrations. The trajectories for Sept. 30 and Oct. 30 move in a westerly
direction from Commerce and would also require interpolation at the
destination point.
81
-------
The decision to obtain Los Angeles County Air Pollution Control
District data and to carry out validation studies for these days forced
changes in our fundamental approach. Because of critical dependence of
model results on initial conditions, we sought trajectories beginning at
* . .
stations having hydrocarbon data in addition to measurements of other-
pollutants. With one exception, the trajectories were source-oriented
rather than receptor-oriented. As mentioned previously, weighted averages
of air quality values were computed from station measurements for each
hourly node of each trajectory.
Three days were designated to allow adjustments in parameters.
These so-called "hands-on" days are 9/29/69, 9/30/69, and 10/29/69. The
remaining three days are reserved for testing the fully adjusted model
without manipulation of coefficients. These so-called "hands-off" days
are 9/11/69, 10/30/69, and 11/4/69. The hands-off days have lower peak
oxidant (generally) than the hands-on days.
5.2 PROGRAM CONVERSION
Conversion of the DIFKIN program for operation on the IBM 360/50
was completed early in the present phase of the work. Specific tasks
carried out can be grouped into the following categories:
1. Elimination of nonstandard software which is incompatible
with the IBM 360/50.
2. Inclusion of FORTRAN software which is characteristic of
the IBM 360/50 and cannot be used in our CDC 6400
3. Testing the accuracy of the solution to a sample problem
to determine the effect of th6, reduced precision of the
IBM 360/50 (see Sec. 5.2.1)
4. Determining the differences in running time between CDC 6400
and IBM 360/50 (see Sec. 5.2.2)
*
The stations are Commerce, El Monte, Downtown, East San Gabriel Valley,
and West San Gabriel Valley.
82
-------
5. Producing a punched deck of cards whose character code is that
used by the IBM 360/50. (Extended BCD for IBM 360/50 compared
to BCD for CDC 6400.)
6. Eliminating diagnostic statements used when the program was
being developed.
7- Using a capability of our computer center to generate a source
deck with all the statement numbers sequentially ordered for
maximum readability.
5.2.1 Precision Test
A sample problem consisting of a polluted air parcel sweeping over
a heavily used freeway was run in single-precision mode on both the
CDC 6400 and the IBM 360/50 and the results compared. For reference
purposes, we recall that the precision of the CDC 6400 and the IBM 360/50
is 14 and 7 decimal digits, respectively. Thus we might expect some dif-
ference in the results produced by each machine. Analyzing the answers,
we found that they agree to four significant figures and differ by at most
two units in the fifth (least) significant digit. Thus the difference is
2 parts in 10,000 at worst. The table below shows the frequency and magni"
tude of the divergence for one of the eleven species computed; the total
number of points sampled is 300.
Units of Difference in
Least Significant Digit Number of Points
0 65
1 227
2 8
Regarding the conversion, two comments are in order. The first is
that since the computational technique is unconditionally stable, no
problems are anticipated due to amplification of the roundoff error. The
second is that for the major species, i.e., NO, N02, HC, 03, CO, the three
most significant figures are sufficient for comparison with experimental
data.
83
-------
5.2.2 Timing Test
The test problem ran for a considerably longer time in the IBM 360/50
than in the CDC 6400. For example, total running time for the IBM 360/50
was 18.6 minutes to compute 30 minutes of real time, for a real time/
computer time ratio of 1.6:1. For the CDC 6400, the figures are 4.4 minutes
of computer time to compute 60 minutes of real time, for a ratio of 14:1.
Thus the running time was increased by more than a factor of 8. Recent
improvements have greatly improved the speed ratios since these timing
tests were conducted. Consequently, the running time for either system
should be reduced considerably to reflect current practice.
5.3 ATMOSPHERIC VALIDATION TESTS
5.3.1 Introduction
In this section, we report the results of validation tests carried
out for four trajectories on each of the following six days: Sept. 11,
29, 30, Oct. 29, 30, and Nov. 4, all in 1969. Three of the trajectories
also serve as baseline cases for the transportation control strategy
study which is described in a separate volume.
September 29, 30, and Oct. 29 have been designated to be so-called
"hands-on" days for validation purposes. This means that parameter adjust-
ments can be made in order to improve simulations of the measured concen-
tration histories of the various pollutants. The adjustable .'parameters
are the diffusion coefficients and the rate constant of the reaction
OH + HC -> (b2) R02 , designated by k, . The experience gained from working
with the hands-on days is then used to develop guidelines for parameter
selection to simulate air quality for the remaining three days, Sept. 11,
Oct. 30, and Nov. 4. The latter are thus designated "hands-off"'days.
The guidelines and the inputs that are necessary to apply them are
described later in the report.
84
-------
In addition to the test results, this section also contains descrip-
tions of the test procedures and criteria used in evaluating the model.
As a necessary adjunct to model evaluation, data base errors are examined
and illustrative examples given. The test results are also assessed on
the basis of statistical correlations between predicted and observed
concentrations.
A map of the Los Angeles Basin which contains the location of the
monitoring stations is included in Fig. 5.1; the legend for the abbrevia-
tions used on the map is found in Table 5.1. A list of the trajectories
is shown in Table 5.2 and the initial concentrations used in the simula-
tions are in Table 5.3.
5.3.2 Approach and Criteria for Model Evaluation
The approach taken for the hands-on cases was to use CO concentra-
tion histories to determine the diffusivity parameters for each trajectory.
The diffusivities thus obtained were then applied without change to the
same trajectory with the reactive species. Because of spatial variability
of meteorological conditions, all the trajectories for a-single day could
not generally be described by the same diffusivity coefficients. In
addition,, it was necessary to adjust the rate constant k, for the reac-
tion OH + HC ->• (b.) RO for some of the reactive cases. Subsequently,
guidelines were worked out for determining the diffusivities and k. to
be used in the hands-off cases. These guidelines appear in Sec. 5.4.
Some remarks are in order about the nature of the test that is
applied to the computed concentrations in relation to observed concentra-
tions. Let us begin by reviewing the simulation process: the computed
results represent the history of the pollutants in an air parcel which
traverses a geographical area following a path determined by the local
hourly average wind speed and direction.
85
-------
J7W HOL
102 W-
WEST*
• 75W
DOLA
34WVERB OCOM
116WPASA m
•79WSGV ^
lltW
1 ALMS O ELM
97W
AZU
/61WONT
1QBW/
POMA
r— /
81W RVA
106WWNTTI
• 22W
"RLA
112WCOMA
LOS ANGELES COUNTY
AIR POLLUTION CONTROL DISTRICT
METEOROLOGICAL NETWORK
JANUARY 1972
72SC
A —101WLONB/
80W -^ /
OOM 21W /
LGB /
80SE |~
• 114W
WHTR
BgRVVKFI "~— ~™ ""*"" Los Angeles Basin Limits
. Upper Santa Clara River
Valley Basin Limits
Antelope Valley Basin Limits
• County Boundary
APCD Air Monitoring and
Meteorological Stations
APCD Meteorological Stations
Cooperating Meteorological
Stations
Scott Research
Laboratories Stations
Figure 5.1. Air Quality and Meteorological Monitoring Network in the
Los Angeles Basin
86
-------
TABLE 5.1
DIRECTORY
MONITORING
Abbreviation
ALMS
AZU
BRT
BUR
BURK
GOMA
CPK
DOLA
DOM
ELM
ENC
FOX
HOL
KFI
LACA
LANG
LAX
LENX
LGB
LONB
MALC
MDR
OF AIR QUALITY AND METEOROLOGICAL
STATIONS IN THE LOS ANGELES BASIN
Location
Alhambra
Azusa - East San Gabriel Valley
Brackett
Hollywood-Burbank Airport
Burbank - East San Fernando Valley
Compton
Canoga Park
Downtown Los Angeles - LAAPGD Headquarters
Dominguez
El Monte
Encino
Gen. Wm. J. Fox Airfield
Hollywood
KFI Transmitter
La Canada
Lancaster
Los Angeles International Airport
Lennox
Long Beach Airport
Long Beach - South Coast
Malibu
Marina del Ray
87
-------
TABLE 5.1 (Cont.)
Abbreviatirn
MISH
MWS
NEW
NP
NTB
ONT
PASA
PICO
PMD
POMA
RB
RESD
RLA
RVA
SAU
SM
SP
VEN
VER
WEST
WHTR
WNTT
ZUM
Location
Mission Hills
Mount Wilson
Newhall
Newport Beach
Los Alamitos Naval Air Station
Ontario International Airport
Pasadena - West San Gabriel Valley
Pico
Palmdale Airport
Pomona
Redondo Beach
Reseda
Rancho Los Amigos
Rivera
Saugus
Santa Monica
San Pedro
Venice
Vernon
West Los Angeles
Whittier
Walnut
Zuma Beach
88
-------
TABLE 5.2
TRAJECTORY IDENTIFICATION TABLE
Trajectory
Number
1
2
3
4
5
6
A
7
*
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
*
24
Date.
Sep 11
Sep 11
Sep 11
Sep 11
Sep 29
Sep 29
Sep 29
Sep 29
Sep 30
Sep 30
Sep 30
Sep 30
Oct 29
Oct 29
Oct 29
Oct 29
Oct 30
Oct 30
Get 30
Oct 30
Nov 4
Nov 4
Nov 4
Nov 4
Origin
Commerce
Commerce
Downtown Los Angeles
Downtown Los Angeles
Commerce
Commerce
Downtown Los Angeles
Near Coast
Commerce
Commerce
Downtown Los Angeles
Downtown Los Angeles
Downtown Los Angeles
Downtown Los Angeles
Commerce
El Monte
Pasadena
Commerce
El Monte
Downtown Los Angeles
Commerce
Commerce
Pasadena
Downtown Los Angeles
Start
Time
0530
0630
0530
0630
0530
0630
0530
0230
0530
0630
0430
0530
0530
0630
0630
0830
0530
0630
0630
0830
0530
0630
0530
0530
Closest
Terminal Station
Rancho Los Araigos
La Canada
West San Fernando
Valley
Mission Hills
Azusa
East San Gabriel
Valley
Walnut
Anaheim
La Canada
La Canada
La Canada
Burbank
Whittier
Walnut
Walnut
Azusa
Orange County
Long Beach
Orange County
Long Beach
El Monte
La Canada
La Canada
Mission Hills
Final
Time
1330
1330
1230
1330
1230
1330
1230
1230
1130
1130
1130
1030
1230
1330
1330
1330
1330
1030
1430
1230
1130
1230
1130
1430
Trajectories used in transportation control strategy study.
89
-------
TABLE 5.3
INITIAL CONCENTRATIONS USED IN ATMOSPHERIC SIMULATIONS
Trajectory
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
CO, r
5.3
8
11
14
9
12
7
7
15.9
22
13
19
9
11
11
6
6
8
8.5
8
16
23.2
6
14
J-ilJ- L. J.CL-L. OULll
NO, pphm
25
17
10
14
39.9
43.5
16
2
67.5
70
35
38
5
25
30
12
21
48
23
13
36
72.6
3
35
-CllULclLXULlS
HC , pphm
60
60
25
25
67
77
40
18
117
129
45
65
30
30
24
30
30
63
27
25
100
130
25
40
NO , pphm
10
11.6
3
3
11.1
10.1
4
9
9.8
12
5
3
11
15
10
20
13
11
13.5
6
18
15.4
9
13
90
-------
The test that is applied to the computed concentrations consists
of comparing the model's results with data which are obtained from air
monitoring stations. This is a severe test inasmuch as an attempt is
being made to match computation and data in magnitude, time phasing, and
space. Consideration of data base errors becomes important in the appli-
cation of such a test and these are discussed in the next section.
For CO, the criterion of goodness of fit consists of matching the
observed concentrations with the computed results. For the reactive
species, we focused on matching the computed ozone concentration with
the observations. The close photochemical coupling among NO, NO , and
0- in many instances precluded a good match for all three species, since
a parameter adjustment made to improve the fit of one of the three would
degrade the match of the others. (Such close coupling is obscured in
the averaged atmospheric samples used in the comparison process since
turbulence effects may interfere—see Appendix A.) Since ozone is
generally considered to be the key indicator of photochemical smog, it
was felt that a good ozone fit at the possible expense of the others was
justified.
5.3.3 Data Base Errors
As was mentioned above, the test of the model consists of comparing
the computed concentrations with observations along an air trajectory.
However, because of the sparseness of the air quality monitoring network,
the monitoring stations will not be on the path of the trajectory as a
general rule. Hence, the data used for comparison must be obtained by
spatial interpolation (in pur case, we use inverse-distance weighting)
from those monitoring stations which are nearest to the trajectory's
nodes• The model tests compare concentrations computed by the model
with concentrations which are interpolated from air monitoring data.
Therefore, the spatial interpolation itself can result in large errors
that bear no relationship to the validity of the model. An estimate of
these errors can be obtained by performing the interpolation for a known
91
-------
monitoring station and then comparing the actual and computed quantities.
Such an exercise was carried out using LAAPCD stations 1 (Downtown Los
Angeles), 60 (East San Gabriel Valley), and 69 (East San Fernando Valley)
for the measured data, and station 79 (West San Gahriel Valley) as the
point where the concentration is assumed unknown. The results for ozone
showed that the relative error, i.e., true-computed|/true, ranged from
6% to 63% with an average error of 35% for Sept. 29, 1969. For ozone on
Nov. 4, 1969, the error ranged from 0% to 33% with an average of 20%.
However, the absolute differences in concentration ranged from 0.3 to
25.8 pphm for Sept. 29 and from zero to 4.1 pphm on Nov. 4. In presenting
the results in graphical form as a comparison of model vs data it is the
absolute difference, rather than the relative error, which is immediately
apparent to the eye. This is clear from Figs. 5.2 and 5.3, which show
the interpolated and actual concentrations for ozone on the two days
mentioned previously. The actual concentration is measured at station 79
(West San Gabriel Valley) and the stations used for interpolation are 1
(Downtown Los Angeles), 60 (East San Gabriel Valley), and 69 (East San
Fernando Valley) . Some sources of error in the interpolation are the
spatial inhomogeneities in source distribution, meteorological conditions,
and terrain. All of these affect the production and flow of pollutants
in the area in question and thereby influence the pollutant levels.
Indeed, the appropriateness of assuming that the pollutant concentrations
at a point are representative of the levels in a region is the basic tenet
that comes under scrutiny when one considers the sources of error in the
interpolation. Mathematical models such as the one being evaluated in
this work may actually help to solve the problem of representativeness
by interpolating under constraints which take into consideration the
spatial inhomogeneities mentioned above.
The interpolation discussed above was performed using inverse-
distance weighting. The same calculation was performed using inverse-
distance-squared weighting and the two results were found to be essentially
indistinguishable. However, it is unwarranted to draw general conclusions
on the basis of this limited test.
92
-------
40
30
OBSERVED AT STATION 79
(WEST SAN GABRIEL VALLEY)
LJJ
O
1
O
o
t-J
O
20
10
/INTERPOLATED FROM STATIONS 1
(DOWNTOWN LOS ANGELES),
60 (EAST SAN GABRIEL VALLEY),
69 (EAST SAN FERNANDO VALLEY)
0600
0800
1000
1200
TIME, PST
Figure 5.2. Interpolated and Observed Ozone Concentration at a Monitoring
Station in the West San Gabriel Valley~on September 29, 1969
93
-------
20
OBSERVED AT STATION 79
(WEST SAN GABRIEL VALLEY)
, ^INTERPOLATED FROM STATIONS 1
(DOWNTOWN LOS ANGELES),
60 (EAST SAN GABRIEL VALLEY),
69 (EAST SAN FERNANDO VALLEY)
0600
0800
1000
TIME, PST
1200
Figure 5.3. Interpolated and Observed Ozone Concentration at a Monitoring
Station in the West San Gabriel Valley on November 4, 1969
5.3.4 Assessment of Model Performance
In this section we present a general assessment of the performance
of the model. The results for each trajectory are included in Sees. 5.3.5
through 5.3.28, where each case is illustrated by a map of the trajectory
and three plots of computed and observed concentrations: one for CO, one
for NO and N02> and one for ozone. It should be noted that results for
reactive hydrocarbon are not shown. This is due to the paucity of hydro^
carbon data available for comparison, there being only three hydrocarbon^
monitoring stations in Los Angeles County.
One of the most important factors which affect the performance 'of
the model is the accuracy of the initial concentrations of the three
fundamental species NO, N02> and reactive hydrocarbon.- The influence of
initial values is especially significant for a trajectory which starts
around 0600 or later and runs for eight hours or less. In such a case,
94
-------
the mass of the emitted pollutants Is considerably less than the initial
mass and as a consequence the computation is greatly influenced by the
initial values. Most of the trajectories studied in this report fall
in this category. One notable exception is trajectory number 8, which
starts at 0230 and has very low initial concentrations. Availability of
hydrocarbon measurements constrains the selection of a starting location
since there are fewer hydrocarbon-monitoring stations than there are for
NO or CO. Hence we attempted to reduce the uncertainties in the initial
X
conditions by starting trajectories at places of best-known hydrocarbon
levels. In connection with the problem of uncertainty in the initial
values, we note that generally we seemed to obtain better results with
trajectories which started at Commerce than with those which began in
Downtown Los Angeles. This could be attributed to the higher quality
of the Commerce data.
The plots of ozone concentration shown in subsequent sections
illustrate that the computations matched the observations remarkably
well. In order to achieve this, we had to settle for poorer fits for
NO and NO.. Usually, the model results for NO matched the data better
than the NO predictions fit the NO, measurements. (We recall that this
situation also prevailed, although to a smaller extent, in the simulation
of the smog chamber experiments.) In several cases, early morning peaks
of NO were difficult to reproduce. On the other hand, the NO buildup
was generally accurate, but the decay was poorly reproduced, with the
N0? tending to linger at relatively high concentrations late in the day.
This behavior of NO- may be due to inadequacies in the kinetics or mixing
model for late-time NO behavior. Addition of the reaction of NO with
particulates (reaction 2.16) Improved the late-time NO decay, but the
improvement was small. Increasing k , is not- the answer to this problem,
because this Interferes with late-time RO control, with the result
being anomalously high concentrations of R0? . Additional research Is
needed to improve the late-time behavior of N0?.
95
-------
In the simulation process, we had to reduce NO fluxes consistently
to 1/4 of the value estimated from source inventories. This adjustment
is consistent with our previous work. The necessity of scaling down the
NO emissions arises from the fact that both the NO balance and the ozone
X
production diverge greatly from the observed values when the full NO flux
is used; Reducing the NO flux to 1/4 of its full value results in pre-
dictions which fit the data much more accurately. As discussed previously,
these flux reductions may reflect atmospheric loss mechanisms such as
surface reactions that are as yet unidentified in any of the field programs.
Statistical Correlation of Computations and Observations
One way of evaluating the aggregate performance of the model is
to measure the correlation between computations and observations. This
contrasts with the case-by-case presentation of results contained in the
next several sections. By examining the relationship between predictions
and measurements in a highly aggregated form, it is easy to discern trends
in the model's performance. These trends may be useful in obtaining
correction factors to improve the predictions. Naturally, the initial
concentrations have been excluded from the statistical analysis since
their inclusion would bias the correlations.
Correlation coefficients for computed and observed concentrations
were obtained for CO and ozone for the set of hands-on cases, the set of
hands-off cases, and for both sets together. Table 5.4 shows the values
of the coefficients for each of the three groupings. It is clear from
TABLE 5.4
CORRELATION COEFFICIENTS FOR CO AND OZONE
Species Hands-on Hands-off Composite
CO 0.90 0.63 0.80
Ozone 0.94 0.88 0.92
96
-------
the table that the coefficient for the hands-on cases provides an upper
J!
bound of the expected performance of the'model, and the coefficient for
the hands-off cases, a lower bound. For ozone, the difference between
the upper and lower bounds is small, thus indicating a good performance
in either situation. The difference is greater for CO, however, indicating
a need for additional adjustment in diffusivity parameters in the hands-off
cases. It is interesting to note that the lowest correlation coefficient
obtained here for CO (in the hands-off cases) matches the highest corre-
lation coefficient obtained by other investigators using another CO
diffusion model which employs a combination of Gaussian plume and box
A 1 36
models.
Figures 5.4 and 5.5 are scatter diagrams of observed vs predicted
concentrations of CO and ozone, respectively. The results of all 24
trajectories are contained in these graphs. The figures also include
plots of the least-squares regression line.
30.00 -i
25.00 -
20.00 -
o
o
CO
CD
O
15.00 -
10.00 -
5.00 -
0.00
O.QO
5.00 10
.00 16.00 20.
COMPUTED CO tPPH)
20.00 26-00 33.00
Figure 5.4. Observed Versus Computed Carbon Monoxide Concentration.
(Number of Points = 149)
97
-------
10-00 -
35-00 -
30.00 -
| es.00-
Q_
O
g 90.00 -
0
UJ
Of.
UJ
CO
oa 15.00 -
o
10.00 -
5.00 -
0
0 ° t
0 % ffl y
oo f o
ff1 a/*o
I 00
o-oo -f r i 1 : 1 1—'- r~ -i r
0.00 5.00 10.00 15-00 20-00 25.00 30.00 36-00 40.00
COMPUTED OZONE (PPHMJ
Figure 5.5. Observed Versus Computed Ozone Concentration.
(Number of Points = 151)
It is apparent from Fig. 5.4 that for CO the model tends to over-
estimate the low concentrations; these generally occur in the afternoon.
The CO peaks, on the other hand, are underpredicted. For ozone, Fig. 5.5,
the model tends to underestimate the low concentrations but to be accurate
at medium levels. High ozone concentrations are slightly overpredicted.
The regression lines shown on Figs. 5.4 and 5.5 can be usec>t:o
correct the predictions and thus obtain a better estimate of the actual
concentration. Table 5.5 shows the equations of the regression lines for
CO and ozone, together with the standard error of estimate. If we ask
what the actual concentration, y , is likely to be given a predicted
concentration y. , we can obtain an answer from the regression equation.
The value of a , the standard error of estimate, provides a measure of
the tolerance which may be assigned to the corrected prediction.
98
-------
TABLE 5.5
REGRESSION EQUATIONS FOR CO AND OZONE
Regression Equation
y = observed Standard Error
Species x = computed of Estimate (a)
CO y = 1.007'x - 0.346 3.663 ppm
Ozone y = 0.840x + 2.307 2.109 pphm
2
Cy' - Y ) > y = observed concentration
y = concentration computed using regression
equation
99
-------
5.3.5 Trajectory No. 1, September 11. 1969. Starts at Commerce at 0530
(Hands-Off)
The simulation of CO agreed very well with the data. However, the
results for the reactive species matched the data poorly, with the NO
balance predicted by the model being greater than is shown by the data.
Also, Fig. 5.9 shows that the predicted ozone concentration, denoted by
the solid line, is considerably higher than is indicated by the data.
This computation was done using the clear-day value of k^ . However,
examination of Fig. 5.102 (Sec. 5.5) shows that for Sept. 11 the value
of k.. at Commerce is much lower than the clear-day value. Since the
trajectory meanders around Commerce for most of the day, it is possible
that the low values of ozone may be due to the low k.. . Substituting
the Commerce k.. for the clear-day k.. yielded the dashed curve shown
in Fig. 5.9. The lover k.. produced much lower values of ozone, but
the relatively small ozone peaks shown by the data were still not repro-
duced by the computation. Nevertheless, the dramatic reduction in ozone
obtained using .the new k1 is indicative of the large variations to
-' J_
which the predictions are subject due to the uncertainty in the fundamental
3 —1 —1
input k^ . Finally, in this case k, = 4 x 10 ppm min
•LACA 108W is
r nn ~rc\ i 0830
CAP 75W . ]030
79 WSGV
PASA new
Figure 5.6. September 11, 1969 Trajectory Starting at Commerce at 0530 (No. 1)
100
-------
MODEL RESULTS Łj
O INTERPOLATED 3
STATION DATA t
Figure 5.7- Trajectory No. 1—Computed and Observed CO Concentrations
MODEL RESULTS —
INTERPOLATED STATION - NO O
Figure 5.8. Trajectory No. 1—Computed and Observed NO and NO Concentrations
Figure 5.9. Trajectory No. 1—Computed and Observed Ozone Concentrations
101
-------
5.3.6 Trajectory No. 2, September 11, 1969, Starts^t^Conimerce at Q6.3Q
(Hands-Off)
Figure 5.11 shows that the early-morning CO buildup and decay are
well reproduced by the model. The low CO concentrations in the afternoon
are overestimated, however. (There is no data point at 1230 because it
was missing from the observations.) On the other hand, Figs. 5,12 and
5.13 show that the predicted concentrations of the reactive species match
the data very well. The computed NO balance is very good and the ozone
X 3-1 -1
prediction is superior. For this trajectory, k, = 4 * 10 ppm min
LACA 108 W
* 1330
tx
UQ
69 E SFV
1230
1130
60 E SSV
ELM
BKTD
1 CAP 75 W
0930
1030
0630 COM
0830 0730 .RVA 81 W
LA INTERNATIONAL
AIRPORT
KFID
Figure 5.10. September.il, 1969 Trajectory Starting at Commerce
at 0630 (No. 2)
102
-------
___ MODEL RESULTS
O INTERPOLATED
STATION DATA
1430
Figure 5.11. Trajectory No. 2—Computed and Observed CO Concentrations
MODEL RESULTS .
INTERPOLATED STATION - NO O
NO, A
Figure 5.12. Trajectory No. 2—Computed and Observed NO and NO Concentrations
—— MODEL RESULTS
O INTERPOLATED STATION DATA
Figure 5.13. Trajectory No. 2—Computed and Observed Ozone Concentrations
103
-------
5.3.7 Trajectory No. 3, September 11, 1969, Starts in Downtown Los Angeles
at 0530 (Hands-Off)
Figure 5.15 shows that the simulation of CO matched the data well.
However, the NO computation does not match the data, the peaks of NO
X i;
and NO- being much lower than the observations shown in Fig. 5.16. On
the other hand, the ozone simulation is remarkably accurate, despite the
poor quality of the NO predictions. This case is one example where
X
parameter adjustments made to improve the NO computations degrade
the ozone and vice versa. The value of k, used in this case is
3 -1 -1 4
4 x 10 ppm min
1230
74 W SFV»
86 W ENC
41 WBURK .108 W LACA
1130 ^V69 E SFV
• 79 W SGV
0730
27 W HOL« V0630
0530
1 DOLA
>89 COM
Figure 5.14. September 11, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 (No. 3)
104
-------
MODEL RESULTS
O INTERPOLATED STATION DATA
0930
TIME (PST)
1130
1330
Figure 5.15. Trajectory No. 3—Computed and Observed CO Concentrations
S 20
S 15
MODEL RESULTS
INTERPOLATED - NO O
STATION DATA .... A
g
0930
TIME (PST)
Figure 5.16. Trajectory No. 3—Computed and Observed NO and N0_ Concentrations
Figure 5.17- Trajectory No. 3—Computed and Observed Ozone Concentrations
105
-------
5.3.8 Trajectory No. 4, September 11, 1969, Starts in Downtown Los
Angeles at 0630 (Haads-Off)
The computed CO, Fig. 5.19, agreed with the data to within 3 ppm,
except for the last point which is about 5 ppm higher than the data. As
in the previous trajectory, the computed NO balance is lower than indi-
X
cated by the data, but this time the NO predicted by the model matches
the data well. Figure 5.21 shows that the computed ozone is generally
lower than the measurements, but the differences are certainly within the
error margin of interpolation and experimental uncertainty. The value
3 -1 -1
of k used in this simulation was k, = 4 x 10 ppm min
SAUD
i MISH113W
LACA 108W
69 E SFV
1030
0930 .79WSGV
0830
0730
• COM
BKTD
DKFI
Figure 5.18. September 11, 1969 Trajectory Starting in Downtown Los Angeles
at 0630 (No. 4)
106
-------
20
i 15
MODEL RESULTS
O INTERPOLATED STATION DATA
1030
TIME (PST)
Figure 5.19. Trajectory No. 4—Computed and Observed CO Concentration
MODEL RESULTS ' ' ' '
INTERPOLATED - NO
STATION DATA
Figure 5i20j Trajectory No. 4—Computed arid Observed NO and N0? Concehtratitins
_ i's -
MODEL RESULTS
O INTERPOLATED STATION DATA
Figure 5*21, Trajectory No1. 4—Computed and Observed Ozone Concentrations
107
-------
5.3.9 Trajectory No. 5, September 29, 1969, Starting at Commerce at
0530 (Hands-On)
This trajectory has the characteristic that it stays in the vicinity
of Commerce most of the time from 0530 to 1030. This gives us a greater
degree of confidence in the data obtained by interpolation from the various
stations along the way.
The CO simulation shows a low peak, but the remaining concentrations
are predicted accurately. The simulation of the reactive species shows
good agreement with the data, especially for the NO and ozone. The NO
buildup is accurately predicted, but the NO decay fits the data poorly.
3 -1 -1
The value .of k. used was 8 x 10 ppm min
79 WSGV
1230,
1130/»ELM
60 ESGV
COM
0730*f»RVA
0830^0930
•
RLA
®80 SE
DKF1
Figure 5.22. September 29, 1969 Trajectory Starting at Commerce at 0530 (No. 5)
108
-------
O INTERPOLATED STATION DATA
Ol
0530
0730 0930
TIME (PST)
Figure 5.23. Trajectory No. 5—Computed and Observed CO Concentrations
MODEL RESULTS
INTERPOLATED STATION DATA D NO
A NO,
Figure 5.24. Trajectory No. 5—Computed and Observed NO and N0? Concentrations
MODEL RESULTS
INTERPOLATED STATION DATA O
Figure 5.25. Trajectory No. 5—Computed and Observed Ozone Concentrations
109
-------
5.3.10 Trajectory No. 6, September 29, 1969, Starting at Commerce at
0630 (Hands-On) ~~ --_ -
This trajectory is similar to the previous one. This time, however,
the computed CO buildup and peak match the data closely. The predicted
CO decay is accurate until 1030. The model overprediets the 1130 and
1230 CO concentrations but at 1330 the approximation is good.
The reactive species, Figs. 5,28 and 5.29, show a, behavior similar
to that in trajectory no. 5: the NO .and ozone curves are very well
modeled, but for N0~ only the buildup is close to the data. The value
~ -i i
of k. was 8 x 1CH ppm min
79 WSGV*
BOESGVi
ZUM
0830
1330
DBKT
DOLA
COM
VER« \— An 30
O730t XRVA
1030
50
•=C
0930 »80 SE WHTR
n KFI
• 72 SC'LONB
Figure 5.26. September 29, 1969 Trajectory Starting at Commerce
at 0630 (No. 6)
110
-------
MODEL RESULTS TTT-
INTERPOLATED STATION DATA O
0830 1030
TIME (PST)
1230
Figure 5.27. Trajectory No. 6—Computed and Observed GO Concentrations
q NO
TIME (PST)
Figu.r-e 5,. 28. Trajectory No. 6--Computed and Observed NO and N0_ Concentrations
..... ^ ^- - . _ - 2-
TIME (PST)
1230
5,.29;. Trajectq.ry No. 6--Computed a,nd Observed O.zone Concentrations
111
-------
5.3.11 Trajectory No, 71_Segjteinber_29J 1969, Starting in Downtown Los
Angeles at 0530 (Hands-on)
The results obtained for CO are shown in Fig. 5.31. The reported
0530 concentration at the downtown station was 3 ppm. Using this initial
value yielded very low concentrations throughout the whole day. In order
to obtain a better approximation to the data, the initial value was
adjusted to 7 ppm and this yielded improved results. However, the pre-
dicted peak value of 12 ppm was still 4 ppm lower than the data indicate.
Figure 5.32 shows that the predicted NO balance diverges con-
siderably from that shown by the data. In spite of this, the simulation
of ozone was very accurate. We might add that no adjustment of the initial
concentrations was necessary for simulating the reactive species. The
value of k, was 5 x 10 ppm min
60 ESGV
m
>ELM
0530f DOLA
•80 SE
75 EGA
• WNTT
10
Ki
IX
Figure 5.30. September 29, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 (No. 7)
112
-------
s '"
i
MODEL RESULTS ——
INTERPOLATED STATION DATA O
0730 0930
TIME (PST)
J 1 I I
Figure 5.31. Trajectory No. 1—Computed and
Observed CO Concentrations
MODEL RESULTS
INTERPOLATED STUTIOK DATA AHO,
Figure 5.32. Trajectory No. 7—Computed and
Observed NO and NO Concentrations
Figure 5.33. Trajectory No. 7—Computed and
Observed Ozone Concentrations
-------
5.3.12 Trajectory No. 8, September 29, 19,69, Starting Near Coast at 0230
(Hands-on)
This trajectory was reverse-calculated from the Anaheim station at
1230. The computation of pollution concentrations commenced at 0230 near
the Lennox station of the LAAPCD. As was the case with the previous tra-
jectory, we could not fit the CO data using the reported initial concen-
tration of 3 ppm. Adjustment of the initial CO to 7 ppm produced' satis-
factory results from 0530 to 1130, as is shown in Fig. 5.35.
As with CO , the initial values of NO and hydrocarbon were
adjusted upward by the ratio 7/3. However, the original initial values
were very low and the adjustment did not cause large changes in the simu-
lation. Figure 5.36 shows that the NO simulation does not agree well
with the data. It should be noted that the NO data show a second peak
at 0930 which is suspect. The computed NO. , on the other hand, shows
close agreement with the data after 0630. For ozone, Fig. 5.37 shows
that the simulation and the data are very closely correlated. The value
4 -1 -1
used for k, was 10 ppm min
80 SE
fe;
1230
ANAHEIM
1030 1130
ORANGE CO.
AIRPORT
Figure 5.34. September 29, 1969 Trajectory Starting Near the Coast
at 0230 (No. 8)
114
-------
S ,0
s
0 O
MODEL RESULTS
INTERPOLATED STATION DATA O
0630 1030
TINE (PST)
Figure 5.35. Trajectory No. 8—Computed and Observed CO Concentrations
MODEL RESULTS
INTERPOLATED STATION DATA O NO
AND,
0230 0430
0630 0830
TIME (PST)
1030 1230
Figure 5.36. Trajectory No. 8—Computed and Observed NO and N0_ Concentrations
MODEL RESULTS
INTERPOLATED STATION DATA O
SUNRISf
0230. 0430 0630 : 0830 1030 1230
TIME, PST
Figure 5.37. Trajectory No. 8—Computed and Observed Ozone Concentrations
115
-------
5.3.13 Trajectory No. 9, September 30, 1969, Starting at Commerce at 0530
(Hands-on)
Figure 5.39 shows that the predicted CO fits the data well, with
the exception of the data point at 0930. In addition, the computed con-
centration at 0630 is low by about 6 ppm. Figure 5.40 illustrates that
the predicted NO matches the data closely but that the N02 is con-
siderably overestimated. This situation is typical of the unsatisfactory
NO balances encountered in the simulations. The computed ozone (Fig.
5.41) is seen to underestimate the early-morning concentrations and then
overshoot the 1130 value by 8 pphm. We note that the relatively high
ozone concentrations indicated by the observations before 1030 are puzzling
inasmuch as very little NO -> NO conversion has taken place prior to
/ ^ -i __ -I
1030. The value of k is 10 ppm min
SAUD
• MISH
1130
BUR*
BURK«
• LACA
r!030
.0930 «PASA
HOL«
WEST
DOLA
0730
0530 COM
LAX<
LENX
BKTD
KFIO
Figure 5.38. September 30, 1969 Trajectory Starting at Commerce
at 0530 (No. 9)
116
-------
30r
i\- MODEL RESULTS
O INTERPOLATED STATION DATA
0530
Figure 5,39.
TIME (PST)
0930
1130
Trajectory No. 9—Computed and
Observed CO Concentration
MODEL RESULTS
O INTERPOLATED STATION DATA
0730 0930
TIME (PST)
1130
Figure 5.40. Trajectory No. 9—Computed and
Observed NO and N02 Concentrations
-Figure 5.41. Trajectory No. 9—Computed and
Observed Ozone Concentrations
TIME (PST)
-------
5.3.14 Trajectory No. 10, September 30, 1969, Starting at Commerce at
0630 (Hands-on) \ ;
i
This trajectory follows approximately the same'path as trajectory
No. 9. However, this time the early-morning values of CO are reproduced
very well, as can be seen in Fig. 5.43, As was the case with the previous
trajectory, the model is unable to follow the sharp drop in CO concen-
tration which occurs- at 0930. For NO it can be seen in Fig. 5.44 that
the peak is too high but the decay is fairly accurate. The NO buildup
is well reproduced, but the decay is very poor. For ozone, Fig. 5.45, it
can be seen that once again the low concentrations are underestimated.
In contrast with the previous case, the model is very accurate at the
4 -1 -1
higher ozone levels. The value of k. is 10 ppm min
1130
RESD
•
69 E SFV«
09 30 \
HOL 27 W«
to
LO
tx
fo
I
LACA
79 W SGV
PASA
08303
DOLA<
0730' ^« 0630 COM
34 W
Figure 5.42. September 30, 1969 Trajectory Starting at Commerce
at 0630 (No. 10)
118
-------
MODEL RSSULTS
INTERPOLATED STATION DATA O
TIME (PST)
Figure 5.43.
Trajectory No. 10—Computed
and Observed CO Concentrations
MODEL RESULTS
O INTERPOLATED STATION DATA
MODEL RESULTS
INTERPOLATED STATION - NO O
DATA „„ A
TIME (PST)
Figure 5.44. Trajectory No. 10—Computed and
Observed NO and N0? Concentrations
Figure 5.45. Trajectory No. 10—Computed and
Observed Ozone Concentrations
TIME (PST)
-------
5.3.15 Trajectory No. 11, September 30, 1969, Starting in Downtown Los
Angeles at 0430 (Hands-On)
Figure 5.47 illustrates that the CO peak is about 6 ppm too low.
This may indicate a deficiency in the strength of the emissions since the
diffusivities were set at the lowest value used for very stable atmos-
3 2
pheric conditions, 2.5 * 10 cm /second . The model is accurate from
0730 to 0930, but does not reproduce the apparent peak after 0930. The
reactive species are well modeled. This time the simulated NO balance
is excellent wi.th the exception of one very high observed concentration
of NO at 1130. The ozone is only slightly underestimated throughout
X 4-1-1
the trajectory. The value of k, used in this case is 10 ppm min
69 E SFV»
71 NWC
1130 »LACA
00
'79 U SGV
0830
LAX 0730
0430 DOLA
0530 . »COM
VER
Figure 5.46. ^^30,^1969 Trajectory Starting in Downtown Los Angeles
120
-------
MODEL RESULTS
O INTERPOLATED STATION DATA
0830
TIME (PST)
Figure 5.47. Trajectory No. 11—Computed and Observed CO Concentrations
MODEL RESULTS
INTERPOLATED STATION - NO O
DATA N02 A
Figure 5.48. Trajectory No. 11—Computed and Observed NO and N0_ Concentrations
o
—— MODEL RESULTS
O INTERPOLATED STATION OATA
Figure 5.49. Trajectory No. 11—Computed and Observed Ozone Concentrations
121
-------
5.3.16 Trajectory No. 12, September 30, 1969, Starting in Downtown Los
Angeles at 0530 (Hands-On)
The CO simulation shown on Fig. 5.51 indicates good agreement
between prediction and observation, with the lower concentrations being
overpredicted. On Fig. 5.52, it can be seen that the predicted NO con-
centrations are low but that the NO computations match the observations
accurately. The ozone, Fig. 5.53, is underestimated throughout most of
the morning, but the error is small. It should be noted that the observed
ozone concentrations which occur from 0730 to 0930 are difficult to recon-
cile with the high NO concentrations indicated by the data. For k. ,
4-1-1
the value used was 10 ppm min
SAUD
RESD
BUR*
69 E SFV
HOL 27 W»
•1030 »LACA
t0930
0830
0730
0630
0530
DOLA
• COM
•79WSGV DASA
KFID
BKTQ
Figure 5.50. September 30, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 (No. 12)
122
-------
— MODEL RESULTS
Q INTERPOLATED STATION
DATA :
0730 0930
TIME (PST)
Figure 5.51. Trajectory No. 12—Computed and Observed CO Concentrations
MODEL RESULTS
INTERPOLATED STATION - NO O
DATA NO, A
1130
TIME (PST)
Figure 5.52. .Trajectory No. 12—Computed and Observed NO and NO- Concentrations
0730 0930
TIME (PST)
Figure 5.53. Trajectory No. 12—Computed and Observed Ozone Concentrations
123
-------
5.3.17 Trajectory No. 13, October 29, 1969, Starting in Downtown Los
Angeles at 0530 (Hands-On)
Figure 5.55 shows that the simulated CO approximates the data
closely. However, from Fig. 5.56 we can see that the NO simulation
diverges considerably from the data. Despite the low quality of the NO
simulation, the computed ozone fits the data closely. This is, of course,
the result of our emphasis on obtaining accurate ozone predictions. For
this trajectoryj the value of k used was 4 x 10 ppm min
'60 ESGV
2
ELM
D
71 NWC
0530
1 DOLA
,0930
Figure 5.54. .October 29, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 (No. 13)
124
-------
—- MODEL RESULTS
O INTERPOLATED STATION
DATA
0530
0730 - 0930
TIME (PST)
Figure 5.55. Trajectory No. 13 — Computed and Observed CO Concentrations
MODEL RESULTS -
INTERPOLATED STATION NO O
DATA N0 A
0930
TIME (PST)
Figure 5.56. Trajectory No. 13—Computed and Observed NO and NO Concentrations
o
1130
1330
Figure 5.57.
0930
TIME (PST)
Trajectory No. 13—Computed and Observed Ozone Concentrations
125
-------
5.3.18 Trajectory No. 14, October 29, 1969, Starting in Downtown Los
Angeles at 0630 (Hand's-On)
From Fig. 5.59 it can be seen that the CO simulation fits the
data fairly closely. The CO peak produced by the model is low by 3 ppm.
The early-afternoon concentrations of CO are overpredicted.
is reproduced very accurately. On
NO,, peak
Figure 5.60 shows that the NO
the other hand, the simulated ">-> v»j.v«_i.gwo j-j-^m ^n^ uu>.u, !_,.!,_ '••^o
being low by about 4 pphm and occurring an hour later than indicated by
the data. Figure 5.61 shows that the modeling of ozone is remarkably
NO diverges from the data, the
accurate. For this trajectory, the value of k. used was
1fl4 -1 . -1
10 ppm mm
DOLA
0630
• ELM
DBKT
•POMA
230 1330
i VINTt
01
!~H
to
fe:
Figure 5.58. October 29, 1969 Trajectory Starting at 0630 in Downtown
Los-Angeles (No. 14)
126
-------
IS3
—I
18r O
MODEL RESULTS
INTERPOLATED STATION DATA O
0630 0730 0830
0930 1030
TIME (PST)
1130 1230 1330
Figure 5.59. Trajectory No. 14—Computed and
Observed CO Concentrations
0830 1030
TINE (PST)
Figure 5.60. Trajectory No. 14—Computed and
Observed NO and N0_ Concentrations
Figure 5.61. Trajectory No. 14—Computed and
Observed Ozone Concentrations
0830 1030
TIME (PST)
-------
5.3.19 Trajectory No. 15, October 29, 1969, Starting at Commerce at 0630
(Hands-On)
The CO simulation shown in Fig. 5.63 is especially interesting be-
cause the double peak shown by the data is reproduced by the model. The
maximum difference between data and simulation occurs near the end of the
trajectory and is approximately 2.5 ppm. This could be due to greater
dispersion near the mountains.
The photochemical simulations (Figs. 5.64 and 5.65) show very good
modeling of NO and 0 . The simulated N0? peak is about 3 pphm too
low and occurs one hour later than is shown by the data. Inaccuracies in
the NO emissions inventory may account for this. The computed end value
of NO is 21 pphm compared to about 7 pphm for the data. The high ter-
minal values of NO appear to be a recurring problem in atmospheric
photochemical simulations. It is likely that this problem arises from
departures from expected photo-chemical equilibrium conditions (see
Appendix, Sec. A.2.2). The value of k. used in this trajectory was
10 ppm min
• PASA
0730
DOLA«
0630
VER»COM
0930
• ELM
>AZU
1330
DBKT
• POMA
>WNTT
to
CO
Figure 5.62. October 29, 1969 Trajectory Starting at Commerce
at 0630 (No. 15)
128
-------
Figure 5.63. Trajectory No. 15—Computed
and Observed CO
Concentrations
— 25
S
5 15
MODEL RESULTS •
INTERPOLATED STATION DATA OHO
1030
TIME (PST)
Figure 5.64. Trajectory No. 15—Compute
and Observed NO and N0?
Concentrations
Figure 5.65. Trajectory No. 15—Computed and Observed Ozone Concentrations
129
-------
5.3.20 Trajectory No. 16, October 29,- 1969, Starting at El MQnte at 0830
This trajectory shows a reversal in direction of almost 180° at 1030,
This may be seen in Fig. 5.66. The high value obtained from the data at
1030 is mostly due to high concentrations reported at the downtown station
(1DOLA). In view of the reversal in direction, it is likely that this
interpolated quantity is badly off the mark.
The high values shown by the data at the end of the trajectory could
not be reproduced by the simulation since the trajectory enters some areas
with no sources at all. Location of the sampling sites can induce large
deviations from the average for the air mass. The ground value will con-
tinue to decrease due to diffusion when source strengths are small.
Figure 5.68 shows that the simulation of NO is very accuratej but
that for N0? the model produced very poor results. As usual, the ozone
simulation, shown in Fig. 5.69, fits the data closely. The value used
for k was 10 ppm min
PASA A^U
1030><-^ 1130
ELM_ ^X<1330 BKT
POMA
0930, ™°Xf23o"""' °
to
to
I
Figure 5;66. October 29, 1969 Trajectory Starting at El Monte
at 0830 (No. 16)
130
-------
M8DEL RESULTS-
INTERPOLATED STATION DATA O
1230
TIME (PST)
Figure 5.67. Trajectory No. 16—Computed
and Observed CO Concentrations
0830
MODEL RESULTS -^——
INTERPOLATED STATION - NO O
OAT* „„. „.
1030 1230
TIKE (PST)
Figure 5.68.
Trajectory No. 16—Computed
and Observed NO and N02
Concentrations
— MODEL RESULTS
O INTERPOLATED STATION DATA
0830
Figure 5.69.
TIKE (PST)
Trajectory No. 16—Computed
and Observed Ozone Concentrations
-------
5.3.21 Trajectory No. 17, October 30. 1969. Starting at Pasadena at 0530
(Hands-Off)
The trajectory was started at 0730 southeast of Pasadena because all
of the CO data for Pasadena are missing for this date. As a result, the
initial values used for the various species were obtained by interpolation
from neighboring stations.
The results for CO show good reproduction of the buildup phase
from 0730 to 0930. The decay part of the concentration curve has the
right shape, but the concentrations are about 3 ppm too high. The low
points in mid-morning measurements could be due to horizontal intrusions
of air directly from the ocean.
The photochemical results show good NO and 0 simulations. The
NO buildup is well modeled, but the
NO. decay shows a very poor fit
NO interpolated values may be
of the data. The late morning dip of
due to the same dilution mechanisms suggested above for the CO points.
The value used for k.
was 10 ppm min
79
0530 .PASADENA • 60
0630^
• EL MONTE
DBKT
LA COUNTY (_f—'
/
l» KVA ^
10930 WHTR I—* !
RLA« \ • j LA HABRA
COMPTON •
ANAHEIM
1330
ORANGE COUNTY
AIRPORT
to
to
Figure 5.70. October 30, 1969 Trajectory Starting at Pasadena
at 0530 (No. 17)
132
-------
UJ
u>
0730
0930 1130
TIME (PST)
Figure 5.71. Trajectory No. 17—Computed
and Observed CO Concentrations
-MODEL RESULTS
N02 A INTERPOLATED STATION DATA
NO O
0930 1130
TIME (PST)
Figure 5.72.
Trajectory No. 17—Computed
and Observed NO and N0?
Concentrations
INTERPOLATED STATION DATA
MODEL RESULTS
0930 1130
TIME (PST)
Figure 5.73.
Trajectory No. 17—Computed
and Observed Ozone
Concentrations
-------
5.3.22 Trajectory No. 18, October 30, 1969, Starting at Commerce at 0630
(Hands-Off) "" """ "
This trajectory is rather short, only 4 hours long, and is interest-
ing because it travels toward the coast early in-the morning.
Figure 5.75 illustrates the simulation of CO , The time phasing
and shape of the curve agree with the data, but the predicted CO peak
is about 2.5 ppm lower than the data. It is noted that there were not
•available data for 0930.
It can be seen from Fig. 5.76 that the behavior of NO is accurately
simulated. The NO buildup is also accurate, but the predicted peak is
too low. This trajectory is one of the few examples in which the computed
ozone does not fit the data. However, it should be noted that the ozone
levels are very low and that the maximum absolute difference is only 4
pphm, although the relative error is considerably greater. The value
4 -1 -1
used for k was 10 ppm min
'79 W SGV
tvj
ELM a
Figure 5.74. October 30, 1969 Trajectory Starting at Commerce
at 0630 (No, 18)
134
-------
MODEL RESULTS •
Figure 5.75. Trajectory No. 18—Computed
and Observed CO
Concentrations
INTERPOLATED STATION - NO O
DATA „,, A
Figure 5.76. Trajectory No. 18—Compute-
and Observed NO and NO
Concentrations
— MODEL RESULTS
O INTERPOLATED STATION
DATA
Figure 5.77. Trajectory No. 18—Computed and Observed Ozone Concentrations
135
-------
5.3.23 Trajectory No. 19, October 30, 1969, Starting at El Monte at
0630 (Hands-off)
The results of the simulation for CO are shown in Fig. 5.79. It
can be seen that the model overestimates CO throughout the day. However,
the maximum difference between model and data is about 2.5 ppm.
As is generally the case with the reactive species, the NO and
ozone predictions are very accurate, as is shown in Figs. 5.80 and 5.81,
respectively. The NO balance is poor after 1030, however. The N0?
simulation exceeds the data after 1030. The low N02 data at 1130 and
1230 are suspect inasmuch as NO -> N0_ conversion continues during this
4 -1 -1
interval. The magnitude of k, was 10 ppm min
79 W SGV
ELM
0630
0730
80 SE
1430
Figure 5.78. October 30, 1969 Trajectory Starting at El Monte
at 0630 (No. 19)
136
-------
-MODEL RESULTS
O INTERPOLATED STATION DATA
1230
TIME (PST)
1030 1230
TIME (PST)
Figure 5.79. Trajectory No. 19—Computed Figure 5.8Q.. Trajectory No. 19—Computed
and Observed CO and Observed NO and NO
Concentrations Concentrations
MODEL RESULTS
O INTERPOLATED STATION
OATA
1030
1230
TIME (PST)
Figure 5.81. Trajectory No. 19—Computed and Observed Ozone Concentrations
137
-------
5.3.24 Trajectory No. 20, October 3.0, 1969, Starting in Downtown Los
Angeles at 0830 (Hands-off)
The shape of the CO simulation depicted on Fig. 5.83 does not match
the data. Nevertheless, the maximum difference between model and data
is 2 ppm. The increase in CO concentration after 1030 is due tq the
filling up of the air parcel, thus reducing the vertical concentration
gradient and therefore the vertical diffusion.
For
NO , the simulation results are inaccurate, as can be seen in
X
Fig. 5.84. However, in Fig. 5.85 the simulated ozone shows once again a
very close fit to the data, despite the low quality of the NO simula-
tion. We note that it is difficult to believe that so much ozone could
coexist with the NO indicated by the data. Finally, the value of k,
3-1-1 •
used was 6 x 10 ppm min
to
1 DOLA
0830
COM
Figure 5.82. October 30, 1969 Trajectory Starting in Downtown Los Angeles
at 0830 (No. 20)
138
-------
10
0830
.MPPEL RESULTS
INTERPOLATED STATION
DATA
I i
1030
TIME (PST)
1230
MODEL RESULTS
INTERPOLATED STATION NO O
U DATA NQ A
Figure fpg3,
Trajectory No. 20^-Computed
and Observed CO
Concentrations
Figure 5.84.
1030
TIME (PST)
1230
Trajectory No. 20—Computed
and Observed NO and N0_
Concentrations
10
2 -
0830
__ MODEL RESULTS
O INTERPOLATED STATION
DATA
1030
TIME (PST)
1230
"Figure 5.85, Trajectory No, 20--Computed and Observed Ozone Concentrations
139
-------
5.3.25 Trajectory No. 21, November 4, 1969, Starting at Commerce at
0530 (Hands-off)
This trajectory exhibits a reversal in direction at 0730 and this
may introduce inaccuracies in the interpolation process used to calculate
the concentrations along the path of the trajectory.
As can be seen from Fig. 5.87, the early-morning CO buildup is
underestimated by the model, with the relative error at the CO peak being
about 21%. The sharp decay and subsequent increase in concentration
seen after 0930 are not properly simulated by the model because the CO
emissions are increasing during this part of the trajectory.
The simulation of NO illustrated in Fig. 5.88 shows that NO
X
is underpredicted but that N0_ fits the data accurately. The relatively
high concentrations of ozone found in the data (Fig. 5.89) are suspect
because the data show high concentrations of NO present throughout the
trajectory. Nevertheless, the ozone simulation is accurate, especially
at the higher levels. In this trajectory the value assigned to k. was
3-1-1
5 x 10 ppm min
PASA* «AZU
1130tr, « BKTD
1 DOLA 0930
°83ii^^3ulo7^^30
• RVA
Figure 5.86. November 4, 1969 Trajectory Starting at Commerce
at 053;0 (No. 21)
140
-------
CO
MODEL RESULTS
O INTERPOLATED STATION
DATA
j L ,_
0730 0930
TIME (PST)
Figure 5.87. Trajectory No. 21—Computed
and Observed CO Concentrations
MODEL RESULTS
O INTERPOLATED STATION
DATA
MODEL RESULTS -
INTERPOLATED STATION - NO O
DATA
N02 A
0730 0930
TIME (PST)
1130
Figure 5.88. Trajectory No. 21—Computed
and Observed NO and N02
Concentrations
0530
0730 0'J30
TIME (PST)
Figure 5.89. Trajectory No. 21—Computed
and Observed Ozone Concentrations
-------
5.3.26 Trajectory No. 22, November 4, 1969, Starting at Commerce at
0630 (Hands-off)
This trajectory also exhibits a reversal in direction at 0730
similar to that found in trajectory No. 21. Thus the caveats mentioned
previously regarding the accuracy of the interpolation also apply here.
In contrast with the previous case, Fig. 5.91 shows that the CO
buildup is reproduced accurately by the model. However, the CO decay is
greatly overestimated by the computation. The CO emissions are high until
0900, at which time they drop to about one-half of the 0900 value.
Figure 5.92 shows that the initial NO level is very high and from
Table 5.3 we can see that the initial hydrocarbon concentration is also
very high (130 pphm). This hydrocarbon-NO combination portends high N0?
levels and this is precisely what the simulation produces, as can be seen
in Fig. 5.92. The ozone simulation is generally accurate, however, although
the ozone peak is overestimated by about 28%. For this trajectory, we
4 -1 -1
used a value of k, equal to 10 ppm min
LACA §
U3
c^
t-O
1130 a:
PASA
S-1030
D BKT
QKFI
Figure 5.90. November 4, 1969 Trajectory Starting at Commerce
at 0630 (No. 22)
142
-------
0830 1030
TIME (PST)
Figure 5.91. Trajectory 22—Computed and
Observed CO Concentrations
MODEL RESULTS .
INTERPOLATED STATION - NO O
DATA
N02 A
0830 1030
TIME (PST)
Figure 5.92.
Trajectory No. 22—Computed
and Observed NO and N0~
Concentrations
Figure 5.93. Trajectory No. 22—Computed and
Observed Ozone Concentrations
TIME (PST)
-------
5.3.27 Trajectory No. 23, November 4, 1969. Starting in. Pasadena at
0530 (Hands-off)
Figure 5.95 shows that the model underestimates CO concentrations
throughout the trajectory. In this case, the strength of the emissions
was too low to produce the high values indicated by the data even under
highly stable meteorological conditions. Upon investigating the data, it
became apparent that the high concentrations from 0630 to 0830 are due to
the downtown and Commerce stations which, in view of the path of the
trajectory, makes these interpolated concentrations suspect. The sus-
picion is heightened upon noting that at 0930, when the air parcel is again
close to Pasadena, the predicted and observed concentrations are closely.
matched.
Figure.5.9o shows that the computed NO balance is very poor, with
X
the predicted NO and NO- considerably below their apparent observed values.
As can be seen in Fig. 5.97, the computed ozone accurately reproduces the
observations-,- the largest deviation occurring at 0930. This is puzzling
in view of the results obtained for CO at 0930, but with chemical
processes at work, it is not surprising. The value used for k, was
3-1-1 4
4 x 10 ppm min
LACAt
• BURK
• HOL
-_ .^_ *
^ " ~\ 1 DOLA
. ,1130 g
\1030 7
I PASA I
0930|^n530
loeV .ELM DBKT
0830*
-------
MODEL RESULTS
O INTERPOLATED STATION
DATA
TIME (PST)
Figure 5.95. Trajectory No. 23—Computed and Observed CO Concentrations
_. 40
§
.jl»
-------
5.3.28 Trajectory No. 24, November 4, 1969, Starting in Downtown Los
Angeles at 0530 (Hands-off)
The simulation of CO yielded the results shown in Fig. 5.99. The
early-morning buildup of CO is accurately simulated. Once again, however,
the lower concentrations-of CO are overpredicted by the model.
NO balance is relatively
X
In Fig. 5.100, it can be seen that the
accurate until 0930, although the NO peak is underestimated. After 0930,
the computed NO. remains too high, while the NO fits the data correctly
For ozone, we see in Fig. 5.101 that the predicted ozone buildup until
1230 is accurate, but after 1230 the model exceeds the data by a maximum
4 -1 -1
of 10 pphm. The value of k. used was 10 ppm min
DSAU
• NEWHALL
ZUM
CANOGA
PARK
MISSION HILLS
1430
RESEDA
•
ENCINO*
L
HI
N
.ES
0 5 10 1
,EL MONTE
BKT
Q
Figure 5.98. November 4, 1969 Trajectory Starting in Downtown Los Angeles
at 0530 (No. 24)
146
-------
MODEL RESULTS
INTERPOLATED STATION DAtt O
Figure 5.99. Trajectory No. 24—Computed and
Observed CO Concentrations
o
0430
MODEL RESULTS _____
INTERPOLATED STATION O NO
0630 0830 1030
TIME (PST)
1230 1430
Figure 5.100. Trajectory No. 24—Computed
and Observed NO and NO '
Concentrations
D -INTERPOLATED STATION DATA
MODEL RESULTS
0430 • 0630 0830 1030
TIME (PST)
Figure 5.101. Trajectory No. 24—Computed and Observed Ozone Concentrations
-------
5.4 TECHNIQUES FOR MODEL OPERATION
In this section, we describe several basic features of the atmospheric
model to assist the prospective model user.
5.4.1 Kinetic Model for Atmospheric Simulation
The chemical model used in atmospheric simulation is basic'ally the
same one used: to model the smog chamber experiment using a dilute auto
exhaust mixture from a vehicle with exhaust hydrocarbon and carbon
monoxide emission controls (experiment 231; cf. pp. 37-40 and p. 43).
This kinetic model was chosen for atmospheric applications because the
smog's kinetics are not likely to be influenced by CO because of its concen-
trations in experiment 231 and in the atmosphere (cf. p.-36).
The branching factors are features of the smog chamber model which
have been retained on moving to atmospheric simulation. Thus we have
b = b2 = 8 , and b_ = 1 . The OH yield factor, y , remains equal to
1/8.
The kinetic model for the atmosphere differs from the smog .chamber
model mentioned above in that a. single hydrocarbon class is used rather
than two. The rate constants for the basic atmospheric model are shown
in Table 5.6. The constants for reactions 3 and 5 shown in Table 5.6 are
rounded mole-weighted averages of the two reaction pairs (3, 3a) and
(5, 5a) shown in Table 2.7. The first-order constant for reaction 16
shown in Table 2.7 was halved in the process of parameter adjustment
required for atmospheric simulation. (See p. 95 for a discussion of
the model's sensitivity to k, ,.)
16
The value of the rate constant of the first reaction depends on the
intensity of ultraviolet light in the wavelength range 2900-3850 A. The
magnitude of ^ is treated in the simulation as a function of solar
zenith angle, and hence of time of day. See p. 150 for an explanation
of the derivation of k^ and the sources of uncertainty associated with
the value of this rate constant.
148
-------
TABLE 5.6
RATE CONSTANTS USED IN ATMOSPHERIC MODELING STUDIES
*
Reaction No. Rate Constant
la 1.32 (-5) ppnf2 min'1
2 2.67(+l)
3 2. 76 (+2)
4 ***
5 4.0(-3)
6 1.0(4-5)
7 2.0(+2)
8 1.5 (+3)
9 3.0(4-3)
10 t
11 1.0(-3)
12 5.0(-3)
13 4. 5 (+3)
14 1.4(4-1) min"1
15 6.05(4-1) min"1
16 1.0 (-3) min"1
* -1-1
Units are ppm min unless otherwise specified.
**
This rate constant depends on sunlight intensity and in the simulation
it is treated as a function of solar zenith angle, and hence time of
day.
***
The rate constant for this reaction is given in the section which
describes each of the cases tested.
This reaction is the photodissociation of HONO and its rate constant
is obtained from the relation k1Q = 3.75 * 10~3 k^.
149
-------
The rate constant for the photodissociation of HONO (reaction 10)
_3
is obtained from ^ using the equation k1Q = 3.75 * 10 ^ . This
relation results from the assumption that the ratio ^g^l "^S a constant
equal to the ratio of the quantities used in the smog chamber simulations.
In atmospheric modeling, k1Q is thus a function of time.
It became necessary to adjust the rate constant of reaction 4
(OH + HC ->• (b?)R09) during the simulation process conducted under atmospheric
conditions. For this reason, the value of k, is given in the sections
describing the results of the simulations.
5.4.2 Chemical Inputs
Two inputs are basic to the operation of the chemical model. These
are k , the rate constant of the reaction hv + N02 •> NO + 0 , and
i . . the rate constant of OH + HC ->- (b0)R00.
H / Z
The value of k.. is of course a function of the intensity of ultra-
violet light with wavelength in the range 2900-3850A. The ultraviolet
intensity depends on time of day, time of year, geographical location,
and weather conditions. In our simulations, we have used values of k
which correspond to clear-day conditions, i.e., no overcast. Thus the
magnitude of k used is an upper bound of the actual value. This can
be seen in Figs. 5.102-5.107 in Sec. 5.5.
The derivation of the clear-day value of k is accomplished by
means of the relationship between k and solar zenith angle given by
•I Q i
Leighton. The solar zenith angle is a function of time of day, time
of year, and geographical location. Using solar ephemeris tables, a
table of solar zenith angles is generated for the specific times and
places of interest. Given the solar zenith angle, the value of k is
then determined. In our model, we update k at 10-minute intervals,
but this is arbitrary and can be changed to match the integration .step
size.
150
-------
The value of k, is determined by the concentrations of NO, NO ,
cl •Ł•
and reactive hydrocarbon at the start of the trajectory or at sunrise,
whichever is later. Thus k, is dependent on the initial mixture of
pollutants. We recall that this is in agreement with findings in the
validation of the chemical model using smog chamber data for the more
reactive mixtures such as propylene/NO and auto exhaust/NO . During
X 3 X 4 -1 -1
the atmospheric simulations, k, ranged from 4 x 10 to 10 ppm min
The rules for determining k, will not be exact because of uncertainties
in the initial concentrations. The variability of emissions from case
to case is another factor which affects the choice of k, . Thus a plot
of k, on a graph with ordinate equal to NO concentration and abscissa
T1 X
equal to reactive hydrocarbon concentration reveals some scatter in the
values of k. relative to the quantities HC/NO and HC + NO . How-
4 x x
ever, some trends are apparent from such a plot and we have used these
to establish our guidelines. In any event, we note that the range of
values of k, is not large, and that by far the most frequent values of
3 4-1-1
k, are 4 x 10 and 10 ppm min . Based on these frequencies of
appearance, we can formulate a general rule: if HC/NO >_ 1.7 , then
"3 _ "1 1 / 1 1
k, = 4 x 10 ppm min ; otherwise, k, = 10 ppm min . However, a
more involved, but more exact set of guidelines for selecting k, is as
follows :
1.
2.
3.
If HC/NO >_
X
If HC/NO <
X
If HC/NO <_
X
-1 . -1
ppm mm
1.7
1.7
1.2
, then
, then
or if
4 x
6 x
HC
103 <
103<
+ NO >
x
k, < 5 x
4 —
k. < 104
4 —
1.8 ppm
in3 -1 - -1
10 ppm mm
-1 . -1
ppm mm
4
, then k4 = 10
If 1.4 < HC/NO < 1.5 and 1.2 ppm < HC + NO < 1.4 ppm ,
X 3 -1 -1 X
then k, = 8 x 10 ppm min
151
-------
5.4.3 Meteorological Inputs
The three basic meteorological inputs are the maximum inversion base
height, a table of inversion heights as functions of time, and the
diffusivity parameters. Establishing general guidelines for selecting
these inputs is a difficult task; this is especially true for the last
two quantities. The spatial and temporal variability of meteorological
conditions impose a high degree of difficulty in trying to devise general
rules. Thus one would expect the inversion height, for example, to
attain different maximum levels at different locations. Similarly, the
changes of inversion height with time can be expected to show spatial
variability. Therefore, we must have information about inversion base
height as a function of time for each trajectory. This information can
be obtained from plots of inversion height isopleths for various times
for an ^entire geographical region. Given a description of the path of the
trajectory, the required inversion base data are then obtained from these
plots for the trajectory in question in a straightforward manner.
Determining the diffusivity parameters is somewhat more difficult.
In this case, the required data consist of profiles of temperature as a
function of height and time. These data determine the stability classes
and from these the diffusivity coefficients can be obtained using
Fig. 3.7- The time variation of the stability class determines the time
dependence of the diffusivity:coefficients. Results from our simulations
indicate that a likely value of diffusivity in the very stable case is
32 32
2.5 x 10 cm /a , rather than 5 x 10 cm /s as shown in Fig. 3.7. The
other values of diffusivity shown in Fig. 3.7 were used without change in
the simulations.
5.5 SOURCES OF UNCERTAINTY DUE TO SOLAR RADIATION AND PARTICULATE
REACTIONS
In the course of any serious evaluation process, any systematic
validation-check of a simulation model demands these two things:
1. A quantitative characterization of each uncertainty
entering the model
152
-------
2. A clear identification of the sensitivity of the model outputs
to each uncertainty in the inputs
Many efforts in the past several years have been concentrated ou the
second aspect enumerated above. Indeed, we and others have run parametric
analyses until the results are almost intuitive to any worker in the field.
Unanswered questions still surround the first requirement, however. These
questions need not be a source of mystery because straightforward well-
planned investigations can be designed to get the answers. No new dis-
coveries of laws of physics or chemistry will be necessary, but rather,
the commitment of resources that concentrate on finding out what is really
happening. Suggestions for the future are embodied in Sec. 6; however,
for now, we must content ourselves with identifying and assessing the
uncertainties in both the input data and the validation data base.
The primary process in the production of photochemical smog is the
photodissociation of nitrogen dioxide to form nitric oxide and atomic
oxygen. Our rate constants for this reaction are derived in two ways.
In the first, we assume clear skies and deal with the atmospheric trans-
mission of radiation in the dissociation bands according to a zenith angle
derived from time of day, day of year, and location on the earth's surface.
f\ /•
In the second method, Eppley ultraviolet detector readings are calibrated
to the clear-day curve of k. versus solar zenith angle. The cosine
correction presupposes the preponderant contributions to be from direct
rather than scattered ultraviolet radiation. . As will be seen, this
leads to significant errors only at large zenith angles.
Using these two methods, we have explored the uncertainty in k
that occurs for predictions that assume clear day values. Figures 5.102
through 5.107 show diurnal k. variations for each of the six data
days.
It can be seen from the figures that the departures of the actual
k from the theoretical clear-day values can be rather large. As might
153
-------
in
IO
-.00
SEP 11
CLEAR DAY
6.00 6.00 7.00
—1 1 1 1 1 1 1 1 1 1
a.oo 9.00 10.on 11.00 12.00 n.oo m.oo 15.00 is.oo n.oo
TIME tPST) (X1Q 2)
Figure 5.102. N02 Photolysis Rate Constant, k , for September 11, 1969
-------
:MD -
.30 -
.25 -
tn
in
.10-
SEP 29
1 1
6.QD 6.00 7.DO B.OO
—I 1 1 1
9.00 iQ.oo a.oo 12.00
TIME (PST) (X10 2)
13.00 IM.OO 15.00 16.00 17.00
01
Ln
Figure 5.103. NO Photolysis Rate Constant, k , for September 29, 1969
-------
.26 -
-"-
.10 -
.OS -
-.00
SEP 30
6.00 6.00 7.00 8.00 9.00 10.BD 11.00 12.00 13.00 1M.DO 15.00 16.00 17.00
10. HI 11.00 12..
TIME (PST) (X1Q 2
Figure 5.104. N02 Photolysis Rate Constant, k.. , for September 30,
1969
-------
.60 -i
.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00
5.00
Figure 5.105. N0? Photolysis Rate Constant, k.. , for October 29,
1969
-------
Ln
OO
to
g
ED
I
O.
6.00 6.00
16.M n.on
Figure 5.106. N02 Photolysis Rate Constant, k , for October 30, 1969
-------
.10 -
0)
z
o
a.oo B.oo
10.00 1L.OO 12.00
TIME (PST) (XIQ
16.00 17.00
Figure 5.107. N0? Photolysis Rate Constant, k , for November 4, 1969
-------
be expected, there also exist marked differences in the value of k^
at different locations in the Los Angeles Basin, in this case Commerce
and El Monte. The ratio of (theoretical ^/actual k^ ranges from unity
at a few points to 8/1 on September 11. As a general rule, the theoretical
k is higher by a factor of 1.5 or 2 for Commerce and by about 1.3 for
El Monte.
To ascertain the magnitude of the uncertainty introduced, we ran
parallel cases using the Commerce values of k.. as well as the clear-day
values for the trajectory starting near Pasadena at 0730 on October 30,
1969 and ending west of Pomona at 1330. This is by no means the worst
case, as is easily ascertained from the Fig. 5.106. The result of the
test was that the concentrations of NO were perturbed very little, but
X
the peak ozone concentration was lowered from 13 pphm using the clear-
day k1 to 9 pphm using the Commerce k , a relative change of about
31%. Such a margin of error is significant and must be borne in mind in
assessing the expected accuracy of the computed concentrations in view
of the uncertainties in the inputs. Interestingly enough, however, in
this particular case (see Fig. 5.73), the simulation using the clear-day
k. is actually very close to the measured data and using the Commerce
k.. would have degraded the simulation if no other rate constants were
adjusted to compensate for the lower ozone values. In closing, we note
that another case has been tested using the data for September 11 since
it shows the largest departures from theoretical values (see Sec. 5.3.5).
Another source of uncertainty is the influence of N0? + particulate
reactions. It will be recalled that such a lumped reaction, reaction
(2.16), was introduced in the kinetic model in simulating smog chamber
experiments involving auto exhaust in which the presence of aerosol was
observed (see Sec. 2.5.4). Introducing this reaction produced a better
simulation of the disappearance of NO from the system.
x
160
-------
The sensitivity of the model to changes in the rate constant of
this reaction was tested during the atmospheric validation runs. The
effect of increasing k-, was to lower the NCL concentration and to
increase the ozone concentration late in the day. The sensitivity of the
NO,, peak was negligible compared to the sensitivity of ozone. However,
the sensitivity of the end value of N0~ almost matches that of CL .
This interaction can be explained by realizing that in the kinetic model,
the reaction N0_ + CL is very strong at late times when N0« and CL
are high; thus removing NCL will cause the CL to increase. As an
example, we ran tests using the November 4, 1969 trajectory with
-3 -1 -3 -1
k, ft = 10 min and k., = 2 x 10 min . The peak ozone went from
25 to 32 pphm and the peak N0_ underwent a negligible change. The
concentration of NO- at the end of the trajectory was reduced from
21 to 16 pphm.
161
-------
6 CONCLUDING REMARKS
These findings establish progress milestones in the understanding
of chemical kinetics and atmospheric dynamics underlying air pollution
models.
Systematic investigations of laboratory smog chamber experiments
have provided significant insights into the air chemistry of urban atmps=
pheres. Although the artifices of the apparatus prevent direct applica-
bility to the atmosphere, important qualitative features are highlighted
by computer simulations of chamber studies. Mixtures of various hydrq-
carbons with nitric oxide were modeled. The hydrocarbons included pro^
pylene, toluene, toluene/n-butane, and auto exhaust from vehicles with
and without exhaust emission controls. Added chain-breaking reactions
for smog chambers improved the predictions markedly. These, along with
previously added OH + HC reactions, give consistent behavior for
various initial mixtures placed in the chamber. Incorporation of recently
reported rate data narrowed the options for adjustable parameters (accord-
ing to our ground rules, at least), but improved the validations overall.
As might be expected for a complex nonlinear system, the nonuniqueness
of the rate constants was discovered for validation of a given experiment;
that is, an entire set of rate constants could be moved through several
orders of magnitude and preserve the same computed results. Selection
of the best set was made using benchmark values that have recently come
out of the laboratory. Rank order and proportionality of hydrocarbon
reactivity indices were shown to bear a direct relationship to the set
of rate constants for hydroxyl attack of the hydrocarbons. This built
o -i
further confidence in our previously adopted procedure of reactivity
scaling of smog chamber reaction rates to model atmospheric systems.
Model methodology was substantially improved in several areas. The
logic for air trajectory computation has been systematized by generaliza-
tions derived from many hand calculations using actual data. Consistency
checks between tetroon trajectories and calculated ground trajectories
162
-------
revealed some large sources of uncertainty that are not considered in air
pollution models. The dominance of stratification over wind shear was
incorporated in changes in the calculation of vertical turbulent mixing
coefficients. While previous formulations stressed wind speed dependence,
the newly adopted methods depend on potential temperature gradients.
The impermeable inversion base assumption was abandoned for upper boundary
conditions. Vertical mesh intervals were extended well above the inver-
sion base with assignment of vertical diffusion coefficients controlling
upward mixing according to local stability conditions. Thus the inver-
sion base was traced through the mesh by varying the diffusion coefficient
in time and space. While the present work did not reexamine the source
inventories, many changes were made to assure direct comparison with other
model results. Extensive improvements in the computer implementation of
source models permitted a high degree of responsiveness to the frequent
alterations.
Turbulent diffusion transverse to the wind was assessed to evaluate
possible errors due to mass exchange between neighboring stream tubes.
As suspected previously, only minor perturbations are introduced by
lateral mixing perpendicular to the path of an air parcel. The case of
side-by-side trajectories was tested using the GRC three-dimensional time-
dependent LAPS code which is especially adapted to treat conditions of
large transverse gradients in emission fluxes that typify localized large
sources. Worst-case carbon monoxide area sources were tested using the
extreme values of fluxes determined from dozens of actual air trajectories
over the Los Angeles Basin. The spread of power plant stack plumes from
off-trajectory sources into the control volume was investigated for a wide
range of parametric conditions of point emission and area emission strengths
To insure realism, the parametric values were selected from the actual
inventory statistics. Even the largest percentage errors are not likely
to exceed other uncertainties due to model inputs. Systematic examina-
tion of validation trajectories relative to these findings showed that
transverse diffusion errors are far smaller than other sources of uncer-
tainty in the model inputs.
163
-------
With the chemical and meteorological improvements, the validation
results have been more gratifying than ever before. Consistency in
assumptions for both the kinetic rates and the diffusion parameters can
be maintained to model a wide variety of cases without ad hoc adjustments.
Our abandonment of the single-receptor validation criterion placed much
more rigorous constraints on the model tests than had been originally
anticipated. Each hourly trajectory node has an interpolated set of
concentrations based on neighboring station values. Therefore, we seek
to match the shape of each pollution history rather than just matching
values at the end point. Despite these more severe requirements, both
diffusion and photochemical validations were remarkably successful.
Some problems remain that are critical to the future success of
simulation modeling. It is likely that they stand alone by now as the
main obstacles to further fidelity improvements. Following this reason-
ing, one realizes the need to establish closer coupling between modelers
and measurers than has been the case in the past. The continuity that
is thus assured will build the scientific foundations needed to attack
air pollution abatement problems on a rational basis.
Finally, of the myriad topics for additional research which are
worthy of note, we wish to single out a few which should improve future
models when the problems posed have been solved. In what follows, the
order of discussion of the topics has no particular significance.
The first subject that merits some discussion is the need for a new
measurement of the rate constant of the reaction NO +0 -> NO +0 .
^ -J O Ł,
In our model, this rate constant had to be reduced by a factor of 10 in
order to. achieve accurate reproduction of smog chamber experiments. Since
this reaction appears to be rate-controlling at late times when NO and
03 have reached high concentrations, and since the available measurements
of this rate constant are rather old (1949 and 1957) a new measurement
is warranted.
164
-------
A second topic of interest is whether HONO exists in the atmos-
phere in significant amounts (around 1 pphm). Should this not be the
case, the chain-breaking structure of the kinetic model will have to be
revised. A corollary to this question is that if HONO is found in the
atmosphere, then the rate constant of the reaction hv 4- HONO ^ OH + NO
needs to be determined. There appears to be no measurement of this rate
constant at the present time.
Thirdly, additional investigation is needed to ascertain the nature
of NO^-scavenging processes in the atmosphere. The possibility that
heterogeneous reactions may play a significant role must be included in
such a study. Such knowledge would help us to improve the late-time
behavior of NO in the simulations.
165
-------
APPENDIX A
A VIEW OF FUTURE PROBLEMS IN AIR POLLUTION MODELING
*
This first appeared as General Research Corporation TM-1631, March 1972
and was subsequently published in Proceedings of Summer Computer Simula-
tion Conference, Simulation Councils, Inc., LaJolla, Calif., June 1972,
pp. 1013-1027- Research reported in this document was originated through
independent efforts, not under a Government contract or program.
167
-------
A.I INTRODUCTION
Vigorous steps in abating air pollution demand heavy investments
both in the public and private sectors. In addition, second-generation
cleanup measures are likely to add personal inconveniences to the already
growing financial burdens. Large dollar outlays and the need for public
support demand that decision-makers have reliable means of evaluating
alternative abatement strategies.
Mathematical air quality models are quantitative tools that will
play a central role in evaluating the environmental aspects of decisions.
These decisions may take the form of regulations aimed at rolling back
existing pollution or of minimizing the environmental damage potential of
future public works projects. For abating existing sources, implementa-
tion planning must show how control regions will achieve ambient air
quality standards within a specific number of years. This requires
predictions of absolute levels of air pollution. For assessing impact
of projected sources, a statement must be filed that demonstrates that
these sources will not cause environmental damage. This, too, requires
predictions of absolute levels of air pollution. These requirements
impose stringent demands on the best air quality models presently
available.
For planning long-range strategies on a national scale, the objective
is to choose rationally from among a field of alternative abatement
actions. The selection criterion is built around maximum benefit/cost
ratios. Thus, the long-range considerations require measurements of
alternatives on a relative scale. Currently available air quality models
will be very useful in fulfilling this less stringent requirement.
Unfortunately, the time scale on implementation plans and impact
statements is much shorter than that on national planning goals. It is
already becoming evident, however, that the costs of really massive roll-
back strategies far outstrip our ability to pay over the short time
168
-------
intervals required by recent statutes. Consequently, we will be forced
to analyze less ambitious plans in order to define optimum steps toward
pollution abatement. This realization may be many months or even years
away, but when its effects are felt, it will require the best products
of the modeler's art.
This paper takes two directions in assessing these future needs:
first, it highlights some scientific problem areas that need immediate
attention; and, second, it suggests some ways of adapting models for
abatement analysis applications. Any approach to the unsolved phenomenology
problems inevitably leads to greater degrees of complexity in the model.
In direct opposition to this trend stands the need for sweeping simplifi-
cations in practical adaptations of the models. More likely than not,
the air quality model will be but one of the many modules in any realistic
abatement simulation. The conflict between pure and applied efforts can
be resolved only by a high degree of communication between the researchers
and the planners »-
In the realm of phenomenology, we examine two potential sources of
error in predicting atmospheric reaction rates. One involves gas-solid
interactions on urban surfaces and on aerosol particles. The other arises
from turbulent fluctuations of reactant gas concentration. Experimental
evidence of the problems is cited and research approaches to its solution
are outlined in each of the two cases. Analytical corrections to the air
quality models are proposed.
Systems- implementation schemes are then considered to define the
fidelity level that the models must achieve. A hierarchy of different
versions emerges from the various objectives that are set forth. For
some cases, the versions are already available, but for others, only a
sketch plan of the specifications can be given.
169
-------
A. 2 PHYSICAL INTERACTIONS WITH POLLUTANT REACTION RATES
A.2.1 Gas-Solid Interactions
Urban surfaces near ground level and particle surface's distributed
through the mixing layer can serve as reaction sites for gas molecules
impinging on them. If these effects compete significantly with homogeneous
reactions, appropriate sink mechanisms must be introduced into atmdspheric
models. Indeed, this has already been done for oxides of nitrogen in our
27
earlier work. Briefly summarizing, we noted that the observed buildups
of carbon monoxide and hydrocarbon during morning peak traffic were well
modeled by the values of emission fluxes and atmospheric diffusion coef-
ficients in the literature. On the other hand, the sum of (NO -f NO^)
was grossly overpredicted as shown in Fig. A.I (Curve A), On the* graph,
the symbol "r" refers to the oxidation fate reduction (beldw that of pure'
propylene) and symbol "f" to the fraction of the inventory value NO
X
emission flux used.
160
I
a
a
z~
O
1-
140
120
100
CURVE
A
B
C
D
r
1/2
1/3
1/2
1
f
1
1/4
1/4
1/4
LU
O
80
_ 60
CM
O
z
40
20
r- CURVE PARAMETERS
r = RATIO OF HYDROCARBON OXIDATION RATE
TO PROPYLENE RATE
f = NO EMISSION SCALE-DOWN RATIO
_L
_L
_L
_L
_L
0600 0800 1000 1200
TIME OF DAY, PST
1400
Figure A.I, (NO + N02> - Concentration Ground Level Buntingto'n Park
170
-------
Due to an apparent rapid removal of nitrogen oxides from the gas
phase, the published values of emission rates had to be reduced by a factor
of four to offset the losses. Note that the first 2.5 hours of buildup
are practically unaffected by the choice of reaction rates (over a factor
of three). Although flux reduction is an ad hoc correction for these
results, it may not be generally applicable to all types of surfaces or
to all types of days or even to all times during a given day. Figure A.2
illustrates the difficulty very clearly. It displays averages of the
CO/NO mole ratios for groups of 1968 data measured by Scott Research
X 35
Laboratories. It is intended that CO be regarded as an inert tracer
so that variations between observed ratios and source ratios reflect
loss of NO . This removes uncertainties due to dilution and other
X
interferences. Types 1, 2, and 3 denote high oxidant days and Type 0,
low oxidant days. The factor of four (between source values and air
70
60
50
40
O 30
20
10
(CURVES FAIRED THROUGH DATA POINTS)
AVERAGE OF TYPE '1, 2, AND 3' DAYS (HIGH OXIDANT)
AVERAGE OF TYPE '0' DAY LOW OXIDANT
ALL SOURCES
0800
0800
1000
TIME OF DAY, PST
1200
1400
Figure A.2. CO/NO Ratios for Huntington Park 1968
X
171
-------
values) shows up clearly at the morning traffic peak for high oxidant days.
Low oxidant days exhibit far lower discrepancies between source ratios
and ambient air ratios.
These bits of evidence show the nature of the problem but they do
not indicate a truly general solution. Indeed, the deficit of gas phase
nitrogen oxides has been observed for years in laboratory photooxidation
9
experiments in smog chambers. Gay and Bufalini give an excellent review
of these problems and demonstrate that analysis of surface-adsorbed
products greatly improves the nitrogen balance in these experiments.
Substantial uptake of pollutants in soils has been demonstrated in
OT OQ OQ
experimental studies '' reported in the literature. From these results,
ground absorption can be deduced or estimated. They average out to the
2
following very approximate values (in mg/m «hr): ~1 for ozone, ~8 for
carbon monoxide and ~3 for nitrogen dioxide. It is important to note that
the ozone value is for atmospheric background concentration (-0,05 ppm)
while the CO and N0_ values are for tens or hundreds of ppm in a 10-
37
liter vessel. Aldaz assumes that the surface reaction rate must be first-
38
order in the ozone concentration, but the Inman and Ingersoll results
show a linear concentration decay of CO suggesting zero-order. Before
any of the reported values can be used in models, the concentration (and
possibly temperature) dependence of the uptake rates must be determined.
It is of interest, nevertheless, to compare these values with the emission
fluxes averaged over the Los Angeles basin area. For oxides of nitrogen
o
(as NO^) the emissions are approximately 10 mg/m • hr (against 3 estimated
for soil uptake) and for carbon monoxide it is approximately 100 (against
8 estimated for soil uptake). Therefore, these preliminary indications
are consistent with our model adjustments and with the observed atmospheric
C0/N0x ratios; namely, that there could be significant perturbation of
the N0x~balance, but only minor effects on the CO-balance. Surface
uptake of ozone has been regarded as the major global sink for this gas;
however, Ripperton and Vukovich indicate that gas phase removal may .also
be important even at background conditions. Some of our early calculations
for a fully catalytic ground surface suggest that gas-phase reactions
172
-------
dominate the ozone balance in polluted air to within a few meters of the
ground.
Modeling the uptake at the ground is a simple matter of adding
boundary conditions to the finite-difference approaches that simulate air
quality for distributed sources. As indicated above, the lack of reliable
data is the main obstacle. Algorithms are already developed for the flux
boundary condition at (or near) the ground level. In general, this logic
is suitable only for zero-order surface reactions, but minor modifications
would generalize it to the rith-order uptake processes. The modifications
would substitute kc for the constant flux now in the boundary conditions,
Now turning to the gas-solid interactions at particulate surfaces,
41
we begin with the evidence cited by Lundgren:
On days of heavy smog, very hygroscopic, crystalline-
like particles were found to comprise over half of
the particulate dry weight in the 0.5-1.5 ym
diameter size range. These crystalline particles
were analyzed by X-ray diffraction and identified
as ammonium nitrate.
It has been known for some time (see for example, Ref. 42) that
Los Angeles aerosol has anomalously high fractions of nitrate when com-
3
pared with that from other cities. Altshuller and Bufalini cite smog
chamber experiments with auto exhaust emphasizing the rapidity of conversion
of nitrogen oxides to particulate nitrate. They urge that further field
work be undertaken to determine atmospheric rates of conversion.
Oxygen-atom reactions with particulate matter were discounted by
43
Leighton and Perkins. They carried out a calculation of the mean displace-
ment during the gas-phase lifetime of an 0-atom. For a high particle
3
loading (1 mg/m ), they then estimated the total volume of the spheres
-4
of influence of the particles. It accounted for less than 10 of the
volume of the gas, thereby forming a basis for neglecting the influence
on 0-atom chemistry.
173
-------
We can undertake a somewhat more detailed analysis of other pollutants
by carrying out some calculations for simultaneous gas-phase and surface
reactions in the neighborhood of a particle. The physical picture is a
steady state diffusion of a reactive species toward the particle surface.
For steady state, the governing equation is
DV2n. + w± = 0 (A.I)
where D = Diffusion coefficient
n. = Number density of ^th species
w. = Net production rate of _ith species in the gas phase
The form of the production term is predicated on a high level of chemical
activity for species i . This is based on a constant strong source s.
opposed by a series of fast reverse reactions that consume species i
at a rate of k n. where k is a composite of gas-phase rate constants
o o
(for reactions with q other species) and concentrations given by
k, = > .0- - ^J^.n, (A.2)
where 6.. = 1 ; i = i
ij J
«±j - 0 ; i / j
so that for the gas phase production term we use
Wi = Si " kgni (A<3)
Outside of the range of influence of the particle, we can calculate the
homogeneous stationary state number density of species i by setting
w = 0 giving
si
n°° ~ k (A. 4)
g
174
-------
Incorporating spherical symmetry and substituting Eq. A.3 in Eq. A.I, we
get
- k n. = -s
(A. 5)
The heterogeneous reaction enters the problem in the boundary
condition at the particle surface (r = rn) where we assume a first-order
reaction balancing the influx of the jLth species,
dn
d
- "Vic
(A.6)
where k is the rate constant for the surface reaction.
S " " " --..---. .....
The homogeneous reaction is approached at an infinite distance
from the particle
lim n. = n
(A. 7)
which corresponds tQ the vanishing first and second derivatives on the
left=hand side of Eq. A.5. Solving Eq. A.5 subject to Eqs. A.6 and A.7,
we can obtain by elementary methods the result
n
- exp
r -
(A. 8)
175
-------
which closely resembles the classical Debye equation for an ion in an
electrolyte solution. Equation A.8 expresses an exponential decay from
the ambient level into the surface of the particle. The e-folding distance
is approximately the length a typical molecule travels before it reacts.
At small fractions of the e-folding distance from the particle surface,
the reciprocal r-dependence dominates. The coefficient of the decay
term brings in the influence of the surface reaction.
It is instructive to examine limiting cases of the surface concentra-
tion by setting r = r in Eq. A.8. For a very small diffusion coefficient
(D) the molecules of species i are insulated from the surface by a thin
film. At the surface the concentration is zero because every molecule
that penetrates the film is consumed at the surface on arrival. This is
the diffusion-controlled limit, and it results in a zero surface concentra-
tion. At high values of k , the presence of the surface or the diffusion
o
process has very little effect and we have approached the homogeneous
reaction limit. The free-molecule limiting case where surface reaction
rate is the controlling feature occurs with very small reactive particles.
Thus if rQ « D/kg and if k » k D , this limit is approached.
For cases of interest in air pollution, it is useful to compare the
magnitudes of the terms in Eq. A. 8. Assume that the mean particle
diameter is 0.5 ym. The diffusion coefficient is of the order of 0.1 cm2/s,
and for active species in photochemical air pollution k ~ 0.004 s .
o
The least certain of all the constants is the surface reaction coefficient
kg . Kinetic theory sets the upper limit on k by the frequency of
molecular collisions T. , on a unit area.
s
(A.9)
176
-------
where m = Molecular mass of jLth species
k = Boltzmann's constant
T = Absolute temperature
Now k can be derived directly from Eq. A.9 by applying a collision
efficiency n . This is the fraction of collisions that result in surface
reaction. The value of n , a measure of chemical surface efficiency, is
determined by the surface coverage of reactant partners and by the activa-
tion energy required to make a reaction occur. Combining Eqs. A.6 and A.9
with the concept of collision efficiency, we obtain
(A. 10)
For nitric oxide at 300°K, this gives a value of ~(1.2 x 10 n) cm-s~ .
The measurements of Aldaz ahow values of ri for ozone on active surfaces
-4
in excess of 10 (for the case of juniper). At the time of this writing,
-4
we do not have values available for oxides of nitrogen, but if 10 is
used, then k is of the order of unity.
s
in Eq. A. 8: (D/k r ) = 3 x 103, (k D)1'2/k = 2 x 10 2 , and /D/k - 5
Using these estimates, we get the following values for the parameters
cm
This indicates that the second term (which is the fractional species de-
ficit in the neighborhood of the particle) is always small and that the
full value of concentration n is exposed to the surface. Thus we
are in the regime that is dominated by surface- reactions even if collision
_2
efficiencies range as high as n - 10 . This simplifies particle up-
take estimations considerably because Eq. A.9 can be used directly with
the ambient gas concentration at the surface (n = n^) if the efficiency
factor is applied to the particle flux T .
Let us turn to the task of placing an upper limit on the role of
particles in competition with gas phase reactions in photochemical smog.
To obtain numerical comparisons between the rates, we must specify some
177
-------
gas phase reactions and assign some concentration values. Table A.I
summarizes some typical conditions on a smoggy day in Los Angeles.
TABLE A.I
CONCENTRATIONS (ppm) AND GAS PHASE RATE CONSTANT ASSUMED FOR COMPARATIVE
ANALYSIS
Species
Ozone
Nitric Oxide
Nitrogen Dioxide
Oxygen Atom
NO. Dissociation
Rate Constant
Early Time (~6-9 AM)
-3*
5.5 x 10
1 * 10"1
1 x lo'1
_QA
8 x 10
0.22 min"1
Late Time (-midday)
2 x iQ"1
-3*
2 x 10
5 x 10~2
-9*
5 x 10
0.30 min~
Computed from stationary state relationships.
The footnote on the table refers to the reaction steps chosen to
characterize gas phase rates. These elementary processes are:
11
hv + NO,
NO + 0
12
M
NO + 03 -^ N°2 + °2
(A.11)
(A.12)
(A.13)
where species M is any collision partner. This mechanism comprises
the fastest gas phase reactions that influence -the species in Table A.I.
178
-------
For calculating the starred concentrations, we assumed stationarity for
0-atom giving
kllCN09
c. = 1 — (A. 14)
° k!2c02CM
and for ozone giving
kllCNO
co
U3
where c denotes ppm concentration of the subscript species. Rate constants
—5 —2 -1
for reactions (A. 12) and (A. 13) were assumed to be 1.3'2 x 10 ppm rain
-1 -1 44
and 40 ppm min consistent with some of our modeling calculations..
In a later section, we express caution about using Eq. A. 15 because of possi-
ble turbulence interference effects; however, for the present purposes
we will assume that we have properly averaged concentrations for use in
stationary state calculations. Since all of the stoichiometric coefficients
in the reaction cycle are unity, stationarity permits us to calculate a
single gas phase reaction rate characterizing all of the transformations
in the cycle.
The gas reaction rate for reactions (A. 12) and (A. 13) at early
-2 -1 -1 -2 -1
time is 2.2 x 10 ppm min , and at late time it is 1.5 x 10 ppm
min . To get upper bounds on the surface rates using Eq. A. 9, we will
need to know the surface area per unit volume. Following Leighton and
43
Perkins, we assume r = 0.25 ym and a specific gravity of unity. For
an aerosol loading of 200 yg/m (which is typical for Los Angeles smoggy
— fi 9 *^
days) this gives a surface area of 6.08 x 10 cm per cm , Using T
for each species from Eq. A. 9, calculations can be made for the collision
rate of each species on the aerosol surfaces. This gives the upper limit
for the surface reaction effect. Ratios of surface to gas phase rates
179
-------
are shown for this limit in Table A.2. If the ground level ozone flux
o 7
measurements of Aldaz apply also to aerosol surfaces, efficiency factors
in the range of 10 to 10~ reduce all of the numbers in Table A.2 to
insignificant levels. On the other hand, if there is a moderate degree
_2
of surface activity on the aerosol particles (say 10 collision
efficiency), we see from Table A.2 that the morning NO levels could
2t
be seriously affected.
TABLE A.2
UPPER LIMIT OF (SURFACE RATE/GAS PHASE RATE) RATIO
Species
Ozone
Nitric Oxide
Nitrogen Dioxide
Oxygen Atom
Early Time (~6-9 AM)
8.5 x 10"1
2.0 x 10+1
1.6 x 10+1
-6
2.1 x 10
Late Time (midday)
4.5 x io+1
6.0 x lo"1
1.1 x 10+1
-6
2.0 x 10 °
* - 3
For conditions in Table A.I and for 200 ug/m of 0.5-ym-diameter particles.
This possibility deserves serious consideration in view of the
atmospheric nitrogen balance results we cited above. In some of the smog
chamber experiments reported by Gay and Bufalini, the majority of the
oxides of nitrogen loaded initially show up in the solid phase after only
a few hours of irradiation. Finally, it should be noted from Table A.2
that our results agree with those of Leighton and Perkins43 for oxygen
atoms; namely, that gas-particle interactions are negligible by several
orders of magnitudes even if oxygen atoms react at 100% efficiency with
the aerosol. The influence of surface reactions on ozone concentration
can become moderate in the later stages of smog formation.
180
-------
The critical factor in assessing the impact of these findings is
the surface reaction efficiency as well as the order of the reaction.
If these quantities were more accurately known, it would be possible to
tell which of the surface processes must be included. Not covered in
the above discussion are the large families of organic radicals and
compounds that may also react with aerosol surfaces. Certainly the
hydrocarbon reaction rates are much smaller than those discussed; there-
fore, heterogeneous reactions could contribute very significantly. The
intermediates such as RO, R02, and HONO and the products like aldehydes
and alkyl nitrates must also be investigated.
Having identified the significant processes, we can incorporate
them into the air quality model by adding sink terms to the continuity
equation for each species. Such a term would depend on a rate constant,
the particulate level, and probably the species concentration raised to
some power greater than zero. These reaction terms could be lumped in
with others. If, for example, a high degree of correlation were found
between ozone and aerosol levels, the ozone terms could be augmented to
account for heterogeneous removal mechanisms. There may also exist
product species which become detached from the particles. If a surface
is coated with loosely bound B-molecules, an impact of an A-molecule on
the surface might be followed by ejection of an AB-molecule. It is
clear that a great deal of research is yet to be done in this field.
Before leaving the question of gas-solid interactions, we should
not fail to mention the possible effects of attenuation of incident sun-
light by the aerosol, particularly in the ultraviolet. Since the photo-
dissociation primary processes are most sensitive to the ultraviolet input,
there may be a significant reduction of the reaction rates without a
commensurate reduction in total incident solar energy. In one version
of our model, Eppley detector readings can be used directly to get the
&
rate constants. This automatically accounts for ultraviolet attenuation
down to ground level. When such data are not available, any significant
A '
This calibration was based on a particular filter system having a 0.3-
0.4 ym bandpass.
181
-------
reduction in rate constant must be accounted for by solving a radiation
transfer equation over the dissociation band of wavelengths. To the best
of our knowledge, this is not a part of existing atmospheric pollution
models. Of course, before extensive development of any new simulation
logic is undertaken, the class of cases where the effect is important
must first be identified. This done, it will be necessary to delineate
the regime with certain parametric criteria.
A.2.2 Turbulent Fluctuation Interactions
The second area of needed research is the influence of concentration
inhomogeneities upon atmospheric reaction rate calculations. As an air
parcel moves over an array of different emission sources, turbulence folds
in the gases of different composition. The nonuniformity manifests itself
as blobs that become less and less distinct due to the combined action
of diffusion and reaction. Attempts to describe such phenomena have formed
a body of theory for turbulent fields of scalar quantities. Unfortunately,
little supporting evidence in the form of actual observations is available
for the testing of such theory. We can, nevertheless, utilize some of
the grosser aspects of the theory for model improvements if some appro-
priate experimentation is done.
The specific problems arising due to fluctuating concentrations can
be clarified by considering the reaction between the pollutants NO and
0 . The reaction of ozone and nitric oxide in polluted air rapidly
produces nitrogen dioxide and molecular oxygen. Characteristic reaction
times for part-per-hundred-million level concentrations of the reactants
(0^ and NO) might range from several seconds to a minute or more. This
reaction is opposed by the photodissociation of NO which has a two or
three minute characteristic reaction time in bright sunlight. The result-
ing quasi-equilibrium, therefore, can respond relatively rapidly to
changing reactant stoichiometry. (Reactions (A.11), (A.12), and
(A.13) in the previous section form the mechanism in question.) If con-
centration changes are induced by turbulent mixing, which may be
182
-------
characterized by the same time scales, there is an interaction between
the turbulent fluctuations and the mean reaction rates.
Here is what is observed: suppose we station ourselves at some
point and measure the atmospheric reactant concentrations with good time
resolution. Typical records (see Fig. A.3) will show that there is
significant anticorrelation between 0. and NO concentration. For
example, one-minute 0_ and NO readings from the first 20 minutes of
this sample have a cross-correlation coefficient of -0.742 (see Table A. 3)
Consequently, if we calculate the mean reaction rate from
dcNO .
dt ~ ~kCNOC0,
(A.16)
we find that the fluctuations introduce a correction into the rate
because
CNO
"NO
(A.17)
and
(A.18)
where bars denote means and primes, fluctuations. Insertion of Eqs. A. 17
and A.18 in Eq. A.16, followed by averaging, gives
dc
NO
dt
-k
CNOC0,
1 +
(A.19)
Note that if there is vanishing correlation between the two concentrations,
the parenthetical factor is unity. Negative correlation will clearly
reduce the reaction rate in Eq. A.19 (the amount depends on fluctuation intensity)
183
-------
0.4 P-
8 °-1
o-U
0400
NITRIC OXIDE
00
CXI
0410 0420
TIME OF DAY, EST
0430
Figure A.3a. Chemiluminescent Measurements in New York - 1970
(44)
0.010 r-
0.009 -
0.008 -
LU
O
z
g 0.007
0.006 U
0400
0410 0420
TIME OF DAY, EST
0430
Figure A.3b. Chemiluminescent Measurements in New York - 1970
184
-------
TABLE A.3
FOSD/NEW YORK DATA44 (FIRST 20 MINUTES)
Input Data
ID
Time After 0400 EST
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
Mean Values =
Correlation Coefficient =
X
NO ppm
0.150000
0.117000
0.052000
0.090000
0.057000
0.070000
0.063000
0.097000
0.122000
0.055000
0.080000
0.088000
0.058000
0.067000
0.060000
0.09SOOO
0.064000
0.078000
0.054000
0.080000
-0.742
Y
03 ppm
0.006150
0.006150
0.008120
0.007780
0.008240
0.007620
0.007940
0.007330
0.006420
0.007970
0.009140
0.007710
0.007080
0.007480
0.007980
0.007480
0.008480
0.007300
0.009070
0.007655
185
-------
9 A
In earlier work, we tested the quasi-equilibrium of the three-
reaction cycle given in reactions (A. 11), (A.12)i and (A.13). It will
be. recalled that ozone quasistationarity requires that
(A'20)
neglecting the turbulence effects on reaction (A.13). Based on 10-miriute
averaged concentrations, Fig. A. 4 shows thia systematic departure from
Eq. A. 20 in the direction of too high an apparent rate fbr reaction (A.13)
To suppress inaccuracies in small NO readings, we have omitted those
values less than (or equal to) one part per hundred million (pphm) . One
explanation for the departure may be reactions besides (A. 11), (A. 12) and
(A.13) playing a significant competitive role. Mode'ling calculations for
these conditions have as yet failed to reveal such reactions; Another
hypothesis is instrument inaccuracies at high ozone level. Figure A. 5
shows the results of back-calculating what the ozOhe meter (MAST) response
would have to be to bias the measurements as suggested in the data in
Fig. A. 4. To achieve quasiequilibriumj we wduld need an instrument that
would read 20 or 25 pp'hm when the actual value is 5 or 6. This extent
of inaccuracy is unlikely especially since it has been reported that
MAST instrument readings are consistently below the true values. This
correction is even in the wrong direction to support the plausibility of
a response curve passing through the data of Fig; A. 5. Detailed back-
ground on data analysis and the calculations of rate constants is given
24
in ah earlier publication.
Two possible explanations of the breakdown of 0,,/NO/NO,2 quasista-
tionarity will now be considered. The first is the obvious possibility
of interference from competing reactions. The second is the effect of
turbulence on the rate of reaction (A.13). For these investigations, we
will make use of Our mathematical model of the chemical-kinetic aspects
of photochemical smog.
186
-------
1.00
0.75
0.50
g 0.25
O
z
O
z
0.00
O
O -0.25
-0.50
-0.75
-1.00
-1.25
POINTS WITH[NO] > 1 pphm
S:
i
+ + +
+
-H-
LOCUS OF POINTS SATISFYING
QUASI-EQUILBRIUM HYPOTHESIS
+ .++
*+ +
+ + . +
1 1
12 16
OZONE, pphm
20 24 28
Figure A.4. Quasiequilibrium Teat for 1969 Ground Data at El Monte—
High NO Levels
187
-------
0.00 2.50 5.00 7.50 10.00 12.50 15.00 17.50 20.00 22.50 25.00
MEASURED OZONE, pphm
Figure A.5. Ozone Inaccuracies Needed to Explain the Departures from
Quasiequilibrium in Figure A.4
To determine the effect of competing reactions, we used a 14-reaction,
n I ~| Q
10-species model of photochemical smog. As reported in previous work, '
the computed concentrations simulated accurately the experimental work
45
of Altshuller, et al. Using the results of this simulation, we computed
the ratio k [NO ]/k [O.J[NO] with the rate constants fixed at
J-J- ^ J_ J J "1*1
k.... = 0.4 min~l and k.. _ = 0.4 pphm min . A plot of the logarithm
of the ratio versus ozone concentration is shown in Fig. A.6. It is
apparent that the ratio is different from unity and that the departure
increases with increasing ozone. However, the maximum value of the logarithm
is only 0.04, indicating that the ratio is very close to unity compared
with the atmospheric deviations shown previously. It seems reasonable
to conclude from this that the interference from other reactions is
negligible and that 03, NO, and N02 closely approach quasiequilibrium
under the static conditions that prevail in a smog chamber.
188
-------
o
C)
X
o
z
o1
1
0
o
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
~ 1
Q
c
•<;
N
+ + ** + ** 6
* * ^
-
-
,
i i i i i i i i i i i i
10 15 20 25 30 35 40 45 50 55 60
OZONE, pphm
Figure A.6. Quasiequilibrium in a Simulated Smog Chamber Experiment
In Fig. A. 6, it should also be noted that the trend away from
equilibrium is in a different direction from that found in the Los Angeles
data (Fig. A«4). This may be significant inasmuch as it implies an
excess of N09 and a deficiency of NO and 0_ . We are well aware
that this effect could be due to the model. Nevertheless, if the effect
is accepted as real, this implies that what was observed under atmospheric
conditions is probably not due to interfering reactions . This is in
46
agreement with a statement by Schuck and Stephens, who claim that
quasiequilibrium holds in the presence of competing hydrocarbon reactions
inasmuch as the rates for these reactions are very low. However, they
go on to claim that quasiequilibrium holds in the Los Angeles atmosphere,
but offer no evidence to support their assertion. Finally, we note that
the pair of reactions
°
N0
°
'2 3
NO + N0n
N03
2NOr
189
-------
suggested by Leighton42 as likely candidates for interference, would pro-
duce results that contradict the data of Fig. A.4. In fact, these
reactions would perturb quasiequilibrium in the direction indicated in
Fig. A..6.
The above arguments are based on the premise that the smog chamber
experiment represents the principal chemical processes that really occur;
in the atmosphere. Many more known (as well as unknown) chemical inter-
actions take place in urban air and thus it is always possible that some
reactions may be interfering with the 0,/NO/NO cycle in a manner which
produces the observed results.
It is apparent from Eq. A.19 that turbulent fluctuations have the
effect of modifying the rate of reaction (A.13). The apparent validity
of the quasi-equilibrium hypothesis is going to be affected if we use
the product of time-averaged concentrations instead of the time average
of the product of concentrations. This notion is in agreement with
Leighton's caveat about the effect on stationarity of rapid changes in
concentration.
Having identified the turbulence interference problem in air pollution
modeling, let us now consider some methods of attacking it. To provide
further insight into the problem, experimental data like those shown in
Figs. A.3 and A.4 must be obtained simultaneously; namely, time-resolved
concentrations of 0_, NO, NO-, temperature, and ultraviolet input.
The monitoring station should be located at each of two types of
urban environment: one in an area dominated by distributed transportation
emission sources and another generally downwind from intense nitric oxide
sources). The first provides low fluctuation tests and the second, high
fluctuation tests. The data obtained in such a program would provide
direct evidence as to whether the apparent deviations from quasiequilibrium
were due to fluctuation interferences. Diagrams like Fig. A.4 could be
190
-------
constructed both from time-averaged reaction rates and from rates derived
from time-averaged conceit rations. Comparisons would test explicitly the
hypotheses advanced o: :va.
Another phase of needed experimental research must be done in the
laboratory. Its objective is v.h-.: construction of a theoretical approach
for inclusion of rate corrections in computer simulation models. The method
of attack in the experiment must be designed to yield fundamental informa-
tion on turbulent flow reaction rates. Thus it must incorporate a well-
understood (necessarily Sifaplt) reaction system in a controlled turbulent
mixing process such that react ion-time/mixing-time ratios are adjustable
and near unity. This difficult set of specifications is being approached
in an experiment in progress at TRW Systems under the direction of
47
R. G. Batt. It involves the N?0, dissociation reaction in a free
shear layer in a low speed wind tunnel. Optical probes are employed for
concentration measurements in addition to conventional aerodynamic instru-
mentation. Other tunnel experiments for studying the effect may also be
conceived. An expected end result of the experiment is a measure of steady
concentration fields and velocity fields as well as the ensemble of
turbulent statistical properties.
Theoretical treatments based on both the field and lab findings must
be undertaken in parallel in order to give model modifications that account
for any mixing interferences with chemical reactions. Theoretical inter-
pretation of the laboratory experiments will give a detailed treatment
of a reactive plume mixing into a background flow under steady conditions.
Coupled with the flow field through time-averaged velocity and concentra-
tion is a fluctuation budget equation like
3fi
?r- " ' e (A-21)
191
-------
where f - c'2 , the mean square fluctuation of species i
D = A diffusion coefficient for fluctuation transport
D * Turbulence diffusivity
E = Dissipation of fluctuations
This equation for total fluctuation content might incoporate the
statistical approaches of Bugnolo and Corrsin in integrated forms.
It holds for a reaction that is second-order in species i . As an
approximation, one might assume D ~ D . The dissipation term consists
of two components, one due to diffusive smearing of fluctuations and
another due to reactive destruction of fluctuations. We can approximate
these by
- 12 f + (A.22)
which expresses the two effects respectively with its two terms. The
new symbols are defined as follows :
D = Molecular diffusivity
A, = Microscale of concentration fluctuations
k. = Second-border reaction rate constant for ith species
Application of Eq. A. 21 to the laboratory results will show how
chemical dissipation of species fluctuations is influenced by chemistry.
This effect then feeds into the equations for mean concentrations through
the reaction rate terms (e.g., Eq. A. 19). The validity of the theoretical
approach can be checked by computer modeling of the mean and fluctuating
concentrations in the experiment. A key aspect of this validation procedure
is a knowledge of the transport and reaction coefficients. Consequently,
it is essential to keep both the chemical system and the flow field as
simple as possible rather than to attempt to construct a physical scale
model of an urban airshed.
192
-------
Applications of the research findings in air pollution simulation
models will require considerable simplification because inadequate know-
ledge of coefficients and complexities of geometry make an elegant treatment
inadvisable; however, the research is needed to tell us where the approxi-
mations are valid. Quantitative implementation of rate corrections can
take the form of gross parameters like a plume gradient criterion to tell
when a correction is needed and a mixing delay time that is a function
of scale length and velocities. The mixing delay time in turn can be
introduced into a correction factor that is formally applied in the
chemical rate portion of the simulation logic.
All of the details of these formulations are yet to be worked out;
hence, it seems imperative that this be investigated in the immediate
future. The large discrepancies observed between our notions of the
inorganic kinetics and the observations cannot be ignored. Until their
sources are discovered and rectified in the simulation models, it seems
unwise to mount extensive validation efforts that repeatedly apply existing
approaches to more and more time-averaged air quality data.
193
-------
A. 3 SYSTEMS IMPLEMENTATIONS OF PHOTOCHEMICAL/DIFFUSION MODELS
A.3..1 Few Runs/High Fidelity
For evaluation studies of emission source contributions, the
influence of specific control measures can be determined by running only
a few simulations. This approach yields the marginal decrease in air
quality ascribable to a particular emitter and, therefore, indicates what
marginal improvements might be bought by imposing controls. Most frequently
this type of inquiry is served adequately by the simulation of relatively
few pollution scenarios. Choices from an array of technological alterna-
tives for controls will be the result of these simulations; therefore,
the model should have high fidelity for this application.
An example of the "few runs/high fidelity" mode of air quality
simulation is our study of the influence of morning vehicle start emissions
13
on photochemical smog. We sought the answer to the question "What
degradation of air quality is directly ascribable to motor vehicles
starting up in the morning?" Despite many statements to the contrary,
the answer is not merely the fractional emissions due to cold starts.
Diffusion and chemistry introduce non-linearities which preclude simple
scale-up. To investigate the air quality effects, we ran simulations with
and without starting emissions.
The model was used with aerometric data from the Los Angeles basin
to study the buildup of air pollution as it is affected by starting
emissions. The procedure relates meteorological factors, time/space
traffic distributions, and ultraviolet solar radiation with the photo-
chemical atmospheric mechanisms involved in air pollution. (Averaging
over the daily activities of motor vehicles may not give an adequate
description of the most severe conditions.)
Table A.4 shows two of the main findings from the simulation.
First, that air quality effects vary with pollutant and, second, that
194
-------
TABLE A. 4
AIR QUALITY EFFECTS FOR 1974 TRAJECTORY
(Ratios of concentration with cold-start to concentration without cold-
start)
Time
1400 hours
1400 hours
Peak
Species
°3
CO
Spatially
Uniform Start
Distribution
1.039
1.024
1.125
Decentralized Start
Distribution
1.042
1.026
1.136
the geographical distribution of starts has no significant effect. (The
density of morning starts for the nonuniform cases was assumed to be
three times as high at the outer edge of populated areas as it was at the
Federal building downtown varying linearly in a radial direction.) The
numerical comparisons drawn from the table depend on the fidelity of the
model. Some significance can be attached to the larger effect on peak
CO than that on photochemical pollutants. The combined action of reaction
and diffusion attenuates the start-effect for 0 and N02 . These pat-
terns emerged after only a very few simulations were carried out.
A.3.2 Moderate Number of Runs/Moderate Fidelity
Studies of local problems over a wide range of conditions demand
more runs than the example described above. On the other hand, the
reduction in scale from regional to local permits us to use a less detailed
physical and chemical formulation from a simplified version of the photo-
chemical/diffusion model. Exemplifying this type of approach is the analysis
of air quality impact for a proposed high-capacity roadway. The larger
number of cases arises from a multiplicity of factors derived from a many-
dimensioned parameter space. For several miles of roadway near an urban
area, background source intensity and receptor sensitivity might both vary
widely. Hourly traffic loadings change sharply from hour to hour and
195
-------
meteorological conditions can exhibit large seasonal variations. Every
year, the emission characteristics of the vehicle population are altered
by the replacement units that have current control systems. Finally, the
impact of the roadway can be fairly assessed only if we simulate the route
corridor with the facility and compare this with the alternative of
traffic diversion over the surrounding network of surface streets. Indeed,
these variations of parameters multiply into literally hundreds of
specific cases.
An example of a reduced-fidelity model suitable for this task is
LAPS, a code that we have recently developed and put into operation.
The key simplification leading to the efficiency features of the model
is the choice of coordinate system. The Lagrangian frame of reference
is chosen so that the downwind distance coordinate is replaced with time,
with air parcels traced through a streamline system. This coordinate
system is illustrated in Fig. A.7. Streamline curvature can be neglected
in the local areas of freeways; hence, for the free flow above the
roadway, the air parcel representations are planar control surfaces moving
along streamlines. Each control surface that sweeps with the wind along
the streamlines has superimposed upon it a spatial grid. This grid con-
sists of intervals in the vertical and crosswind directions. Finite-
difference methods are used for the vertical diffusion differential
equations. In particular, the Crank-Nicolson technique is employed. For
diffusion in the lateral direction, Gaussian dispersion is employed.
Horizontal diffusion, therefore, is treated as an algebraic correction
that runs concurrently with the finite-difference solutions of the vertical
diffusion equations, permitting spatial variation of the vertical
diffusion coefficient. An optional feature of LAPS is the inorganic
portion of the smog chemical mechanism involving the NO/0 /NO,, cycle.
With two dimensional diffusion and limited chemistry, LAPS is considerably
faster running than our regional code DIFKIN.
—- —
Local Air Pollution Stimulator.
DIFfusion and KINetics.
196
-------
•— Ax = uAt
MOVING CONTROL PLANE
Figure A.7- LAPS Coordinate System
Inputs to the program consist of wind direction, wind speed, vertical
and horizontal diffusion parameters and the whole set of descriptors that
characterize emission sources. Output values of concentrations as func-
tions of time are obtainable by specifying what receptor locations are
of interest. For example, in a particular environmental impact study,
the concent-ration on a nearby school playground might be desired. If that
were the case, the model could give a history of concentration at that
location.
Stagnant conditions are calculated by determining the dispersion
from the roadway under zero wind (but not zero diffusion). This is done
by centering the mesh on the roadway and computing the pollutant spread
in a plane normal to the road centerline.
197
-------
A set of three typical cases serves to demonstrate the capability
of the GRC model to simulate the dispersion of nonreacting pollutants in
the vicinity of a freeway with various wind directions. The freeway, a
six-lane depressed section shown in Fig. A.8, was assumed to carry 100,000
vehicles/day. Using the geometry shown in Fig. A,9, a wind speed of
1.3 mph was simulated at angles (6) of 0°, 30°, and 90° to the roadway.
The resulting CO distributions predicted by the model are shown in
Fig. A.10.
As an example of the speed of the LAPS model, the problem of computing
the CO concentration at 100 points in a vertical plane normal to a free-
way for 8 hours of real time required 3 minutes of central-processor time
on a Control Data 6400 computer. The computation interval for this problem
was 0.1 minute; the output intervals were 1 minute for ground concentra-
tion profiles and 30 minutes for vertical data maps.
To illustrate how the fidelity of the model captures the ozone-
depression effect near a freeway, we can examine the same input conditions
as above for a one mile per hour crosswind. Background levels are
chosen to be typical of the Los Angeles basin well into a midsummer day.
Figure A.11 shows the reduction in ozone as the nitric oxide from the
vehicles mixes in and feeds the NO + 0« -»• NO- + 0 reaction. Downwind
of the roadway, ambient air dilutes the emissions with air containing a
higher level of ozone. Consequently, the nitric oxide decreases back
down to a level near ambient.
Since the chemistry is handled by a simple algorithm based on quasi-
stationary state, this technique is suitable for running the many cases
we will encounter for future problems of air quality impact evaluations.
Turbulent mixing effects of vehicle wakes are treated internally by
aerodynamic formulas for the locally enhanced diffusion coefficients.
Source geometries are not restricted to roadways. Airports, central power
stations and other source concentrations can also be treated as inputs to
the LAPS code.
198
-------
Figure AiS. Cross Section of Depressed Six-Lane Freeway
70 m SOURCE STRIP WIDTH
to
o
03
to
Figure A.9. Wind-Oriented Coordinate System
199
-------
.
O
I
9 = 0'
I— ROADWAY -—|
ZERO BACKGROUND
WIND SPEED = 1.3 mph
1971 VEHICLE MIX
TRAFFIC: 100,000 VEHICLES/DAY
0= 90°
0 50 100 150 200
NORMAL DISTANCE FROM ROADWAY CENTERLINE, m
Figure A.10. CO Concentration Profiles Normal to Roadway at Various
Wind Aspect Angles
100
I
g 10
z
EMISSIONS MODEL: .CALIFORNIA VEHICLE MIX IN 1971
ROADWAY DATA: EIGHT LANE FREEWAY CARRYING
100,000 VEHICLES/DAY
1 mph WIND
Figure A.11. Ozone and Nitric Oxide in an Air Mass Moving Over a Roadway
200
-------
A-3.3 Many Runs/Low Fidelity
The deeper we penetrate into future problems of pollution abatement,
the more respect we gain for the sometimes subtle, but tightly coupled
problems of economic externalities and political realities. Staggering
social costs are now beginning to appear on the horizon as the burden we
must bear to improve the environment. The job of standard-setting, which
has nearly run through its first round, only sets quantitative targets
in this quest for better air and water quality. Actions to achieve the
goals result in multiple feedbacks of money and material, that may either
damage or benefit humanity depending on what value system we adopt.
C Q
Models such as the Implementation Planning Program have endeavored
to relate all of these factors for the steps needed to improve air quality.
Since the air pollution simulation is only a small part of an ensemble
of logical functions, it must, of necessity, be highly simplified. The
sacrifice of fidelity in these applications is believed to be well justi-
fied by the need for many runs. For its scope (particulates and SO
from stationary sources) and for the limited availability of data, the
IPP simulation did an admirable job of laying out the significant issues.
Its weaknesses highlighted the specific needs for better quantitative
information, particularly in the area of damage assessment and the
inclusion of mobile sources.
Future models of this type will need chemistry in the air module
because nearly all pollutants of interest are reactive. Refined cost figures
will be needed in the economic modules to reflect advances in control tech-
nology. Aggregate damage indices may be incorporated in the form of
summations or in integrals such as
where D . = damage index due to the i_th species acting on the jth
type of receptor
201
-------
p = population density of jth receptor
$ = impact function for ith species on jth receptor
13 - -
c. = concentration of ith species
x = location
t = time
a = area
This will be an objective index that goes beyond the mere question "Does
it or does it not exceed ambient standards?" It can be tied to ambient
standards by normalizing the impact function. Say, for example that the
impact of pollutant i on receptor j goes up like the nth power of c
so that the expression
permits a comparative assessment of all species if c, is the ambient
standard. Contributions can be collected by forming a total damage on
receptor j by summing all of the D over all i.
Now the use of this index in systems analysis requires a model
that spreads over all of the receptors, not just a few monitoring stations.
An important feature of future work will be reducing the three-dimensional,
time-dependent air pollution simulations to a usable size. One key to
this reduction lies in a "black box" chemical model. The black box functions
in a chemical sense like the Maxwell Demon operates in the kinetic sense;
i.e., while the Demon sorts out molecules in a certain energy range, the
black box converts NO molecules to N0_ molecules in photochemical smog
according to the hydrocarbon decomposition products that are present.
The black box transfer function will need parameters depending on reactivity j
HC/NOx ratio and, perhaps, temperature, humidity, arid aerosol levels.
The function, which must be obtained from curve fits of extensive kinetie-s
202
-------
simulations, will replace a dozen or more coupled differential equations.
If the Lagrangian fluid dynamic frame is retained in the systems model,
this simplification is essential, since many, many, simulations will be
required to supply values for the integrand of D. over an urban region.
Future applications of this simplified air quality model will find
it coupled with economic input/output models, transportation network simu-
lations, land use models, and energy managment models (possibly all at
the same time). Of course if this line of development is allowed to grow
unchecked, it invariably leads into the fantasy world of some systems
analysts who are unafraid to take on the universe.
Pollution abatement strategies are limited more by human institutions
and resources than by technological advances. Thus, an important class
of future problems in simulation will involve live participants in the
loop. Invariably, actual decisions on incremental changes take directions
other than those selected by system optimization procedures. This happens
because the latter procedures are incomplete with respect to variables
and constraints. With people operating in a simulated abatement scenario,
the quantitative models are used as feedback generators, but they do not
control the action. This application of gaming has been instituted in
the APEX and CITY exercises already under evaluation by the Environmental
Protection Agency. They will serve as training devices as well as
testbeds for policy experiments.
The usefulness of these games will depend heavily on the credibility
of the feedback models that tell the decision maker how much his decision
costs and what effects it will have in many sectors including that of
environmental quality. Nevertheless, these models must simulate many
alternatives rapidly to enhance the value of the game even if it must be
at the expense of some fidelity.
203
-------
A. 4 SUMMATION
Tight implementation schedules and tough regulations have been laid
out to abate pollution. The urgency of policies already adopted has led
to hasty actions in some cases. It is imperative that the research
community concentrate on some of the ill-defined areas that may contain
the key elements of understanding the consequences of air pollution
control decisions.
Mathematical simulation models at the very least provide a logical
framework that highlights the unknowns. At the most, they will serve as
predictive tools in evaluating the impact of rulemaking and decision-making.
As we have moved into the science of air pollution simulation, we have
discovered some serious deficiencies in present-day approaches. The neglect
of heterogeneous processes omits possibly the most important cleanup
processes for oxides of nitrogen. Unexplained shifts in HC/NO ratios
2t
are observed in morning air samples. The largest discrepancies occur on
the worst smog days. Turbulent chemical kinetics are untouched in con-
temporary simulation approaches. These effects may well be responsible
for apparent shifts several hundred percent away from quasiequilibrium
states believed to govern the major pollutants., ozone and oxides of
nitrogen in urban environments.
This paper has assembled some suggestions for the attack on each
problem. These attacks are based on systematic gathering of observational
evidence followed by careful data analysis feeding into refinements or
corrections to the simulation models. Unfortunately, many uses will be
demanded of the models before these issues are faced. Some of the expres-
sions derived in the preceding sections will serve at least as criteria
for setting rough levels of confidence in existing models. The urgency
of the problems at hand must be used as a stimulant for the needed research
rather than an excuse to overlook the deficiencies in our present under-
standing of the problems.
204
-------
With the mature development of certain air quality simulation tech-
niques, a hierarchy of models will emerge embracing a wide range of fidelity
levels. The large variety of applications anticipated places demands on
operating speed in some cases while other cases are characterized by needs
for precision in certain types of predictions. The examples cited in the
closing section highlight these differences.
Despite consultant advertising that touts "complete modeling capa-
bilities for air pollution simulation," we can still see some exciting
possibilities for new research. This new work will fill significant
voids in the basic fiber of which the "complete" models are made. But
simply building models is not enough. They must be molded into useful
forms by creating special algorithms, by inventing approximations, by
designing input/output structures that meet the dynamic demands of air
quality management.
205
-------
REFERENCES
1- A.Q. Eschenroeder and J.R. Martinez, Concepts and Applications of
Photochemical Smog Models, General Research Corporation TM-1516,
June 1971 (to be published as an ACS Monograph in Advances in
Chemistry).
2. A,P. Altshuller and J.J. Bufalini, "Photochemical Aspects of Air
Pollution: A Review," Photochemistry and Photobiology, Vol. 4, 1964,
pp. 97-146.
3. A.P. Altshuller and J.J. Bufalini, "Photochemical Aspects of Air
Pollution: A Review," Environmental Science and Technology, Vol. 5,
No. 5, January 1971, pp. 39-64.
4. K. Westberg and N. Cohen, The Chemical Kinetics of Photochemical
Smog as Analyzed by Computer, AIAA Third Fluid and Plasma Dynamics
Conference Paper No. 70-753, Los Angeles, June 29-July 1, 1970.
5, T.A. Hecht and J.H. Seinfeld, '^Development and Validation of a
Generalized Mechanism for Photochemical Smog," Env. Science and
Technology, Vol. 6, No. 1, January 1972., pp. 47-57.
6. H. Niki, E.E. Daby and B. Weinstock, "Mechanisms of Smog Reactions",
Advances in Chemistry, 1972 (in press).
7. D.H. Stedman, E.D. Morris, Jr., E.E. Daby, H. Niki, and B. Weinstock,
The Role of OH Radicals in Photochemical Smog, American Chemical
Society Division of Water Air and Waste Chemistry, Chicago, Illinois,
September 13-18, 1970.
8, J.R. Holmes, A.D. Sanchez, and A.H. Bockian, Atmospheric Photochemistry:
Some Factors Affecting the Conversion of NO to NO , Pacific Con-
ference^ on Chemistry and Spectroscopy, San Francisco, October 6-9,
1970.
9. B.W. Gay and J.J. Bufalini, "Nitric Acid and the Nitrogen Balance
of Irradiated Hydrocarbons in the Presence of Oxides of Nitrogen,"
Environmental Science and Technology, Vol. 5, No. 5, May 1971,
pp. 422-425.
10. M. Dodge, private communication, May 26, 1972.
11. M. Dodge, private communication, July 11, 1972.
207
-------
12. P.L. Hanst, "Mechanism of Peroxyacetylnitrate Formation", J. Air
Pollution Control Association, Vol. 21, No. 5, May 1961, pp. 269-271.
13. J.R. Martinez, R.A. Nordsieck, and A.Q. Eschenroeder, Morning
Vehicle-Start Effects on Photochemical Smog, General Research
Corporation CR-2-191, June 1971.
14. F. Stuhl, private communication, Ford Motor Co., Scientific Research
Staff, June 1, 1972.
15. J. Anderson, private communication, University of Pittsburgh, June
7, 1972.
,16. E.A. Sutton, "Chemistry of Electrons in Pure-Air Hypersonic Wakes,"
AIAA Journal, Vol. 6, No. 10, October 1968, pp. 1873-1882.
17. K. Schofield, "An Evaluation of Kinetic Rate Data for Reactions of
Neutrals of Atmospheric Interest," Planetary and Space Sciences,
Vol. 15, 1967, pp. 643-670.
18. P.A. Leighton, Photochemistry of Air Pollution, New York Academic
Press, 1961.
19. G. Schott and N. Davidson, "Shock Waves in Chemical Kinetics: The
Decomposition of NO,, at High Temperatures," Journal of American
Chemical Society, Vol. 80, p. 1841 (1958).
20. S. Jaffe and H.W. Ford, "The Photolysis of Nitric Acid at 3660A and
25°," Journal of Physical Chemistry, Vol. 71, p. 1832 (1967).
21. W.R. Greiner, "Hydroxyl Radical Kinetics VI., Reactions with
Alkanes in the Range 300°-500°K, J. Chem. Phys., 53, 1970, pp.
1070-1076.
22. J. Heicklen, K. Westberg, and N. Cohen, The Conversion of NO to NO
in Polluted Atmospheres, Pennsylvania State University Center for
Air Environment Studies Publication 115-69, July 1969.
23. K. Westberg, N. Cohen, K.W. Wilson, "Carbon Monoxide: Its Role in
Photochemical Smog Formation", Science, Vol. 171, March 12, 1971,
pp. 1013-1015.
24. A.Q. Eschenroeder and J.R. Martinez, Further Development of the
Photochemical Smog Model for the Los Angeles Basin, General Research
Corporation CR-1-191, March 1971.
j
25. A.P. Altshuller, "An Evaluation of Techniques for the Determination
of the Photochemical Reactivity of Organic Emissions", JAPCA, Vol.
16, No. 5, May 1966, pp. 257-260.
208
-------
26. A.Q. Eschenroeder and J.R. Martinez, Analysis of Los Angeles Atmos-
pheric Reaction Data from 1968 and 1969, General Research Corpora-
tion CR-1-170, July 1970.
27. A.Q. Eschenroeder and J.R. Martinez, Mathematical Modeling of Photo-
chemical Smog. General Research Corporation IMR-1210, December 1969.
Also a paper presented at the AIAA Eighth Aerospace Sciences Meet-
ing, January 1970.
28. J.K. Angell, D.H. Pack, L. Machta (R. Dickson, and W.H. Hoecker),
"Three-Dimensional Air Trajectories Determined from Tetroon Flights
in the Planetary Boundary Layer of the Los Angeles Basin", Journal
of Applied Meteorology. Vol. 11, No. 3, April 1972, pp. 451-571.
29. M.A. Estoque, "A Numerical Model of the Atmospheric Boundary Layer",
Journal of Geophysical Research, Vol. 68, 1968, pp 1103-113.
30. C.R. Hosier, "Vertical Diffusivity from Random Profiles", Journal
of Geophysical Research, Vol. 74, No. 28, December 20, 1969, pp.
7018.
31. A.Q. Eschenroeder and J.R. Martinez, A Modeling Study to Characterize
Photochemical Atmospheric Reactions to the Los Angeles Basin Area,
General Research Corporation CR-1-152, November 1969, p. 18.
32. F. Pasquill, Atmospheric Diffusion, Van Nostrand, London, 1962.
33. J. Angell and D. Pack, "Mesoscale Diffusion Derived from Tetroon
Flights", USAEC Meteorological Information Meeting, Atomic Energy
of Canada Limited, Chalk River, Ontario, AECL-2787, September 1967.
34. P. Roberts, P. Roth, and C. Nelson, Contaminant Emissions in the Los
Angeles Basin—Their Sources, Rates, and Distributions, Systems
Applications, Inc., Report 71 SA1-6, March 1971 (Appendix A).
35. Final Report on Phase I, Atmospheric Reaction Studies in the Los
Angeles Basin, Vols. I and II, Scott Research Laboratories, June
30, 1969.
36. F.L. Ludwig, A.E. Moon, W.B. Johnson, R.L. Mancuso, A Practical
Multipurpose Urban Diffusion Model for Carbon Monoxide, Stanford
Research Institute, September 1970.
37. L. Aldaz, "Flux Measurements of Ozone Over Land and Water," Journal
of Geophysical Research, Vol. 74, No. 28, December 20, 1969, pp.
6943-6946.
38. R.E. Inman, and R.B. Ingersoll, "Note on the Uptake of Carbon Mono-
xide by Soil Fungi," Journal of the Air Pollution Control Association,
Vol. 21, No. 10, October 1971, pp. 646-647.
209
-------
39. F.B. Abeles, L.E. Craker, I.E. Forrence, and G.R. Leather, "Fate of
Air Pollutants: Removal of Ethylene, Sulfur Dioxide arid Nitrogen
Dioxide by Soil," Science, Vol. 173, No. 4000, September 3, 1971,
pp. 914-916.
40. L.A. Ripperton, and F.M. Vukovich, "Gas Phase Destruction of Trdpo-
spheric Ozone," Journal of Geophysical Research, Vol. 76, No. 30,
October 20, 1971, pp. 7328-7333.
41. D.A. Lundgren, "Atmospheric Aerosol Composition and Concentration
as a Function of Particle Size and of Time," Journal ~of the Air Pol-
lution Control Association, Vol. 20, No. 9, September 1970, pp.
603-608.
42. J. Cholak, L.J. Schafer, D.W. Yaeger, and R.A. Kehoe, "The Nature
of the Suspended Matter," Chapter VIII in Air Pollution Foundation
Report No. 9, An Aerometric Survey of the Los Angeles Basin, August-
November 1954.
43. P.A. Leighton, and W.A. Perkins, Photochemical Secondary Reactions
in Urban Air, Air Pollution Foundation Report No. 24, August 1958.
44. E. Daby, Ford Motor Company, private communication. (Full results
to be published in Journal of Air Pollution Control Association,
Vol. 22, No. 4, April 1972.)
45. A.P. Altshuller, S.L. Kopczynski, W.A. Lonneman, T.L. Becker and R.
Slater, "Chemical Aspects of the Photo-oxidation of the Propylene-
Nitrogen Oxide System," Environmental Science and Technology, Vol. 1,
No. 11, November 1967, pp. 889-914.
46. E.A. Schuck and E.R. Stephens, "Oxides of Nitrogen," Advances in
Environmental Sciences, Vol. 1, New York, John Wiley and Sons,
1969, pp. 73-118.
47. R..G. Batt, T. Kubota and J. Laufer, Experimental Investigations of
the Effect of Shear-Flow Turbulence on a Chemical Reaction, AIAA
^Reacting Turbulent Flows Conference Paper, San Diego, June 1970.
48. D. Bugnolo, "Effects of a 'rMixing-in-Gradient' on the Spectrum of
the Electronic Density in a Turbulent Weakly Ionized Gas," Journal
of Geophysical Research, Vol. 70, No. 15, August 1965, pp. 3725-3734.
49. S. Corrsin, "Statistical Behavior of a Reacting Mixture in Isotropic
Turbulence," Phys. Fluids, Vol. 1, No. 1, January-February 1958 pp
42-47.
50. S. Corrsin, "The Reactant Concentration Spectrum in Turbulent Mixing
with a First-Order Reaction," J. Fluid Mech., 1961, pp. 407-416.
210
-------
51. S. Corrsin, "Further Generalization of Onsager's Cascade Model for
Turbulent Spectra," Phys. Fluids, Vol. 7, No. 8, August 1964, pp.
1156-1159.
52. J. Hinze, Turbulence; An Introduction to Its Mechanism and Theory,
McGraw Hill, New York, 1959, p. 224.
53. Air Quality Implementation Planning Program, TRW Systems Group
SN11130, November 1970.
211
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
-R4-73-012a
3. Recipient's Accession No.
4. Title and Subtitle
Evaluation of a Diffusion Model for Photochemical
Smog Simulation
5. Report Date
October 1972
6.
. Authorfs)
A.Q. Eschenroeder, J.R. Martinez, R.A. Nordsieck
8. Performing Organization Kept.
No- CR-1-273
Performing Organization Name and Address
General Research Corporation
P. 0. Box 3587
Santa Barbara, California 93105
10. Project/Task/Work Unit No.
11. Contract/Grant No.
68-02-0336
12. Sponsoring Organization Name and Address
ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, North Carolina 27711
13. Type of Report & Period
Covered
Final
14.
15. Supplementary Notes
16. Abstracts Exten5ive jimprovements have characterized this evaluation of the GRC Photo-
chemical/Diffusion model. Despite the limitations of smog chamber experimental data,
they have served an essential purpose toward updating the kinetics portion of the
model. Consistency of rates and reactivities is now achievable using recently measured
coefficients for a wide variety of systems. Model methodology revisions have enhanced
the realism of the advective and diffusive descriptions. Previous assumptions regard-
ing transverse (cross-streamline) horizontal diffusion have been confirmed by an ex-
haustive series of parametric tests. Photochemical/diffusion validations were success-
ful for trajectories occurring during four days of the 1969 smog season in Los Angeles.
Our measure of success is concentration-history fidelity with a minimum of adjustments
of diffusion parameters. (Chemical coefficients were scaled from the smog chamber
studies and held fixed for the simulations carried out to date). Future directions
for air pollution model development are discussed in detail in an appendix as informa-
tion supporting the experimental recommendations.
17. Key Words and Document Analysis. 17a. Descriptor
Air pollution
Mathematical models
Photochemical reactions
Diffusion
Advection
Trajectories
Reaction kinetics
Smog
Chambers
Validity
17b. Identifiers/Open-Ended Terms
Diffusion models
Atmospheric modeling
17c. COSAT1 Fjeld/Group 13B
18. Availability Statement
JTIS-35 IREV. 3-72)
Unlimited
19. Security Class (This
Report)
UNCLASSIFIED
20. Security Class (This
Page
UNCLASSIFIED
[1- No. of Pages
230
G. P. O. 1973 — 746-772 / 4196. REGION NO. 4
22. Ptice
USCOMM-DC 14952-P
-------
INSTRUCTIONS FOR COMPLETING FORM HTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for the Federal Government,
PB-180 600).
1. Report Number. Each individually boupd report shall carry a unique alphanumeric designation selected by the performing
organization or provided by the sponsoring organization. Use uppercase letters and Arabic numerals only. Examples
FASEB-NS-S7 and FAA-RD-68-09.
2. Leave blank.
3. Recipient's Accession Number. Reserved for use by each report recipient.
4 Title and Subtitle. Title should indicate clearly and briefly the subject coverage of the report, and be displayed promi-
nently. Set subtitle, if used, in smaller type or otherwise subordinate it to main title. When a report is prepared in more
than one volume, repeat the primary title, add volume number and include subtitle for the specific volume.
5- Report Date. P!ach report shall carry a date indicating at least month and year. Indicate the basis on which it was selected
(e.g., date of issue, date of approval, date of preparation.
6- Performing Organization Code. Leave blank.
7. Author(s). Give name(s) in conventional order (e.g., John R. Doe, or J.Robert Doe). List author's affiliation if it differs
from the performing organization.
8. Performing Organization Report Number. Insert if performing organization wishes to assign this number.
9- Performing Orgonization Name and Address. Give name, street, city, state, and zip code. List no more than two levels of
an organizational hierarchy. Display the name of the organization exactly as it should appear in Government indexes such
as USGRDR-I.
10. Project/Task/Work Unit Number. Use the project, task and work unit numbers under which the report was prepared.
11. Contract/Grant Number. Insert contract or grant number under which report was prepared.
12- Sponsoring Agency Name and Address. Include zip code.
13. Type of Report and Period Covered. Indicate interim, final, etc., and, if applicable, dates covered.
14- Sponsoring Agency Code. Leave blank.
15. Supplementary Notes. Enter information not included elsewhere but useful, such as: Prepared in cooperation with .
Translation of . Presented at conference of . . To be published in . . Supersedes . . . Supplements
3$. Abstract. Include a brief (200 words or less) factual summary of the most significant information contained in the report.
If 'he report contains a significant bibliography or literature survey, mention it here.
17. Kay Words and Document Analysis, (a). Descriptors. Select from the Thesaurus of Engineering and Scientific Terms the
proper authorized terms that identify the major concept of the research and are sufficiently specific and precise to be used
as index entries for cataloging.
(b). Identifiers and Open-Ended Terms. Use identifiers for project names, code names, equipment designators, etc. Use
open-ended terms written in descriptor form for those subjects for which no descriptor exists.
(c). COSATI Field/Group. Field and Group assignments are to be taken from the 1965 COSATI Subject Category List.
Since the majority of documents are mulndisciplinary in nature, the primary Field/Group assignment(s) will be the specific
discipline, area of human endeavor, or type of physical object. The appHcation(s) will be cross-referenced with secondary
Field/Group assignments that will follow the primary posting(s).
18. Distribution Statement. Denote releasability to the public or limitation for reasons other than security for example "Re-
lease unlimited" Cite any availability to the public, with address and price.
19 & 20. Security Classification. Do not submit classified reports to the National Technical
21. Number of Pages. Insert the total number of pages, including this one and unnumbered pages, but excluding distribution
list, if any.
22. Price. Insert the price set by the National Technical Information Service or the Government Printing Office if known
ORM NTIS-33 (REV. 3-72) U SC OMM-DC V4SS 2-P?2
------- |