EPA-R4 73-013a
February 1973 Environmental Monitoring Series
Controlled Evaluation
Of The Reactive
Environmental
Simulation Model (REM)
Volume I - Final Report
Office of Research and Monitoring
U.S. Environmental Protection Agency
Washington, D. C. 20460
-------
EPA-R4-73-013a
Controlled Evaluation
Of The Reactive Environmental
Simulation Model (REM)
Volume I - Final Report
BY
Lowell G. Wayne, Allan Kokin, and Melvin I. Weisburd
Pacific Environmental Services, Inc.
2932 Wilshire Boulevard (Suite 202)
Santa Monica, California 90403
Contract No. 68-02-0345
Program Element No. 1A1009
Project Officer: Ralph C. Sklarew
Meteorology Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
Office of Research and Monitoring
ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D. C. 20460
February 1973
-------
This report has been reviewed by the Environmental Protection Agency
and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
11
-------
Ill
TABLE OF CONTENTS
Pa
ABSTRACT . ............................. • Xi
CHAPTER I. MODELING PHOTOCHEMICAL POLLUTION ........ .... I.I
A. INTRODUCTION ................ .... I.I
B. APPROACH TO MODEL DEVELOPMENT ............ 1.2
1. Potential Uses of Models ............ 1.2
2. Input-Output Considerations ..... ...... 1.2
3. Principles of Simulation .... ........ 1.3.
4. Application for Control Strategy Purposes- • • • 1.4
5. Application for Shor*" rm Forecasting ..... ' 1.6
6. Research, Training and v,Jier Applications. ... 1.8
C. RATIONALE OF THE REACTIVE ENVIRONMENTAL
SIMULATION MODEL (REM) ............... 1.9
1. General Approaches to Model Building . ..... 1.9
2. Reasons for Following the Conservative Approach. 1.9
3. Structure of the Simulation Process ....... 1. 11
D. ROLE OF CHEMICAL REACTIONS IN DETERMINING
AIR QUALITY ..................... 1.12
1. Introduction .................. 1.12
2. Principles of Chemical Kinetics in Relation. . .
to Simulation of Air Quality .......... 1.14
3. Distribution of Contaminants in the Atmosphere,
Resulting from Chemical Reactions and Transport. 1.16
4. Chemical Reactions in the Atmosphere in
Relation to Control Strategy Planning ...... 1.24
5. Formulation -.-f a Chemical Reaction Mechanism
in Terms of ..leuentary Reactions ........ 1.26
E. ROLE OF ATMOSPHERIC TURBULENCE IN SIMULATION OF
AIR QUALITY. . . ................... 1.31
F. STATUS AND FUTURE DEVELOPMENT OF REM ........ 1.33
1. History of REM ................. 1.33
2. Current Project ................. 1.36
3. Recommended Improvements ............ 1.38
G. EVALUATION CRITERIA ................. 1.41
CHAPTER II. DESCRIPTION OF THE REACTIVE ENVIRONMENTAL SIMULATION
MODEL (REM) ....................... . II. 1
A. INTRODUCTION ........... ......... II. 1
B. GRID SYSTEM ................... . . II. 4
-------
IV
C. PROGRAM CODE AND MACHINE USE II. 4
D. OPERATION OF REM II. 4
E. CONTROL STRATEGIES II. 6
F. REM MODULES II. 7
1. Initialization Program II.7
2. Numerical Integration II. 8
3. Chemical Kinetics II.8
4. Meteorology II. 9
5. Barrier Check 11.11
6. Ultraviolet Module 11.13
7. Source Emissions 11.13
8. Output Routine 11.21
9. Plotting Routine 11.21
10. Reverse Trajectory Program 11.21
CHAPTER III. SIMULATION OF IRRADIATION CHAMBER EXPERIMENTS III.l
A. INTRODUCTION III.l
B. SIMULATION OF PROPYLENE-ETHANE-OXIDES OF NITRO-
GEN SYSTEM III. 7
1. By Original Mechanism, Ml III. 7
2. By Revised Mechanism, M2 III. 12
C. SIMULATION OF AUTO EXHAUST PHOTO-OXIDATION SYSTEMS . III.31
D. SIMULATION OF TOLUENE-CONTAINING SYSTEMS III.37
1. The Toluene-Oxides of Nitrogen System III.38
2. The Toluene-Butane-Oxides of Nitrogen System . . III.41
3. Discussion of Results with Toluene-Containing
Systems . III. 41
E. SIMULATION OF PHOTO-OXIDATION PRODUCTS IN THE
IRRADIATION EXPERIMENTS III. 43
F. SUMMARY AND CONCLUSIONS III.43
CHAPTER IV. SIMULATION OF PHOTOCHEMICAL AIR POLLUTION EPISODES ... IV.1
A. INTRODUCTION IV. 1
B. CARBON MONOXIDE SIMULATION IV.2
C. COMPLETE ATMOSPHERIC SIMULATIONS IV.14
1. Simulation by Trajectories for 1330 PST Arrival. IV.15
2. Extended Simulation of a Single Episode IV. 17
3. General Accuracy of Atmospheric Simulations. . . IV.24
D. CONCLUSIONS IV. 30
I
APPENDIX RESULTS OF REM SIMULATIONS A.I
-------
LIST OF FIGURES
Figure Page
II.1 PES Reactive Environmental Model (REM) II.2
II.2 Schematic of the PES Reactive Environmental Model (REM) . II.3
II.3 REM Grid System for the Los Angeles Basin ......... II.5
II.4 Demonstration of. Barrier Check Technique . . . . 11.12
II.5 Temporal Distribution of Daily Street Traffic 11.14
II.6 Temporal Distribution of Daily Freeway Traffic ....... 11.15
II.7 Distribution of Street Traffic by Grid Area
(thousands of vehicle miles per day) 11.16
II.8 Distribution of Freeway Traffic by Grid Area
(thousands of vehicle miles per day) 11.17
II.9 Distribution of Reactive Hydrocarbon Emissions
(average kilograms/hour) by Grid Area. .... 11.18
11.10 Distribution of Oxides of Nitrogen Emissions Taken
as Nitric Oxide (average kilograms/hour) by Grid Area . . 11.19
11.11 Distribution of Less Reactive Hydrocarbon
Emissions (average kilograms/hour) by Grid Area 11.20
11.12 Sample REM output 11.21
11.13 Digital Plot of Primary Pollutant Time vs.
Concentration Histories 11.23
III.l Concentration of Ethane Related to Time of Sampling,
Experiment 321 (propylene, ethane, and oxides of
nitrogen system). III.5
III.2 Time Profiles of Nitrogen Dioxide Concentrations,
Experiments 222 and 233 (diluted auto exhaust system).
Smoothed experimental profiles III.6
-------
VI
LIST OF FIGURES
Time Profiles of Contaminant Concentrations,
Experiment 329. (Original Mechanism) III.10
III.4 Time Profiles of Contaminant Concentrations,
Experiment. 325. (Original Mechanism) .....*... III.11
III.5 Time Profiles of Contaminant Concentrations,
Experiment 329. (Revised Mechanism) III.15
III.6 Time Profiles of Contaminant Concentrations,
Experiment 325. (Revised Mechanism) III.17
III.7 Time Profiles of Nitrogen Dioxide Concentration
Simulating Experiment 329. , , , III. 19
III.8 Time Profiles of Oxidant (Ozone) Concentration
Simulating Experiment 329 III.20
III.9 Time Profiles of Contaminant Concentrations,
Experiment 318 III.21
III.10 Time Profiles of Nitrogen Dioxide Concentration
Simulating Experiment 318. .... III.23
III.11 Time Profiles of Oxidant (Ozone) Concentration
Simulating Experiment 318 III.24
III.12 Time Profiles of Nitrogen Dioxide Concentration
Simulating Experiment 318 III.25
III.13 Time Profiles for Oxidant (Ozone) Concentration
Simulating Experiment 318 III.26
III.14 Time Profiles of Nitric Oxide Concentration
Simulating Experiment 318 III.27
III.15 Time Profiles for Propylene Concentration
Simulating Experiment 318. ..... III.28
III.16 Time Profiles of Contaminant Concentrations,
Experiment 321 III.30
-------
VI1
Figure Page
III.17 Time Profiles of Contaminant Concentrations,
Experiment 222 III.33
III.18 Time Profiles of Contaminant Concentrations,
Experiment 233 III.34
HI.19 Time Profiles of Contaminant Concentrations,
Experiment 231 III.35
III.20 Time Profiles of Contaminant Concentrations,
Experiment 224 III.36
III.21 Time Profiles of Contaminant Concentrations,
Experiment 250 III.40
IV.1 Comparison of Calculated vs. Observed Values of
Carbon Monoxide at Air Monitoring Stations on
September 11, 1969. IV.4
IV.2 Comparison of Calculated vs. Observed Values of
Carbon Monoxide at Air Monitoring Stations on
September 29, 1969 IV.5
IV.3 Comparison of Calculated vs. Observed Values of
Carbon Monoxide at Air Monitoring Stations on
September 30, 1969 IV.6
IV.4 Comparison of Calculated vs. Observed Values of
Carbon Monoxide at Air Monitoring Stations on
October 29, 1969 IV. 7
IV.5 Comparison of Calculated vs'. Observed Values of
Carbon Monoxide at Air Monitoring Stations on
October 30, 1969 IV. 8
IV.6 Comaprison of Calculated vs..Observed Values of
Carbon Monoxide at Air Monitoring Stations on
November 4, 1969. IV. 9
IV.7 Comparison of Calculated vs. Observed Values of
Total Oxidant and Carbon Monoxide at Downtown
Los Angeles on October 29, 1969. IV. 18
-------
Vlll
Page
Comparison of Calculated vs. Observed Values of
Nitrogen Dioxide, Nitric Oxide, Total Oxidant and
Carbon Monoxide at Azusa on October 29, 1969 IV. 19
IV.9 Comparison of Calculated vs. Observed Values of
Nitrogen Dioxide, Nitric Oxide, Total Oxidant and
Carbon Monoxide at Whittier on October 29, 1969. . . . IV.20
IV.10 Comparison of Calculated vs. Observed Values of
Nitrogen Dioxide, Nitric Oxide, Total Oxidant and
Carbon Monoxide at Pomona on October 29, 1969 IV. 21
IV.11 Scatter Plot of Calculated vs. Observed Carbon
Monoxide for the Atmospheric Simulations IV. 25
IV.12 Scatter Plot of Calculated vs. Observed Nitrogen
Dioxide for the Atmospheric Simulations IV. 27
IV.13 Scatter Plot of Calculated vs. Observed Total
Oxidant for the Atmospheric Simulations IV. 29
A.I Trajectories for September 11, 1969 A. 14
A.2 Trajectories for September 29, 1969 A.15
A.3 Trajectories for September 30, 1969. A.16
A.4 Trajectories for October 29, 1969 A. 17
A.5 Trajectories for October 30, 1969 A. 18
A.6 Trajectories for November 4, 1969 A.19
-------
IX
LIST OF TABLES
Table Page
III.l Parameters Used for Simulation of Various Experiments . . III.3
III.2 Irradiation Chamber Experiments in the Propylene-
Ethane-Oxides of Nitrogen System III.8
HI.3 Original Mechanism (Ml) for the Chemical Dynamics
Module (with Rate Factors for Exp. 329 and 325) III.9
III.4 Revised Mechanism (M2) for the Chemical Dynamics
Module (with Rate Factors for Exp. 329 and 325) III.13
III.5 Irradiation Chamber Experiments with Automobile
Exhaust Gases 111,32
III.6 Irradiation Chamber Experiments in the Toluene-
Oxides of Nitrogen System III.39
III.7 Irradiation Chamber Experiments in the Toluene-
Butane-Oxides of Nitrogen System III.42
III.8 Comparison of Estimated vs. Observed Concentrations
for Formaldehyde, Acetaldehyde, and Peroxyacetyl
Nitrate (after 6 hours irradiation) III.44
IV.1 Comparison of Estimated Mid-Hour Concentrations to
Observed Hourly Average Concentrations for Carbon
Monoxide, for Six Smog Days-at Different Receptor
Locations ..... IV.12
IV.2 Comparisons of Estimated. Mid-Hour Concentrations to
Observed Hourly Average Concentrations for Carbon
Monoxide, for Four Receptor Locations on Different
Smog Days IV. 13
IV.3 Comparison of Estimated Mid-Hour Concentrations to
Observed Hourly Average Concentrations for Carbon
Monoxide, Nitrogen Dioxide, Nitric Oxide and Ozone,
for 1330 PST Trajectories IV.16
IV.4 Comparison of Dosage of Various Contaminants as Esti-
mated from Simulation Model to Dosage Estimated from
Air Monitoring Data, Four Locations, Six Hours on
October 29, 1969 . IV. 23
-------
Tables Page
A.I Carbon Monoxide Simulations A.2
A.2 Photochemical Pollutant Simulation Trajectories ...... A.9
-------
XI
ABSTRACT
The development and validation of an operational version of the
Reactive Environmental Simulation Model (REM) were completed. REM was
specifically designed to handle large chemical mechanisms to assess
the impact on air quality of air pollution control devices, fuels,
propulsion systems, stationary sources, and transportation systems
where thorough evaluation of emissions, emission constituents and
reaction rates are required. The version delivered to the U.S. EPA
under this contract contains a mechanism involving 32 reactions, 12
accumulating species, and 12 non-accumulating species. Larger or smal-
ler mechanisms, also, can be readily inserted into REM for specified
purposes.;
REM is based on a Lagrangian moving coordinate system which enables
the numerical simulation of the chemical reactions that take place in
a parcel!or column of air moving along a dynamic wind trajectory. The
trajectory approach gives REM considerable flexibility and is adap-
table for use in a. number of practical operational situations. These
include short-term air quality forecasting; analysis of the impact of
sources of air pollution at various designated locations on the air
quality |at specified receptor points; interpolation of contaminant
concentrations and dosages at locations not covered by air monitoring
stations, and control strategy evaluations.
REM|contains such user features as reverse and forward trajectory
routines; automatic and objective interpolation from input emission
inventory, meteorological and air quality data bases; a chemical dynam-
ics routine capable of accomodating mechanisms based on elementary
chemical reactions; and automatic estimation of mixing depth and solar
irradiance based on input of local weather and sun angle data.
-------
XI1
REM has attained a running time which makes it cost-effective for
practical use. On an IBM System 370/155, the real-time-to-simulation
ratio for the 31-step mechanism is 150:1; for carbon monoxide alone it
is 3000:1; for some of the shorter mechanisms available it should run
more than 300:1. The program, also, is user-oriented in that it pro-
vides simple input procedures, user documentation, receptor point and
time-of-day selectivity, flexibility in treating specific problems, and
ability to conveniently select any of an infinite number of trajectories
on any number of days of interest. The modular construction of REM,
also, makes it easy to add, replace, or delete individual modules.
REM can be used as an unlimited receptor point model and to chronicle
emission inputs contributing to the air quality at any receptor point.
The validation record of REM over a large number of runs may be
summarized as follows:
• CO, less than a factor of 2 in more than 80% of the
comparisons; 40$ agreement within one part per million.
• 03, within a factor of 2, 75% of the comparisons.
• NO, within a factor of 2, 75% of the comparisons; 90%
agreement within .02 ppm.
• N02, within a factor of 2, 60% of the comparisons.
The shapes of predicted time profiles of contaminant concentrations
are generally similar to those observed at air monitoring stations.
The ease and economy at which REM can be utilized under a variety
of data availability situations make potential improvements in accuracy
and application possible on a cost-effective basis.
-------
I.I
Chapter 1
MODELING PHOTOCHEMICAL POLLUTION
A. INTRODUCTION
This final report covers the work performed under U.S. Environmental
Protection Agency Contract 68-02-0345 to evaluate the Reactive Environ-
mental Simulation Model (REM). REM originally was developed under
Contract CPA 22-69-108, June, 1970. It was subsequently refined under
Contract CPA 70-151, January, 1971.
The evaluation of atmospheric simulation models should consist of
three major parts: (1) evaluation of the validity and applicability
of the modeling approach; (2) validation of the model against known
"real" world conditions, including field measurements; and (3) evalua-
tion of the ease and economy of use of the model.
Given the uncertainties and complexities inherent in environmental
modeling, confidence in a model and in its potentialities may be enhanced
if it is based on established scientific principles to the maximum
feasible extent, and if the empirical aspects are formulated in a
plausible way, consistent with established principles as well as all
available relevant data.
At the same time, the complexity of the problem of simulating air
quality is such that a truly comprehensive air quality model for photo-
chemical contaminants is, at present, out of reach. Further, if such a
model were in existence, its use for most of the ordinary purposes of
simulation would, in all likelihood, be prohibitively expensive. It is
appropriate, in evaluating a model, to consider its characteristics in
relation to its potential uses.
-------
1.2
The evaluation of an air pollution model, therefore, must include a
comprehensive understanding of the approach and its implications. It
should include assessment of both the current status of the model and
its projected line of development.
B. APPROACH TO MODEL DEVELOPMENT
1. Potential Uses of Models
Consideration of the intended use of a photochemical air quality
model may indicate which aspects of the model should be emphasized—
that is, developed in detail—and which may be,de-emphasized.in the
interest of ease of operation or cost-efficiency. If an intended use
is specified, model design may be optimized in view of the needs of the
responsible user and in view of his ability to understand, apply, and
adapt the model for this use.
Currently identified potential uses for air quality models fall
generally into three categories, which are discussed below in further
detail:
1. Control strategy evaluation
2. Short-term forecasting of air quality
3. Research, training and other applications
2. Input-Output Considerations
Any air quality model, for whatever intended use, has as its purpose
the prediction of concentrations of contaminants in the atmosphere,
based.upon input information specific to certain situations of interest.
Predicted concentrations will depend strongly on the time-frame invol-
ved. Thus, for control strategy evaluation, the appropriate time frame
is likely to be a year or one of the seasons of a year, while fore-
casting of air quality for emergency warning systems requires a time-
frame of hours.
-------
1.3
Similarly, the spatial scale of predicted values is /.n important
factor for model design. Predicting conditions in the itauediate vicin-
ity of a roadway or industrial pollution source may require a model
substantially different from one to be used for evaluating the air
quality of an air pollution control region.
Input factors consist, basically, of three types of data: contami-
nant emissions information, initial pollutant concentrations, and weather
information. The time-frame and spatial scale of these input factors
must be consistent with the requirements associated with the intended
use of the model. Thus, a climato-logicai summary of -the weather for a
given region will not-be adequate input for use in a model for short-
term forecasting of air quality. ' •
3. Principles of Simulation
Operating on these elements, the model simulates the forces of
nature which determine actual concentrations of contaminants in a real
region. To the extent that these forces are predictable and understood,
they can be represented by mathematical relations or simulated by
numerical processes and the results may be used with confidence. To
the extent that the relevant forces of nature are not predictable or
not understood, the validity of simulation is unavoidably impaired.
For these contingencies, empirical relations or processes are customar-
ily devised and adjusted to produce results having more or less
reasonable similarity to values observed in the atmosphere.
The greater the content of empiricism embodied in the simulation,
the smaller is the validity of the model and the less confidence may
be placed in the use of the results. For this reason, it is desirable
to base the simulations on established physical and chemical laws wherever
possible. Where necessary in the interest of cost reduction, the next
preferred alternative is to use simplifications based on these laws.
Least desirable, but sometimes necessary, is the resort to empiricism.
-------
1.4
4. Application for Control Strategy Purposes
In evaluating alternative strategies for the abatement of air pollu-
tion problems, the user attempts to optimize estimated air quality
subject to constraints imposed by nature and by society. His time-frame
is often of the order of magnitude of years, in terms of implementation
of the strategies to be compared. With respect to the common sources
of photochemical contaminants, the applicable spatial scale is likely
to be also rather large; for example, the transportation system for a
large city or an entire region may be an important factor.
However, the selected strategy must provide for protection of the
inhabitants of the region against concentrations of certain contaminants
exceeding ambient air standards for periods as short as one hour. For
simulating individual pollution episodes to this degree of detail, in
view of ordinary minimum wind speeds of about 2 miles per hour, the
appropriate spatial resolution is for characteristic distances of 2
miles.
For evaluating strategies for pollution abatement, an air quality
model will be used in a comparative mode. That is, given certain fixed
assumptions regarding the weather in the region under study, air quality
parameters (i.e., contaminant concentrations) will be predicted for
various hypothetical distributions of contaminant emissions, and the
main interest of the exercise lies in comparing the predicted values of
these concentrations. The model output, therefore, must be sensitive
to differences in the details of the emission inventories which are
provided as input, and it must reflect as accurately as possible the
chemical processes in the atmosphere, which alter the concentrations
by using up some of the contaminants and producing others.
On the other hand, the fact, that the required predictions refer
to a hypothetical or indefinite future, ameliorates the necessity for
realistic detail in regard to weather input. What is necessary is that
the types of weather associated with high, moderate, and low pollution
-------
1.5
potential be known, so that hypothetical future weather conditions
which are relevant to the smog problem can be furnished as input to
the model. The most obvious way of accomplishing this, provided ade-
quate weather and air quality records exist for the subject region, is
to use historical weather data for days (or episodes) on which air
quality was severely impaired or which showed other characteristics of
special interest.
•
For example, in a region such as the South Coast Air Basin of
California, where detailed records of both weather variables and con-
taminant concentrations are kept, the model user concerned with a con-
trol strategy would identify the occasion or occasions (within a recent
period of interest) on which the highest oxidant levels were encountered
at various monitored sites. Using the reverse trajectory routine
associated with each of these high oxidant observations, and using
the full model he would obtain model estimates for comparison with the
observed concentrations. Assuming reasonable agreement existed, he
would next rerun the model with the same trajectories and weather con-
ditions, but with revised source emission input, corresponding to the
strategies to be tested.
\
Estimates of the effects of each proposed control strategy could
then be given for each of the selected high-oxidant observations. These
estimates could be expressed either in relative terms, as percent
reduction from the original model estimate, or in absolute terms, by
comparing the predictions with .air quality standards.* The net effect
of this system is to yield, for any number of selected historical obser-
vations for hypothetical reoccurrences of the same weather patterns
but with different emission patterns. Obviously, the same method of
exploration can be extended to determine effects on lower oxidant
levels, and on other contaminants as well.
*In this case, allowance may have to be made for statistically unavoid-
able uncertainty in the model estimates.
-------
1.6
It is important to note that, when this system is used for the
evaluation of proposed control strategies, elaborate simulation of the
dynamics of air flow is not required, since any small errors introduced
in a given model run by a simplified representation of the flow field
are largely balanced by compensating errors in the comparison runs. On
the other hand, the effects of chemical reactions in generating secondary
contaminants are first-order, non-compensating effects which require
simulation on an adequate theoretical basis. Fortunately, the principles
necessary for developing such a simulation are known and, as explained
below, they have been largely incorporated into the REM chemical dynamics
module, which forms the core of REM.
5. Application for Short-Term Forecasting
The short-term forecast process consists of the prediction of pollu-
tion conditions up to perhaps 48 hours in advance in order to notify
the public of expected air quality levels and to issue appropriate
warnings and take appropriate temporary measures in the event than an
air pollution emergency condition is predicted. The problem here is
to predict air quality concentrations from forecasts of large-scale
weather features within rather severe time constraints. Thus, both
accuracy and speed are of crucial importance. In the future, photo-
chemical simulation models may be linked to objective forecasting sys-
tems to assist in making such forecasts and to test the effectiveness
of temporary emergency prevention measures.
The problem is complicated by model input requirements. To the
extent that any air quality model depends for its operation on input
of recorded weather data, that model will be inapplicable as an aid in
forecasting air quality on the necessary day-to-day basis. If, on the
other hand, a model can readily accept and utilize provisional approxi-
mate values of a small set of weather parameters, then prospects of
incorporating the model in a short-term forecasting system would be
considerably brighter.
-------
1.7
Forecasting considers variables on the macroscale and mesoscale
related to the general characteristics and flow of large air masses;
while air quality models often require input conditions at specific
measuring stations and receptor points on the microscale. Models which
require more meteorological detail will have more difficulty in this
regard than models which retain only the type of detail that is most
compatible with current forecasting procedures.
A brief analysis of the forecast process will suggest the compati-
bility requirements. This essentially involves analysis of scale
interrelationships, i.e., summarization of predictive information from
the macroscale to the microscale.
a. Macroscale to Mesoscale
Large scale weather patterns are predicted from current weather
information either manually or by means of National Weather Service
Prognostic Models (barotropic or baroclinic) , e.g., through, the
depiction of changes in height of a 500 millibar surface. From
these data, predictions are. made regarding general air flow charac-
teristics, potential mixing heights and solar radiation (cloud
cover) for the region of interest.
b. Mesoscale to Microscale
From this information, predictions must be made of air flow
and mixing depth characteristics for specific areas within the
region in order that the relationship to air quality can be pro-
perly made (and for a photochemical model to function). This
involves the construction of appropriate trajectories by taking
into account the relationship of the general flow of the air mass
to the surface flow, and mapping of the predicted variations in
mixing height for the region. This is usually accomplished by
trained and experienced meteorologists who associate large and
-------
1.8
small scale conditions by use of long term historical, climato-
logical and statistical records of weather and air quality condi-
tions available to them in the region.
The question of the use of photochemical models in the forecast
process thus will depend either on the establishment of an objective
forecast system which incorporates the model (if the model is indeed
needed), or incorporating the model with the manual procedures routinely
employed by the forecast meteorologist.
In either case, the linkages between the model and the forecast
procedure are the mixing depth variables, trajectory construction
(which embodies both wind speed and direction) and cloud cover. The
Lagrangian approach therefore seems especially adaptable and compatible
with forecast procedures. REM, in particular, is founded on the trajec-
tory approach and can accept exactly the kind of input that a forecast
meteorologist would wish to supply to it. The trajectory model possesses
flexibility and non-redundancy in application; only one or any number
of trajectories can be tested. REM also will readily permit input of
derived and hypothetical trajectories for rapid model computations.
6. Research, Training and Other Applications
Other uses of a model include interpolation of contaminant concen-
trations at locations not covered by an air monitoring network; calcu-
lation of receptor dosages; identification of source areas that contri-
bute to the air quality of a receptor point; testing of postulated func-
tions; and training of personnel. The trajectory model is especially
and efficiently adapted to these uses.
-------
1.9
C. RATIONALE OF THE REACTIVE ENVIRONMENTAL SIMULATION MODEL (REM)
1. General Approaches to Model Building
Two general approaches to model development may be characterized:
Derivative and Conservative.
The derivative approach usually proceeds from the complex to the
simple. It attempts to systematically model all functions—both first
and second order—which the modeler believes must be accounted for in
the total environmental system. The modeler may then modify, delete
or simplify functions as necessary in the course of his effort to
improve the workability of his model. The derivative approach in
photochemical air quality modeling is currently characterized by the
use of the "conservation-of-mass" equation, with elaborate procedures
to take into account fine details of air flow in a three-dimensional
field.
The conservative approach proceeds from the simple to the complex,
as necessary, and emphasizes those functions that correspond most closely
to observed atmospheric behavior and which utilize commonly available
types of information. This approach focuses on major first order
influences. Once the effects of these influences are established,
second order effects are added if needed to improve the accuracy of
the model. In photochemical modeling, this approach is exemplified by
the Lagrangian (moving coordinate) system in which attention is focused
on events which take place in any individual parcel of air as it follows
a natural trajectory.
2. Reasons for Following the Conservative Approach
PES, since the beginning of its model development effort, has adop-
ted a conservative, Lagrangian approach for the following reasons:
a. Because of the scarcity of relevant observational data, appli-
cation of the derivative approach requires many empirical assumptions
-------
1.10
regarding the relation of detailed air movement to the observable sur-
face conditions. Such assumptions must be furnished if the model is to
be made operational; and, frequently, assumptions having no theoretcial
basis may be used for reasons of computational convenience. In a
detailed, all-inclusive model it is difficult to evaluate the actual
effects of such speculative assumptions, singly and in combination,
on the final results. Unless such effects can be isolated, speculative
functions possess the danger of compounding errors.
b. The derivative approach tends to yield complex models which
require comprehensive, detailed input information; in effect, a different
model may be required for each day to be simulated. Such extensive
and complex input data may restrict the flexibility of the model in
its ready use for testing many days of the year. Similarly, the degree
of detail needed to verify and apply the model is greater and front-
end labor and costs tend to be significantly greater than for the simp-
ler models.
c. Although the derivative approach initially tends to satisfy
the sense of completeness, it may render the development effort proble-
matical, requiring costly and expensive revisions. Another danger is
that the effort to model one aspect comprehensively may also hinder
the development of another important or needed mechanism. For example,
should greater resolution be required in emission constituents and
should verified chemical mechanisms be required which are larger than
those currently in use, a complex model may be difficult to modify for
these purposes.
d. Severe limitations in the availability of reliable and accurate
emissions inventories and meteorological information limit the accuracy
attainable in any moral. Unless the comprehensive model is drastically
simplified, to be usable with such input data, cost-effectiveness is
likely to suffer.
-------
1.11
3. Structure of the Simulation Process
The PES Reactive Environmental Model (REM) may be considered in two
parts: (1) the "core model", i.e. those routines which express the
fundamental, invariant principles which can be modeled; and (2) the vali-
dation, experimental and application routines which are for use in
specific cases and experimental conditions. The heart of the PES
approach has been to distinguish these two features of the simulation
process. As explained below, consideration of the chemistry of a parcel
of polluted air leads to an invariant paradigm which can be taken to
be central, thus facilitating the isolation of effects due to the
auxiliary routines which can be modified or improved as new information
becomes available. The realization of the invariant paradigm thus
permits the assessment of the importance of the effects of postulated
functions.
The importance of the chemical mechanism in photochemical modeling
in general, in its function in the "core model" and its role in control
strategies is clear from a number of points of view that are taken
up in other sections and chapters of this report. At the same
time, the chemical module of any photochemical air pollution simulation
model possesses a greater potential for achieving an invariant status
and hence an acceptable degree of validity than do the meteorological or
emission functions. As a body of theory, chemistry possesses a greater
consistency and constancy of principles, with regard to predictability,
than do other disciplines presently bearing on atmospheric modeling.
Further, the large amount of unapplied chemical knowledge and increasing
amount of experimental data available makes it possible to achieve
an acceptable chemical module in a much shorter period of time and at
considerably less expense than would be the case with the other func-
tions of the model.
The basic parameters of elementary chemical reactions are invariant
relative to changes in concentration and light intensity. Once they
have been determined for any given reaction, they need not be reassessed
for application in various concentration ranges. The body of data and
-------
1.12
knowledge available thus increases the probability of developing a valid
chemical mechanism useful for purposes of atmospheric simulation.
The importance of the role of the chemical mechanism in models used
for control strategy comparisons, also, is clear from the fact that since
meteorological conditions are typified and then held constant, the reli-
ability of the comparisons becomes a function of the validity of the
chemical mechanism and the quality of the emission information employed.
An adequate representation of the chemical mechanism, therefore, is
particularly essential whenever the effects of new controls, fuels and
propulsion systems which involve detailed information on emission consti-
tuents and reaction rates, are to be subjected to thorough testing.
The characteristics of the REM chemical module are further described
below, and in Chapter II.
D. ROLE OF CHEMICAL REACTIONS IN DETERMINING AIR QUALITY
1. Introduction
In relation to non-reactive contaminants, air quality is determined
by the rates of emission and the properties of air in motion. These
properties have been subject of much study; theoretical formulations
have been developed which account well for behavior of air as observed
in the laboratory (under carefully controlled conditions).
But these formulations are not easy to apply to simulation of atmos-
pheric conditions, because the necessary parameters are usually not
observable. Therefore, in models which are intended to simulate the
urban atmosphere, various simplifications are made. These attempt to
preserve the formalism of the ideal case while recognizing (and avoiding)
some of the complications of the real case.
One class of existing air quality models is based on "plume" formulas,
which apply to steady wind conditions in wind tunnels, for example. Atmos-
pheric and environmental conditions which interfere include fluctuating
-------
1.13
wind (both speed and direction); vertical turbulence; non-uniformity
of wind with height above ground; roughness of terrain and
intrusion of obstacles (such as buildings). Ignoring these factors
usually leads to serious inaccuracies in predicting air quality, so
various empirical adjustments are devised to account for them, or for
some of them. These take the form, for example, of tables and graphs of
vertical and horizontal diffusion parameters recommended for use in
various typical weather conditions; summation over long periods of time
to take advantage of the smoothing effect of averaging; "calibration" of
the formulae by optimizing an arbitrary factor for best representation of
observed data; in certain cases, optimizing an assumed "half-life", in
analogy to radioactive decay processes.
Simulation of the air quality in an urban atmosphere, by this approach,
requires the superposition of contaminant concentration patterns individ-
ually calculated for each of the large number of emission sources in
the typical city. If enough degrees of freedom are provided in the form
of arbitrary factors, and if the degree of detail required in the repre-
sentation of air quality is kept low enough (e.g., restricting predic-
tions to annual average contaminant level for a specified sensor loca-
tion) the semblance of validation for an atmospheric model can be
provided even without the substance. However, the irreducible require-
ment for plausibility with modeling of this sort is that concentrations
of contaminants appear to decrease with distance from known sources.
The behavior of photochemical product contaminants is incompatible with
this requirement and, therefore, cannot be imitated successfully by
diffusion models based on the plume formulation.
Determining factors for the concentrations of photochemical product
contaminants are, at basis, the rates at which chemical reactions occur.
These rates depend strongly on the concentrations of precursor contami-
nants and on the intensity of sunlight. Furthermore, the rates are
joint functions of the concentrations; this implies that the observed
concentrations of the products must depend upon interactions between
-------
1.14
contaminants emitted into the atmosphere at different locations, and
that the effects of these various sources are not simply additive at
any receptor location. This is especially obvious when the reaction
between ozone and nitric oxide is considered: the first effect of
injecting nitric oxide into an air mass laden with ozone is that ozone
is consumed, with the production of nitrogen dioxide; later, if the sun
is shining, ozone will again accumulate (but this will occur at a
location perhaps far removed from the point where the nitric oxide was
added).
Thus, for photochemical product contaminants, predicting the changes
of concentration in an urban atmosphere requires more detailed infor-
mation than for non-reacting contaminants. Further, unlike the non-
reacting contaminants, photochemical products undergo changes of con-
centration which are in no way related to the mixing and transport of
air; these changes can be demonstrated by studying synthetic atmospheres
immobilized in irradiation chambers, under conditions such that non-
reacting contaminant concentrations remain steady or change only by
dilution. From this it follows that successful simulation of concen-
trations of the reactive contaminants and their products requires con-
sideration of more factors than does simulation of non-reacting contami-
nants.
These additional factors basically involve the phenomena of chemistry
and must be considered in that context.
2. Principles of Chemical Kinetics in Relation to Simulation of Air
Quality
In the science of chemistry, kinetics is the branch which treats of
the rates of occurrence of chemical changes. Modem gas phase kinetics
is based on the well-established molecular collision theory of chemical
reactions. According to this theory, most chemical change is the
result of collisions between molecules, which cause rearrangement of
the atoms which compose the molecules, thus giving rise to new molecules
-------
1.15
having different chemical characteristics.
Whether or not a reaction ensues when two chemical species are mixed
depends, in the first instance, on whether an effective rearrangement of
the atoms involved is chemically feasible. If a.particular type of
rearrangement is feasible, it will take place not on every collision
between molecules of the two species, but for some fraction of such
collisions; this fraction is called the "collision yield." It is impor-
tant because it is universally characteristic of the molecular behavior
of these two species when mixed in the gas phase (for example, in the
atmosphere*). The collision yield is commonly a function of the tempera-
ture of the gas which contains the reacting molecules, but it is
independent of the concentrations of the reacting species and of any
other species present in the mixture, whether these extraneous substan-
ces are reactive or not. It1 is also independent of such atmospheric
variables as wind, humidity, mixing height, barometric pressure and
cloud cover.
The chemical change which takes place in this way, as a direct
result of molecular collision, is known as an "elementary reaction."
When two substances which react in this way are mixed, the rate at which
they react is equivalent to the rate at which suitable molecular col-
lisions occur, multiplied by the collision yield. The collision rate
is proportional to the product of the concentrations of the reactants; it
also depends (not strongly) on the temperature of the gas mixture, but it is
essentially independent of the concentrations of other substances and
of atmospheric variables such as wind, humidity, etc. The rate of
reaction, since it is proportional to the collision rate, is therefore
also independent of the concentrations of extraneous substances and of
the atmospheric variables mentioned.
In a multicomponent system (such as an urban atmosphere), many
different elementary reactions take place simultaneously, but each con-
forms to the kinetic principles just outlined, except for a very few
rather minor complications. The resultant behavior of the system may
*
It should be noted that real atmospheres always contain particles of liquid
and solid, which present surfaces where heterogeneous reactions can take
place. Calculation of rates for such reactions is subject to similar, but
slightly more complicated, rules.
-------
1.16
be mathematically complicated because each of the rates of the elemen-
tary reactions, which cause the concentrations of the reactants to
change, itself depends upon the concentrations of one or more of the
reactants. This resultant behavior can, however, be calculated if all
the elementary reactions and their collision yields are known. The
modern high speed computer enables us to perform such calculations for
some rather complicated systems.
The main difficulty in applying these principles to simulation of
atmospheric photochemistry lies in the fact that our knowledge of the
elementary reactions involved is incomplete, and our estimates of the
collision yields are not very precise, even for those elementary reac-
tions which are well recognized. The advantage of pursuing this approach,
which is embodied in REM, is that the parameters involved are fundamental
invariants of nature; they are required to be consistent with the entire
body of chemical knowledge; they can be tested by experiment, not only
in simulated smog irradiation chambers, but in many other ways; and
since they are independent of many other atmospheric variables including
concentrations, they may be expected to be useful in extending simulation
to lower ranges of pollutant concentrations.
3. Distribution of Contaminants in the Atmosphere, Resulting from Chemical
Reactions and Transport
To understand the behavior of contaminants in the atmosphere in
terms of the changes of concentration with time, it is necessary to know
not only the chemistry of the contaminants as it might be seen in an
irradiation chamber, for example, but also the relevant facts concerning
input of pollutants into the atmosphere and their dispersal and mixing
by physical forces in the moving air. The principles involved in describing
and simulating this behavior are the subject of this section.
A very general approach to the description of this behavior is the
use of the "conservation of mass" equation . This equation represents
Darling, E.M. "Computer Modeling of Transportation-generated Air
Pollution." Report No. DOT-TSC-OST-72-20, June 1972.
-------
1.17
the changes in concentration of any contaminant at any point in the at-
mosphere. It states that the rate of change of concentration at that
point is equal to the net effect of four processes: (a) the change due
to chemical reactions; (b) the change due to emissions from pollution
sources; (c) the change due to movement of the air; and (d) the change
due to molecular diffusion.
In the present state of the art, a general solution of the conser-
vation-of-mass equation is not feasible, partly because of the complexity
of the computations which would be involved but, even more importantly,
because of the lack of some of.the information required to compute the
individual effects of the four processes. Deficiencies in present
knowledge of the effects of chemical reactions are discussed in the
preceding section of this chapter; in the PES model, these deficiencies
are minimized by structuring the calculations in terms of elementary
reactions, so that improvements in the understanding of the chemistry
and in the precision of the parameters leads easily to an improvement
of the representation of the chemical behavior.
The effects of changes due to emissions from pollutant sources are
not difficult to compute, if rates of emission are known. However,
the number of individual sources in an urban area is so large that a
complete accounting is completely unfeasible, even if the necessary
details of the behavior of each source were available. For practical
purposes, therefore, these effects must be estimated in terms of average
behavior for multitudes of small sources, with substantial detail only
on a small number of very major sources. Thus, atmospheric modeling
for air quality requires an emissions model for each region to which
the model is to be applied. The PES atmospheric simulation model pro-
vides for separate emissions modules for several types of sources,
namely, freeway traffic, non-freeway vehicular traffic, major industrial
sources and aggregated other sources. Emission rates are estimated as
a function of location and time. Obviously, the simulated detail
-------
1.18
resulting from the application of such an emissions model must be very
different in many particulars from the real status of emissions at
any moment chosen for simulation. This discrepancy, then, constitutes
an inherent, serious limitation on the achievable accuracy of air quality
modeling.
The effects of changes due to motion of the atmosphere have received
much attention in some recent approaches to air quality modeling. It
*
is common to divide the motions of air into two types: a "transport"
or advective component, which causes net geographical displacement of
air (including both horizontal and vertical displacements beyond some
minimum magnitude); and a "turbulent" component, which causes mixing of
air from parcels in neighboring locations. Even when thus partitioned,
however, the equations of motion remain intractable unless further sim-
plifying assumptions are made.
For solving the equations of motion after the simplifying assump-
tions are introduced, there are two principal approaches, which differ
in the choice of a calculational frame of reference. In one, called
the "Eulerian" method, effects of all motions are computed with reference
to a geographically fixed co-ordinate system. In the other method,
»
called "Lagrangian", calculations are further simplified by computing
only the advective effects relative .to geographic co-ordinates, while
the effects of turbulent motions are referred to a moving co-ordinate
system (moving, specifically, with the velocity of the wind). The PES
model utilizes the Lagrangian approach and focuses attention on pro-
cesses occurring in a moving column of air.
The effects of molecular diffusion are quite generally assumed to
be negligible, because the linear range of-such effects has been shown
by experiment to be small relative t.o both advective and turbulent
atmospheric motions. This assumption is also implicit in REM.
To summarize and symbolize these concepts, we may write the con-
servation-of-mass equation in the form
*
cf. Darling, loc. cit.
-------
1.19
c' = R, + S4 + A, + T. + M, (1)
where c. is the rate of change of concentration of species i_;
R is the rate of change due to chemical reactions;
S is the rate of change due to pollutant emissions;
A. is the rate of change due to advection;
T. is the rate of change due to turbulence;
M. is the rate of change due to molecular diffusion.
In large part, the problem in constructing a useful simulation
model for air quality lies in identifying appropriate assumptions
to simplify the computation of the various terms of Equation 1. As
explained above, M can be taken as zero by assumption, leaving only
four terms to be dealt with. By adopting the Lagrangian approach, A
can be rendered tractable and effectively disengaged from the central
problem of calculating concentrations, leaving for further consideration
the terms R., S , and T.. These terms must be computed, explicitly or
implicitly, for each parcel (i.e., each hypothetical volume element)
of air within the compass of the model.
To develop a cost-effective model, it is desirable to minimize the
computational effort involved in simulating those details of reality
which are of little importance or relevance to the objectives of model-
ing. For example, any chemical reaction which could be shown to proceed
to completion within a few seconds after the primary contaminant
,responsible is emitted into the air could be subsumed in the effects of
sources and withdrawn from consideration in the more complicated chemis-
try represented by the term R . Similarly, if any of the three terms
R , S and T. could be shown to be negligible relative to the total,
then computation for that term could be curtailed or eliminated. For
purposes of the PES model, it is assumed that the effects of turbulence
on rate of change of concentration are negligible, except in the vertical
-------
1.20
dimension. With respect to vertical turbulence, it is assumed that its
effect is to produce homogeneity of contaminant concentration in the
vertical dimension, from the ground to the mixing height. Combined,
these assumptions obviate the necessity for performing calculations to
determine T. as a separate quantity, by defining the appropriate hypo-
thetical volume element as an air parcel comprising a uniform column
having height equal to the mixing depth of the urban atmosphere at the
given time and location.
The effect of eliminating T. as an essential term.of the conser-
vation- of -mass model is to make the calculational scheme for the model
formally equivalent to that of an irradiation chamber of variable
volume in which chemical reactions take place, and into which fresh
quantities of reactants are continuously introduced. The simplified
version of Equation 1, appropriate to the PES model, may therefore
be formulated as
Ci = Ri + Si + Hi (2)'
where H. represents the rate of change of concentration of species i_
due to the variation of mixing height.
This formulation is appropriate, of course, only to the extent that
the effects which have been ignored or subsumed are really small rela-
tive to those which have been retained. This is likely to be more
nearly true in some circumstances than in others. We must, therefore,
examine the typical state of an atmosphere in which photochemical smog
is likely to develop, in order to justify the assumptions on which
REM is based.
1. The assumption that the effect of horizontal components of
turbulent motion is negligible. This assumption would be absolutely
correct if there were no variation in concentration with geographical
location, and it will be nearly correct whenever horizontal gradients
of concentration are sufficiently small, so that the replacement of a
small parcel of air within the postulated column by an equal parcel
-------
1.21
from a short distance away would cause only a negligible change in
the concentrations being calculated.
With regard to large scale inhomogeneities in the concentration of
contaminants in the Los Angeles Basin, two comments are germane. First,
the relatively slow progress of photochemical reactions, when combined
with the rapid rate of small scale mixing such as caused by daytime
breezes and vehicular traffic, tends to smooth out the differences in
concentration of oxidant, ozone and other photochemical products, pro-
ducing rather small concentration gradients over substantial areas
within the urban region. A study of this situation, reported in an
2
authoritative review dealing with photochemical smog in the Los
Angeles Basin, led the authors to conclude: "In spite of...lack of
complete mixing, the maximum daily oxidant at stations within a 600 to
800 square mile area are directly proportional to each other in a
predictable fashion."
Second, the scale of representation of inhomogeneities in the hori-
zontal dimension, in the atmospheric simulation, cannot usefully be
made any finer than the representation of inhomogeneities in the geo-
graphic variation of contaminant emissions. In the present state of the
art, this effectively restricts the resolution to that available on
emission inventory grids, on which the smallest differentiated areas
are of the order of 4 square miles. Gradients permitted by such reso-
lution are, beyond question, negligible in their effects on small air
parcels subject to vertical mixing and chemical reaction.
2. The assumption that the effect of the vertical component of
turbulent mixing is to establish uniform contaminant concentrations in
a column of air between the ground and the mixing height. This assump-
tion is in accord with the conventional meteorological explanation of
2
Schuck, E.A., J.N. Pitts, Jr., and J.K.S. Wan. "Relationships
Between Certain Meteorological Factors and Photochemical Smog," Air
and Water Pollut. Int. J. 1966 Vol. 10, pp 689-711.
-------
1.22
the establishment of normal lapse-rate conditions of vertical tempera-
ture structure below the subsidence inversion in the Los Angeles Basin,
during sunlight hours. This follows from the vertical motion, due to
buoyancy of air parcels warmed by contact with the ground and other
surfaces heated by the sun; cooler air from above sinks to replace the
rising air, and is in turn warmed. This convective effect soon results
in a near uniform temperature gradient below the mixing height.
It is to be expected that this mixing mechanism should tend to
distribute pollutants from sources near the ground (such as motor
vehicles) in a similar manner, as any relatively cool, unpolluted air
above the ground layer soon reaches the ground to receive heat from
the ground and contaminants from the ground-based urban activities.
Little is known of the actual distribution of contaminants under such
circumstances, as few studies have been made.
However, one instance investigated by a research team in the Los
Angeles Basin substantially supports the assumption in question.
Temperature, carbon monoxide and oxidant concentrations were measured
at various altitudes above three air monitoring facilities during
several ascents and descents, on a day (September 29, 1969) when photo-
chemical smog prevailed in the Basin. The vertical distribution of
contaminants showed a marked tendency toward uniformity, increasing as
the day progressed and ground temperatures increased. For example,
the report cites the following observations: at Commerce at 0810, a
strong inversion base near 700 feet; at 700 feet, the CO is half the
ground level; at 1000 feet, the CO is one tenth of the ground level.
At 0900, the inversion had lifted slightly, and the CO at 700 feet had
increased to 80 percent of the ground level. Near noon, the oxidant
exceeded 20 pphm from ground level to 1700 feet at El Monte, but at
Commerce it was 4 pphm or less at all points above 800 feet with con-
centrations of 8 pphm from the ground to 750 feet. At 1300, the inversion
"1969 Atmospheric Reaction Studies in the Los Angeles Basin, Volume I"
Final Report, Contract No. CPA 70-6 (Scott Research Laboratories)
-------
1.23
was gone at El Monte; the temperature profile was close to the adiabatic
lapse rate; the oxidant was nearly uniform from the ground to 2200 feet;
CO was approximately the same at ground level, 700 feet, 1100 feet and
1500 feet.
Although these and other observations support the assumption that
vertical turbulence tends to produce uniform concentrations throughout
a vertical column of air, it is nevertheless clear that a finite time
must be required for the operation of this mechanism. If the time
required is long enough to be comparable with the characteristic time
for development of photochemical products, then the concentrations of
these contaminants should be expected to be substantially different
near the top of the column than near the ground. No doubt the rate of
mixing varies considerably from one day to another, and from one location
to another. However, no method of estimating the effect of such varia-
tion has yet been offered. In the present state of the art, it appears
likely that the assumption of uniform concentrations within the column
is adequate for modeling purposes, under most of the conditions that
give rise to photochemical smog.
As observed above, these assumptions regarding the effects of tur-
bulent air motions render the calculational scheme for the PES model
formally equivalent to that of an irradiation chamber, conceived as a
chamber having variable volume and capable of receiving fresh contami-
nant input during the course of irradiation. Use of this structure for
t .
the model has the incidental advantage that it is readily adaptable for
simulation of actual irradiation chamber experiments. Thus, testing of
any proposed chemical reaction mechanism, with the aid of appropriately
designed experiments, is facilitated. Chapter 3 of this report deals
with the results of an extended series of irradiation chamber simulations,
using data supplied by the contracting agency.
Another implication of the procedure of subsuming effects of turbu-
lence under those of volume variation (cf. Equation 2) is that the effects
-------
1.24
of chemical reactions should be, in reality, more important than those
of atmospheric turbulence. If this is true, then the changes in air
quality as observed in the atmosphere should resemble, to an appreciable
extent, the changes seen in chamber irradiations. The fact that they
do so is well known and attested in the literature relating to photo-
chemical smog. It has recently been reconfirmed in a statistical study
by PES of Los Angeles air quality data for the years 1968 through 1970,
dealing with the average diurnal variation of the photochemical contami-
nants. It is found that, on the average, nitric oxide concentration
decreases continuously from the time of maximum morning vehicular traf-
fic, while nitrogen dioxide increases to a maximum after about two
hours and oxidant reaches a maximum after four or five hours (all at a
particular receptor location in Los Angeles). Such observations tend
further to confirm our contention that the PES model suffers no serious
loss in plausibility from the assumptions used to obviate the calcula-
tion of T. (Equation 1)—a calculation for which, in any event, only
empirical approaches appear to be available.
4. Chemical Reactions in the Atmosphere in Relation to Control Strategy
Planning
One of the principal uses of air quality models is the estimation of
probable effects on urban air quality which would arise from various
possible pollution abatement strategies. REM is designed to emphasize
the importance of those aspects of the model which are most sensitive
to changes in abatement strategy, namely, the effects of contaminant
emissions and the effects of chemical reaction—the terms S and R ,
respectively, of Equation 1.
The relevance of a cogent model of contaminant emissions is, of
course, obvious; no model can produce estimates of air quality effects
which are any more accurate than the supplied estimates of emissions.
However, in some of the extant approaches to air quality modeling,
emphasis is placed on detailed calculation of the effects of advection
and turbulence—the terms A and T—at the expense of a cogent model
-------
1.25
of the chemical reactions.
The importance of an adequate simulation of the chemical aspects of
contaminant development should be particularly clear in the considera-
tion of photochemical product contaminants like ozone, one of the most
toxic constituents of photochemical smog. Referring again to the terms
of Equation 1, it can be seen that, in this case, the term S. is iden-
tically zero, since there are no ozone emissions. It is only through
the term R that the effect of emissions of contaminants other than
ozone can be taken into account in estimating air quality relative to
ozone, and no amount of advection and turbulence can tend in the
slightest degree of account for the production of ozone if the chemical
simulation does not adequately do so.
It must be further observed, that the mode of use of an air quality
model for control strategy planning is basically a comparative mode
(in mathematical terms, a difference or ratio mode). That is, the question
to be answered in applying the model for this purpose is, given such and
such changes in the existing emission inventory for a given region,
what changes in air quality can be expected to ensue? Viewed in this
light, it can be seen that elaborate computational procedures for care-
ful simulation of the atmospheric motions (terms A. and T ) are unlikely
to be cost-effective, since (a) there is no expectation that the pro-
cesses of advection and turbulence will be in any way affected by
changes in the existing emission inventory and (b) the accuracy of
individual simulations is less critical than the relative accuracy of
the comparison.
The relative importance of these aspects is reflected in the detailed
structure of the terms of Equation 1, on closer examination. For
example, the term T in the conservation-of-mass formulation, by either
_I 1
the Eulerian or the Lagrangian method, becomes
-------
1.26
in which there is no requirement or provision for any information regar-
ding contaminants other than species 1^. Thus, in particular, the effect
on ozone concentrations of a change in the emission inventory of nitric
oxide would not be evidenced in any direct way by changes in the calcu-
lated value of T .
With regard to the term R , representing the rate of change of concen-
tration of species jL (ozone, say) as a result of chemical reaction, the
situation is far different. The detailed form of R. is expressed as a
sum of terms for individual elementary reactions*; some of these terms
are negative in magnitude and contain as a factor the concentration c ,
but almost all contain, as additional factors, the concentrations of
other constituents. (In the case of ozone, one of the negative terms
contains the concentration of nitric oxide as a factor. Thus the behavior
of the term R. does reflect in a direct manner the influence of changes
of nitric oxide emissions on ozone accumulation.)
For these reasons, when control strategy planning applications are
envisaged, it is important to select an air quality model which (like
REM) incorporates an adequately detailed chemical mechanism.
5. Formulation of a Chemical Reaction Mechanism in Terms of Elementary
Reactions
Development of an adequate chemical reaction mechanism for simulating
the time-profiles of concentrations of reactive contaminants and reaction
products might be expected to be a relatively simple and straight forward
process, in view of the basic simplicity of the kinetic principles
discussed above (section 2 of this Chapter). The procedure does, how-
ever, contain pitfalls for the unwary, some of which will be discussed
in this section.
One of the most common fallacies committed by would-be modelers who•
do not understand chemical kinetics is to confuse complex reactions with
For further detail, see Volume II of this report
-------
1.27
elementary reactions. For example, it is not uncommon, in discussing
control strategy for photochemical contaminants, to encounter the argu-
ment that maximum ozone concentrations should be proportional to the
product of maximum concentrations of hydrocarbons and oxides of nitrogen,
since these are known to be precursor contaminants for ozone.
That this relation does hot hold true, in general, has been often
demonstrated in irradiation chamber experiments with simulated smog.
Nevertheless, it is4sometimes felt that the assumption, that it is true
in the urban atmosphere, will yield an adequate simulation for air quality
with respect to ozone. This is thought to be plausible in view of (a)
the superficial similarity of this relation to the corresponding expres-
sion of the rate law for elementary reactions and (b) the scarcity of
relevant data regarding hydrocarbon concentrations in urban atmospheres.
The basic fallacy, here, is in contending that the rules for ele-
mentary reaction should hold when, in fact, there is no elementary
reaction, between hydrocarbons and oxides of nitrogen, that has any
appreciable collision yield. This should be obvious, incidentally,
from the fact that these components do not react with each other in
the absence of sunlight. As an indication of how unreliable such an
assumption may be as a quantitative guide, we may cite the classical
investigations of the kinetics of photochlorination of olefinic hydro-
carbons (a somewhat parallel case). Here the rate of change of hydro-
carbon concentration was found to be proportional to the concentration
of chlorine gas, but almost completely independent of the hydrocarbon
concentration—a clear-cut contradiction to the type of simple assump-
tion advocated by these oversimplifiers.
Although elementary reactions are, in fact, responsible for the
chemical changes which occur in both the photochlorination system and
the photochemical smog system, most of these elementary reactions involve
so-called "intermediate" species: very reactive molecules having a high
content of chemical energy, such as free radicals and atoms—basically,
molecular fragments produced in the system by the absorption of energy
-------
1.28
from sunlight. (In the photochlorination system, one such species is
free atomic chlorine, produced by photolysis of ordinary diatomic
molecules of chlorine gas; in the photochemical smog, or "photo-oxida-
tion", system, such species include free atomic oxygen, from photolysis
of nitrogen dioxide, and free hydroxyl radicals, as well as a wide variety
of other free radicals). Because of their high reactivity, their con-
centrations always remain exceedingly small, and cannot be measured by
methods currently available.
Thus, the adequate simulation of the photochemical smog system,
whether in chamber experiments or in the atmosphere, depends on a
reasonable understanding of the nature and properties of these interme-
diate species, and some reasonable estimates of the collision yields of
the elementary reactions which involve them.
As previously mentioned, the state of the art of computer simulation
of chemical kinetics is now such that, using available routines for
numerical integration, concentration-time profiles of many components of
a reaction system can be readily generated, starting with postulated
initial concentrations and a mechanism of as many as a hundred individual
reactions, perhaps more. Thus, if a complete mechanism is postulated,
it may be readily tested for agreement with relevant data from chamber
experiments. If, under a reasonably wide range of initial conditions,
the results of computer simulation are in reasonable agreement with the
results of experiment, then the postulated mechanism may be regarded as
a fairly accurate description of the true chemistry of that reaction-
system, provided that none of the postulated collision yields of the
elementary reactions are known to be in serious conflict with other
chemical information.
However, it is important to note that the postulated mechanism
cannot be considered chemically adequate, unless all of the conditions
listed above are met, using values for the collision yields which are
invariant except for a temperature-dependence of a theoretically-per-
mitted form. A mechanism is not satisfactory in a theoretical sense,
-------
1.29
if (a) different values of the collision yields (or related rate factors)
are required for simulation of experiments with differing initial
concentrations of various components; or (b) individual reactions, as
written, are not compatible with established chemical principles,
including stoichiometry; or (c) molecular formulas are written in
generalized terms, corresponding to more than one chemical species.
Each of these fairly common types of discrepancies invalidates a mechanism,
at least in a theoretical sense, although in practice, some instances
may be less damaging than others.
For a number of relatively simple systems, including the olefin-
photochlorination systems cited above, theoretically adequate reaction
mechanisms have been worked out. In most such cases a relatively small
number of elementary reactions have been involved, usuually less' than
twenty.
The evident complexity of the photochemical smog system, however,
ensures that a theoretically adequate, or nearly adequate, mechanism
will involve a much longer list of elementary reactions. Complete
determination of that list is not presently in sight. Therefore, to
develop a satisfactory mechanism soon for practical purposes, some com-
promises with the purity of a theoretically satisfactory mechanism must
be countenanced.
In developing the mechanism currently employed in REM, PES has
utilized three types of compromises:
(a) Chain-terminating reactions., i.e., reactions between free
radicals and between nitric oxide and free radicals, have been assigned
common values of their rate factors.,, and; the products have been assumed
quantitatively negligible.
(b) Hydrocarbons other than propylene have been lumped into a single
category, assumed to react with oxygen atoms with a single rate factor
and with identical products.
(c) Rate factors for the chain-terminating reactions, and for the
-------
1.30
reaction of generalized hydrocarbons with- oxygen atoms, have been oc-
casionally allowed to vary, in the interest of improving simulation for
certain irradiation chamber experiments.
These assumptions and the manner of their implementation are discussed
4
below (Chapter 3) and in the final report on an earlier project in which
the main features of REM were developed.
At the same time, in developing REM, PES has avoided utilizing
certain types of compromises, regarding chemical validity, which have
been advocated by other modelers . These include:
(1) the use of chain-branching factors as invariants of the
mechanism, in conflict with the fundamental nature of such factors as
derived quantities inherently sensitive to concentration changes. E.g.,
"OH + HC^(b2)R02 where b2=4 for propylene."
(2) the use of variable stoichiometric coefficients, in conflict
with the principle of invariance of collision yields for elementary
reactions in chemical kinetics. E.g., "NO + RO — NO + dOH, where d=l/2
to 1/4 for various concentrations."
Such parameters may in some cases facilitate the adjustment of
simulation results to agree with reference data but, to the extent that
they contravene the established principles of kinetics, they sacrifice
generality and impair the credibility of the model for application to
conditions other than those of the example simulated.
Thus, at the core of REM is a paradigm for the simulation of chemi-
f
cal kinetics which makes it rather easy to test proposed reaction mecha-
j
nisms to determine whether they are compatible with results of chamber
experiments and to utilize the best validated of such mechanisms in
y '
atmospheric simulation. At the same time, the PES approach minimizes
the introduction of discrepancies which can arise from improper postu-
lates as to the relation of the kinetic parameters .to contaminant con-
4
Weisburd, M.I., L.G. Wayne, R. Danchick and A. Kokin, "Development of
a Simulation Model for Estimating Ground Level Concentrations of Photo-
chemical Pollutants." Final Report, Contract No. CPA 70-151 (January 1971)
-------
1.31
centrations.
E. ROLE OF ATMOSPHERIC TURBULENCE IN SIMULATION OF AIR QUALITY
REM is based on a relatively simple view of the motion of air over
an urban area; namely, that air goes where the wind takes it, that in
the process it picks up contaminants from the sources it moves over, and
mixes them thoroughly in the vertical dimension, up.to the mixing depth.
Implementing this approach, REM focuses on hypothetical moving columns
of air and integrates the effects of the emissions, the chemistry, and
the changes in mixing depth as each such air parcel traverses its
appropriate trajectory.
(Practically the same model for air movement was arrived at indepen-
dently by Leahey , and validated for sulfur dioxide in New York City.
Correlation between measurements, using an instrumented helicopter, and
predictions, using the model, was 0.83 for 400 data pairs.)
The PES model calculates mixing depth for any given location and
time as, to a good first approximation, a function.of the temperature at
ground level and of the vertical temperature structure of the atmosphere
over cooler areas of the region. The temperature at ground level is
approximated by interpolation from values observed at nearby measuring
stations. As a result of motion along the trajectory, the elevation of
the base of the air parcel changes because of topography, and the .le-
vation of the top changes because of the changes in temperature at the
ground. The net effect is of a change in volume of the air parcel; this
may be a net increase or decrease, depending on circumstances. Whenever
the volume is increasing, it is assumed that concentrations within the
column decrease to maintain constant total mass; but whenever the volume
decreases, it is assumed that concentration remains steady while total
mass decreases (as if contaminants were being forced out through the top
of the column).
Leahey, D.M., "An Advective Model for Predicting Air Pollution within
an Urban Heat Island," etc., J. Air Pollution Control Assoc. 22, 548-
550 (1972)
-------
1.32
The assumption of homogeneous mixing throughout the column is use-
ful, in that it eliminates the necessity for repetitive computation of
chemical reaction rates at various altitudes, to account for differences
due to inhomogeneity. Another approach which has been advocated is to
assume that the column comprises a number of vertically stacked cells,
within each of which, mixing is instantaneous; then to provide a para-
digm for transfer of material from each cell to the one above it by
virtue of turbulent motions within the column.
This additional complication is likely to be useful only if (a)
the differences in concentration between different levels of an actual
sub-inversion air mass are large enough to cause very substantial
deviations from the rates based on column-average concentrations, and
(b) some plausible algorithm is known, by which the rate of upward .
diffusion due to turbulence can be estimated.
In regard to the first of these conditions, it is clear that real
atmospheres are not really homogeneous in the vertical dimension. How-
ever, we have already adduced evidence (see section D.3 of this chapter)
that, under conditions conducive to photochemical smog, the deviations
from homogeneity are not inconsistent with an impressive degree of
success in validating a model for sulfur dioxide in New York City (v.s.).
It is of special interest, therefore, that in two photochemical air
quality models currently under development, estimates of ground level
concentrations based on the assumption of vertical homogeneity have been
compared with other estimates based on the multilevel approach. Appar-
ently, the differences found have not been very large.
As to the availability of a suitable paradigm for application to the
problem of calculating rates of vertical diffusion, this has been ad-
dressed in the studies of another Environmental Protection Agency con-
tractor , from which it appears that estimates of vertical diffusivity,
in the present state of the art, rest on completely empirical formu-
lations with an extremely sketchy experimental basis. When related to
Contract No. 68-02-0336, Monthly Technical Progress Narrative, Sep-
tember 15, 1972
-------
1.33
wind speed, estimates typically vary from observed values by a factor
of three to ten, and sometimes more. When related to vertical tempera-
ture gradient, the spread of observed values is nearer to a factor of
three, in the single literature source cited for this purpose.
In view of these difficulties, it is pertinent to consider whether
available data indicates that sub-inversion mixing is slow enough to
require separate treatment in the model., Since the typical period
required for development of photochemical smog from one phase to the
next (e.g., from maximum NO to maximum NO-) is about two hours, sub-
stantial mixing within a period of thirty minutes might be considered
likely to produce a reasonable approach to vertical uniformity of
atmospheric composition in the mixing layer. Studies recently publish-
ed showed that vertical velocities as high as 100 feet per minute are
common in the subinversion layer in Los Angeles during the morning,
and that constant-volume balloons were often displaced vertically as
far as 1000 feet in a period of 15 minutes or less. Such vigorous
motions within a mixing layer of, typically, 2000 feet or less, should
provide good mixing.
We conclude that the assumption of ;ertical homogeneity, utilized
in REM, is sufficiently plausible to be accepted for modeling purposes,
at least until alternative formulations are based on more substantial
theory and documented with better field information than is now possible.
F. STATUS AND FUTURE DEVELOPMENT OF REM
1. History of REM
The development of REM first began as a task under contract CPA
22-69-108 to the National Air Pollution Control Administration, '
Angell, J.K., D.H. Pack, L. Machta, C.R. Dickson and W.H. Hoecker,
"Three-Dimensional Air Trajectories Determined from Tetroon Flights
in the Planetary Boundary Layer of the Los Angeles Basin." J.
Applied Meteorology, Vol. 11, p. 451-471 (April, 1972)
-------
1.34
"Comprehensive Technical Report on All Atmospheric Contaminants Asso-
ciated with Photochemical Air Pollution". This project concentrated
on the preparation of the above report, preparation of the "Air Quality
Criteria for Hydrocarbons" document (AP-49), statistical analysis of
pnotochemical data for 6 CAMP cities and 11 SCAN stations, and develop-
ment of an atmospheric pollution simulation model. This work was per-
formed for the Bureau of Criteria and Standards. A completely opera-
tional version was validated and the program of the model, together
with documentation and results of control strategy exercises (the first
ever to be conducted with a photochemical simulation model), were
delivered to NAPCA, June 1970. The results of this effort were repor-
ted in the Journal of the Air Pollution Control Association, Vol. 21,
No. 6, June 1971.
The first version of the model contained a 31-step reaction mechan-
ism (substantially different from the current version) and a stationary
point source diffusion module based on standard diffusion principles.
It utilized Hamming's linear, multistep predictor-corrector method
for numerical integration purposes. With these exceptions, and sub-
sequent, optimizations and refinements, the design of this model has
carried forward through the current project.
The next project was performed under Contract CPA 70-151, "Develop-
ment of a Simulation Model for Estimating Ground Level Concentrations
of Photochemical Pollutants," for the Bureau of Abatement * National
Air Pollution Control Administration, July 1970 through January 1971.
The work performed under this contract included a literature review of
chemical mechanisms, chemical chamber simulations, program modifications
to handle mixing depth information aquired from Scott Research Labora-
tories for September 29, 1969 at El Monte, Commerce and Hawthorne,
revision of the source emission inputs, and conduct of 15 validation
runs for September 29 and 30, 1969. Some 21 sensitivity tests and
-------
1.35
11 control strategy exercises were also conducted in this project. The
results of this effort were extensively reported in "Final Report:
Development of a Simulation Model for Estimating Ground Level Concentra-
tions of Photochemical Pollutants", January, 1971.
For 15 validation runs the percent deviations* by contaminant were
as follows:
% Less than
Contaminant Ave. Range Factor of 2
Ozone 46% 5-79% 47%
Carbon Monoxide 47% 4-81% 54%
Nitric Oxide 52% 0-90% 64%
Nitrogen Dioxide 59% 5-100% 45%
Analysis of the results indicated that the "poor" runs contributing to
these results were associated with the minimum mixing depth assumption
used for surface inversion conditions (which had been Set at 30 meters)
and calculations of breaking inversions and that the "good" runs were
associated with middle range mixing depths of 55-500 meters. Reruns
of some of the poor runs with the minimum mixing depth assumption
changed to 100 meters showed dramatic improvements. Rank order analy-
sis of the data for nine runs involving four stations where the mixing
depth calculations appeared to be reasonable yielded a positive rank
correlation coefficient of .84, significant at the one percent level.
These results indicated that further work was particularly needed on
assumptions regarding mixing depth calculations. This work was carried
forward into the current contract period.
A sensitivity analysis was also conducted to establish the relative
importance of various design, input and program features of the model.
The results are reported in detail in the above mentioned report. In
general:
* Calculated as deviation over larger (calculated or observed) value.
0% = perfect agreement; 100% = maximum deviation possible.
-------
1.36
1. Results are strongly sensitive to minimum mixing depth
assumptions.
2. Results are strongly sensitive to traffic emission rates and
to traffic behavior resulting from movement of the diurnal traffic
curve two hours earlier. Concentrations are affected by as much as 50%.
3. Reductions in ozone are roughly proportional to reductions in
uv intensities.
4. The results are sensitive to background concentrations (initial
values at start of trajectory), but the degree of sensitivity is tra-
jectory dependent.
5. Sensitivities to emission rate increases or decreases are
highly trajectory dependent. Contaminant concentrations are never
linearly proportional to emission reductions. In some instances, sig-
nificant changes in air quality were shown to be highly refractory to
small changes in emission rates.
6. The results exhibited notable sensitivity to assumptions con-
cerning reaction rate constants for the production of singlet oxygen
by reaction between ozone and nitric oxide; slight sensitivity to
changes in certain rate constants; and moderate sensitivity to the
density of the wind station network. The study indicated that the
location of wind stations is as important as the density of such stations.
2. Current Project
The current project began in March, 1972, after more than a year
in which no additional work had been conducted. The following was per-
formed:
1. The existing REM program was completely reviewed and improved
wherever possible from the standpoint of accuracy and efficiency.
2. Special interpolation routines were added for smoothing wind,
relative humidity and temperature data (meteorological module) and
emissions data (emissions module) to eliminate time-step and cross-grid
discontinuities.
-------
1.37
3. A sensitivity study was performed on the minimum mixing depth
assumption and the minimum mixing depth value was set as an input
parameter to 50 meters.
4. A new emissions module was developed which combines area and
vehicular sources. New emissions data, supplied by Systems Appli-
cations Inc., were input to this module.
5. A new 32-step chemical mechanism was developed as a result of
simulations of chamber data for 26 gaseous mixtures that had been
supplied by EPA.
6. A new numerical integration mechanism was adapted to the model.
This is the Adams option of the Gear routine. The back substitution
option of the Gear routine was also made to work. This necessitated
reducing the order of the chemical mechanism matrix from 11 to 7.
7. Carbon monoxide validations were performed on 67 trajectories
at selected receptor points for 6 days in the Los Angeles Basin.*
8. Ground level concentrations for all photochemical contaminants
were computed at four receptor points for 6 days (24 trajectories) in the
Los Angeles Basin, and for an additional 8 trajectories for one of these days.*
9. A training session was successfully conducted by PES staff at
EPA in the Research Triangle Park on October 25, 26 and 27, 1972.
The model was easily operated by EPA personnel on the IBM System
370/165.
As a result of the above efforts, the new version of REM exhibited
considerable improvement in speed and accuracy. For example:
• CO is less than a factor of 2 in more than 80% of the compari-
sons; 40% agreement within one part per million. The 80% figure may
compared to 54% in the previous version.
• 0_ is within a factor of 2 in more than 70% of the runs as
compared to 47% in.the previous version.
• NO is in 90% agreement within .02 pphm. In a previous version,
NO was within a factor of 2 in 64% of the comparisons.
All simulations were conducted "hands-off", i.e., no runs were "fitted"
or "calibrated", hence the validation results are not influenced by these
kinds of manipulations.
-------
1.38
• NO- is within a factor of 2 in 60% of the runs as compared to
45% in the previous version.
Continuous improvements in the speed and usability of REM were also
attained. Computer speeds are expressed in terms of real-time-to-
simulation ratios. For the IBM Systems 360/65 and 370/155, the follo-
wing average ratios were attained:
1969, first contract 10:1 Hamming method
1970, second contract 50:1 Hamming method
1972, current project 150:1 Gear method, Adams option
When carbon monoxide simulations are conducted alone, REM runs at
about 3000:1. The type of computer used, of course, affects speed.
On EPA's IBM System 370/165, a full chemical simulation of REM over 6 1/2
simulated hours ran at a speed of 550:1, or 40-45 seconds for a 6 1/2
hour trajectory. For carbon monoxide alone, the speed would be 11,500:1.
3. Recommended Improvements
The results of this and past projects indicate that REM is a more
realistic, accurate and useful method for testing control strategies
than are rollback or other available methods. Unlike the rollback
technique, REM can determine the impact on air quality of various
source reduction schemes, singly and in combination, by time and
location taking into account varying configurations of point and area
sources, both moving and stationary, and the chemical reactions that
take place among air pollutants under varying physical and environ-
mental influences.
The user may attain the confidence level he may require as an
outgrowth of his use of the model. Through the normal processes of
validating the model for any air quality control region, providing
improved data or modifying input parameters, REM becomes highly
adaptable to local conditions, and its accuracy can be continuously
improved. REM is readily modified; for example, the chemical mechanism
*Full chemical simulations can approach this speed depending on the
chemical mechanism employed. Chemical mechanisms smaller than those
used in this project are available. These would substantially improve
running time.
-------
1.39
can be altered or replaced entirely, various transport and mixing
assumptions can be attempted and modules can be deleted or added.
The greatest spur to the practical refinement of the model, in
fact, would be in its widespread use for control strategy and opera-
tional purposes since the natural uses of the model inherently involve
validation and sensitivity testing, the results of which will indicate
the further improvements that can be made to the model.
The degree of .validation achieved by REM and possibly other photo-
chemical models appears to be well within the degree of accuracy that
has been attained by diffusion models. Further improvement in accuracy
and speed by working with the model as it is now constituted, of
course, is possible, but major improvements will depend on obtaining
more and better field and chamber data.
A summary of desirable further work on photochemical modeling in
general, and REM specifically, is described below:
1. In the future a closer interplay between chamber irradiations
and simulations of chamber results should take place. New experiments
should be designed to answer the questions about the mechanism that
need answering, and continuous critique and feedback should be pro-
vided for.
2. An improved method of estimating mixing height for the eastern
portions of the South Coast Air Basin should be developed. This will
likely depend on more extensive field study of vertical temperature
structure in the region.
3. An improved representation of the hydrocarbon mix
for input to the chemical mechanism, and appropriate elementary reac-
tions, should be studied and developed.
4. More experimental work is needed to determine the effects of
w?*:er vapor and sulfur dioxide on the chemistry. Generation of
aerosols by photochemical activity and the effects of aerosols on
-------
1.40
photochemical activity should also be determined.
5. An option should be provided to. utilize carbon monoxide as a
ventilation index taking into account emissions input.
- 6. Additional elementary reactions should be explored (by cut-
and-try methods) in an attempt to cut down the excessive simulated rate
of conversion of NO to N0_ and to reduce the overestimation of nitrogen
dioxide in the current REM chemical mechanism.
7. Improvement of existing, and search for a new, numerical inte-
gration procedure should continue. This could include use of system
identification methods, examination of tailor-made vs. general purpose
procedures, further chemical mechanism parameter optimization, mini-
mizing set size in validated mechanisms, and balancing accuracy vs.
speed tests.
8. Alternative methods for inputting mixing height information,
including multiple radiosonde information, aircraft monitoring data,
and assumed mixing height information should be provided.
9. Photochemical models should be linked with routine 24-48 hour
air pollution forecast and emergency warning procedures. This will
require analysis of objective forecasting procedures, study of mega-
meso-microscale relations, and linkage to air quality and meteorologi-
cal data handling systems.
10. A synthetic trajectory application capability is required to
permit use of the model in all data availability situations.
11. Automatic trajectory plotting or computer graphics to minimize
manual plotting of trajectories from output would be desirable.
12. A universal U.V. Module which would generate solar angle as
function of time, geographic location, and calendar date is desirable.
13. An improved version of the REM point-source diffusion routine
should be developed.
-------
1.41
14. Add user conveniences including displays of trajectory emissions
output, automatic validation comparisons, summaries of emissions input
to air parcels, etc.
15. Linkage of reverse trajectory routine to the main model, and
packaging of REM as an automatic multi-receptor point episode model.
G. EVALUATION CRITERIA
The establishment of objective criteria for the evaluation of atmos-
pheric simulation models, particularly photochemical models, is an
extremely difficult matter. Both quantitative and qualitative factors
must be considered, as well as subjective judgment. Further, it is
difficult to evaluate models based on limited applications.
PES, throughout its experience in developing REM, has kept the
following criteria in mind:
1. Cogency - the employment of fundamental and realistic principles
consistent with knowledge of physical and chemical reality
such that the model responds realistically and sensitively to
the wide range of input conditions to which it may be subject.
The question of cogency is described in greater detail in
previous sections of this report.
2. Validation Accuracy - these criteria may be looked at from the
standpoint of the statistician and the "responsible user" of
the model. The statistical evaluation of REM is treated through-
out this report. The responsible user is considered to be the
air pollution control officer, urban planner or other individual
with administrative responsibility who must decide whether or
not to use the model and what degree of confidence can be
placed in its use. Given the limits of modeling, the most
important criteria that can be applied may be in part quanti-
tative, in part qualitative. PES assumes the following cri-
teria in their order of importance.
a. The model should be capable of distinguishing among high
and low oxidant days. It should exhibit a high degree of
success in predicting episode days.
b. The variation in predicted contaminant levels by time of
-------
1.42
day should be parallel to variation in the observed levels by
time of day even if significant quantitative differences
may occur; the model should exhibit similar results for
dosage calculations.
c. The variation in predicted contaminants by general geo-
graphical location should be consistent with the observed
values by location.
3. Potential for Improvement and Usability - these are very•impor-
tant criteria, particularly with respect to the current genera-
tion of photochemical models. The model should be (a) easy
to use; (b) easy to modify; (c) adaptable to many urban areas;
(d) adaptable to different degrees of data availability and
conditions of use; and (e) economical to use.
Frequently, some of these criteria are ignored in model
development. Yet, the accuracy of a model at any particular
point in its stage of development may be misleading. Given
a cogent and cost-effective model, further development in
accuracy, speed, and application will depend more on its
widespread use in practical situations than on a preconceived
development program. For accelerated development to take place,
the model should be easy to use and understand, and should be
made available for general use.
-------
II.1
Chapter II
DESCRIPTION OT1 THE REACTIVE
ENVIRONMENTAL SIMULATION MODEL (REM)
A. INTRODUCTION
The REM effort has been directed towards the construction of a
sophisticated photochemical mechanism which is capable of closely
replicating smog chamber data and then extending this capability to
the urban atmosphere. REM is based on a Lagrangian (moving coordinate)
system. This system enables the numerical computation of the chemical
i
reactions to take place in a moving parcel or column of air. The
parcel Is bounded by the inversion base above and the ground below.
The locations, of the base of the column at successive points in time
describe the path or trajectory that the air parcel traverses across
the region. The trajectory Is computed by special routines contained
in the REM program from wind velocity information contained in the data
base as a function of time of day and location. An overview of the
model's operation is presented in Figure II-l.
The computer program has been designed as a detailed episode model
for chemical kinetics, with, additional modules linked up to form a
complete atmospheric system. The modules presently in the system deter-
mine the necessary meteorological parameters, the rate of absorption
of actinic light by N02, and emissions due to traffic and area sources.
A schematic diagram of REM appears in Figure II-2. The facility exists
to incorporate modules to calculate the emissions due to large point
sources and the effect of vertical diffusion should practical and
effectual methods be found to perform these tasks. A mathematical
description of REM may be found in Volume II of this report.
-------
II.2
PHOTOCHEMICAL MODELING
Meteorological stations provide
current data on winds, tempera-
ture, pressure, dew point, etc.,
to the model for predictive
movement of air parcels through
the various grids.
For pollution modeling, the air space within each
geographic grid is assumed to emit an approximate
mean value of pollutants for a given time. The
model includes nitrogen oxides, high and low re-
active hydrocarbons and carbon monoxide. Values
are provided by local environmental control
agencies.
Pollutant monitoring stations
measure types and amounts of
pollutants, i.e., oxidants, hydro-
carbons, nitrogen oxides, car-
bon monoxide.
TERRAIN
INVERSION LAYER
The column of air
is one square meter
in cross-sectional
area. Its height is
determined by ter-
rain and inversion
layer conditions.
The latter are de-
rived directly from
radiosonde meas-
urements.
The model predicts the amount and
type of photochemicals present in a
column of air as It traverses a com-
puted trajectory path to a pollutant
monitoring station. At this point
predicted and observed values are
compared.
This model simulates a region's photochemical pollutant content
by representing in a computer the chemical reactions occurring
in a column of air as it traverses a dynamic wind trajectory. The
model is tested by acquiring appropriate meteorological and
pollutant data from a dense network of monitoring stations.
The path of a moving parcel of air Is computed on the basis of
wind station data. As the parcel passes over the urban area It
picks up pollutants from vehicular traffic, power plants, factories
and other sources. Pollutants, or emission rates, for each grid are
contained in the data base. With the sun's energy acting as the
catalyst, the pollutants are transformed Into a variety of photo-
chemical pollutants, producing the well known effects of smog.
Predicted concentrations for these pollutants are printed out for
points along the air parcel path and compared with values meas-
ured at actual air monitoring stations.
COMPUTER
In a given area this photo-
chemical model could predict the effects of
new smog control devices for any given time
period after Introduction, determine a pro-
posed freeway's effect on pollutant concen-
trations, or the contributory effects of new
industries and proposed rapid transit sys-
tems. Decreases in pollutants that would
result if modified or advanced propulsion
systems are used In vehicles or if nuclear
power plants are installed also could be
predicted by the model.
Figure II-l. PES Reactive Environmental Model (HEM)
-------
II. 3
DATA BASE
WIND, TEMPERATURE, HUMIDITY,
EMISSIONS INVENTORY, RADIO-
SONDE CURVE, SOLAR ZENITH
CURVES, EMISSI~ON~FACTORS"
/INITIAL
CONCEN-
TRATIONS/
METEOROLOGICAL
EVALUATION
DETERMINATION
OF EMISSIONS
INPUT INTO COLUMN
MIXING
DEPTH, TEM
HUMIDITY
WIND
ABSORPTION
RATE
OF NO
EMISSIONS
RATES
RATES OF
CHANGE OF
CONCENTRA
TIONSAND
LOCATION
SMGOUT&SMGPLT
OUTPUT
TABLES & GRAPHS
Figure II.2.
Schematic of the PES Reactive
Environmental Model (REM)
-------
II. 4
B. GRID SYSTEM
The geometric setting for REM is a standard two-dimensionsl coor-
dinate system. REM utilizes the mile as the standard unit of measure
with all urban grid squares chosen as 4-square miles. The kilometer
can also be used as the standard.
To set up the grid system, the origin is placed at a point such
that the region of interest resides entirely in the positive quadrant.
The coordinate positions of all air monitoring and meteorological sta-
tions are established in either miles or kilometers. These values are
then fed into the model at execution time. In this manner, REM can be
applied in any region.
Figure 11-3 illustrates the grid system used by REM for the Los
Angeles Basin. The locations of the air monitoring stations are desig-
nated as circles and the coordinate points are given in miles.
C. PROGRAM CODE AND MACHINE USE
REM is coded in the FORTRAN IV computer language and has operated
with both the Level G and H compilers. It is thus operational on any
computer system with 150K in core and a standard FORTRAN compiler.
The Los Angeles validation version of the model has been success-
fully run on the IBM System 360, Models 50 and 65, and the IBM System
370, Models 155 and 165. The ratio of simulation time to computer time
used was approximately 150:1 on the 370/155 and 550:1 on the 370/165.
In regions where much less data is available than in Los Angeles, REM
will operate at even greater speeds.
D. OPERATION OF REM
REM has been designed to operate as a multi-receptor point model.
The user is required to provide a minimum number of inputs after the
data base has been prepared. In its execution mode, the model trans-
ports a parcel of air, bounded by the ground and the inversion layer,
-------
II.5
;7 29 31 13 35 37 » 41 <3 45 47
j
Figure II.3. REM Grid System for the Los Angeles Basin
-------
II. 6
along a dynamic trajectory that is generated utilizing wind data from
numerous meteorological stations. At each point along the path, the
speed and direction of the trajectory are recalculated.
The user may select the starting point or the terminus for each
simulation run. In either case, he must indicate the length and starting
time (of day) of the simulation, and initial values for the major pollu-
tants.
If the user selects the terminus, he must first operate the reverse
trajectory program (see Section F.10) to determine the correct starting
position and major pollutant values. The starting location is necessary
because this point must be input into REM in order for the trajectory
to reach the selected terminus at the end of the full simulation run.
The initial pollutant concentrations are determined by interpolation of
observed values at air monitoring stations to the trajectory starting
point. Once the location and starting values have been calculated for
a simulation run, the same trajectory may be run many times over with-
out utilizing the reverse trajectory program again.
If the user executes the model by selecting a starting point, he
may wish to choose a monitoring station where the initial pollutant
values will be known. As an alternative, he may select another point of
interest, or pick a completely random location. For these latter cases,
some assumptions may be made regarding the starting values, or minimum
or standard background values may be input.
E. CONTROL STRATEGIES
After a given full simulation run has been made, it may be consi-
dered as a standard run for the purposes of evaluating particular con-
trol strategies. Control strategies are implemented to reduce pollu-
tant concentrations and to attain required air quality standards.
-------
II. 7
By varying any of the numerous emissions inputs to the REM, either
individually or in combination, the possible effects of control strategies
can be determined before they are actually implemented. These strategies
may include for example: reducing total traffic, reducing traffic in
specified areas, reducing emission rates, varying traffic patterns, or
altering operating times of area sources.
F. REM MODULES
The modular construction of REM has facilitated the model's
development and alteration, when necessary. Any given module may be
unplugged and replaced as desired; or special modules may be created
and added to handle special situations.
1. Initialization Program
The initialization program reads in the data base and the infor-
mation needed for a complete simulation run. This program converts
all of the data, except station locations, to the metric system for
easy utilization by the other modules, and calls the numerical integra-
tion, output and plotting routines as required by monitoring the simu-
lation time. At the end of the run, it checks to see if another simu-
lation is to take place. To make additional runs, only control informa-
tion must be input thereby facilitating the operation of REM for the
user.
The initialization routine outputs the data base that has been
read in to provide the user with a permanent record of the information
employed in the simulation. A complete list of the data categories
required for input follows:
• Trajectory start position and time of day;
• Length of simulation;
• Starting pollutant concentration values;
-------
II. 8
• Control parameters for numerical integration routine;
• Meteorological stations - location and elevation;
• Meteorological variables - hourly data (wind speed and
direction, temperature, dew;point);
• Emission factors for freeway and street traffic in grams
per vehicle mile;
• Grid distribution of hourly area emissions;
• Radiosonde data - temperature vs. elevation profile; and
• Solar zenith angle as a function of time.
2. Numerical Integration
The numerical integration module is the heart of REM. This routine
accepts as input derivatives calculated by the kinetics and meteoro-
logical modules and computes pollutant concentrations and the parcel
position forward in simulated time. The kinetics module supplies the
derivative of each of the photochemical pollutants computed as a
function of time. The meteorological module provides wind velocity
outputs as the time derivatives of the position vectors.
The economic use of REM is dependent upon the numerical integration
routine. By attaining larger step sizes, the program will execute
faster, use less computer time, and be less expensive to operate.
REM utilizes the Adams method of integrating ordinary differential
equations of the Gear numerical integration package. A sensitivity
analysis has been conducted on the' Adams'routine in REM. As a result
of this examination, it was found -that the Adams method executes most
economically with the maximum step" size at 0.1 minutes and the error
limit at .05 in pollutant concentration units.
3. Chemical Kinetics
The chemical kinetics module has operated on reaction mechanisms
C. W. Gear, DIFSUB for Solution of Ordinary Differential Equations,
Communications of the ACM, Volume 14, Number 3, March 1971.
-------
II.9
containing 13-41 reactions and is Based on a set of stoichiometrically
valid elementary reactions. The current version contains a 32-reaction
mechanism that preserves the flexibility of chemical detail necessary
for adequate simulation of Both primary and secondary contaminants
Csee Chapter III). The set of elementary reactions, rate constants and
chemical species is easily modified and expanded as required.
This routine calculates the elementary reaction rates and the
derivatives of each of the accumulating chemical species as they change
with time. It calls the meteorological, ultraviolet and emissions
modules into operation, as necessary. The kinetics program adds the
perturbations from the traffic and area sources to the derivatives
computed for the chemistry for input into the numerical integration
routine.
4. Meteorology
This module consists of a series of related routines that employ
wind speed and velocity, temperature, relative humidity, elevation,
and radiosonde data to calculate the meteorological values, mixing
depth and rates of change in the wind at the pollutant column position.
a. Spatial Interpolation .
Wind speed and direction, temperature, and relative humidity
undergo Both spatial and temporal interpolations from numerous
meteorological stations to the trajectory location of the air
parcel. The spatial interpolation is inversely proportional to
the square of the distance from station to parcel. This procedure
has two advantages over simply using a ratio inversely propor-
tional to the distance involved. Tirstly, it ensures that the
meteorological values assigned to the trajectory position are
more closely in keeping with the data at the nearest stations; and
secondly, it permits the computer program to execute faster By
eliminating the necessity of computing square roots.
-------
11.10
b. Temporal Interpolation
\
The temporal interpolation is linear. The meteorological \
items for the current hour and the next hour are utilized in this
process. The portion of each item contributing to the result is\
directly proportional to the number of minutes in the hour that
have passed. For example, at 10:15 A.M. one-quarter of the value
for hour 10 would be added to three-quarters of the value for hour 11.
c. Mixing Depth
The model calculates mixing depth on the basis of a temperature
versus height profile derived from temperature soundings at various
altitudes aloft recorded by radiosonde measurement or aircraft.
The mixing height is taken as a function of temperature at the
ground and the most recent vertical temperature soundings. The
mixing depth is then obtained by subtracting ground elevation at
the location of interest. The ground temperature is interpolated
at the trajectory point from meteorological data at each station.
In this technique the temperature versus height profile is assumed
fixed during the day unless more frequent soundings are provided.
The mixing depth changes with variations in ground temperature and
elevation.
d. Trajectory Derivatives
The rates of change of the wind components at the air parcel
position are utilized by the numerical integration routine to
calculate the next trajectory position. Two differential equations
are reserved for the trajectory calculation. They employ these
derivatives in the same manner in .which the integration of the
differential equations for the chemistry occurs.
-------
11.11
5. Barrier Check
A "barrier" is aft imaginary line or set of connected line segments
superimposed on the region of interest with the coordinate system for
that area. Its purpose is to represent a topographical feature such as
a mountain range which may serve to separate sections of an air quality
region. The mountain range may tend to influence differing climato-
logical characteristics and wind regimes. F,or example, on a given day
in the Los Angeles Basin, the meteorological characteristics of the
Westwood area are likely to differ from those in Van Nuys due, in part,
to the intervention of the Santa Monica Mountains. Experience in the
Los Angeles area has shown that significant temperature and wind regime
differences between the San Fernando Valley and the central Basin tend
to occur on any given day.
REM contains two such barriers. One bisects the Santa Monica
Mountains and separates the San Fernando Valley from the main section
of Los Angeles. The other barrier passes through the Puente and San
Jose Hills at the east side of the Basin.
The main function of the barrier routine is to determine the
"permissible" meteorological stations. Permissible stations are simply
stations which are located in the same general section of the Basin as
the trajectory point of interest, as defined by the barriers. That
is, both the trajectory point and all stations to be used in the cal-
culation of the trajectory path and the interpolated meteorological
Rvalues at any trajectory point are located on the same side of each
barrier. This is graphically shown in Figure II.4.
The barrier check is intended to ensure that a station which is
very close in distance to a trajectory point of interest, but separated
from the trajectory point by a mountain range does not exert any
influence on the calculation of the trajectory path, temperature, and
relative humidity at the trajectory point.
-------
M
(Wind Direct!onV
M
8
•
M9
Wind direction
Line T - T. is the trajectory path
a D
T, - 1 trajectory point of interest
' G
^2 " ^a Downwind trajectory point along the same trajectory
M. meteorological stations
P-, Permissible meteorological stations
1 for T, (M4, M5, M6, V(Jt Mg, Mg)
P? Permissible meteorological stations
• for T2 (M2, M3, M5, Mg, M?, Mg, Mg)
Figure II. 4 Demonstration of Barrier Check Technique
-------
11.13
6. Ultraviolet Module
This module calculates a diurnal ultraviolet irradiance function
based on measurement of sky cover, latitude and local calendar time.
The irradiance determines the specific ultraviolet absorption rate of
N02» which is used in the kinetics module.
7. Source Emissions
This routine computes emissions of NO, CO, reactive hydrocarbons
and less reactive hydrocarbons for freeway and street traffic, and
calculates emissions from area sources of NO, reactive hydrocarbons,
and less reactive hydrocarbons.
The inputs to this program for vehicular sources include two
2
diurnal traffic curves (see Figures II.5 and II.6), emission factors
for reactive hydrocarbons, less reactive hydrocarbons, CO and NO in
grams per vehicle mile, and distributions of daily vehicle traffic on
streets and freeways given in a system of 650 4-square mile grids that
2
cover the Los Angeles Basin (see Figures II.7 and II.8). The emissions
of the above pollutants are computed as a function of time by applying
their respective emission factors to the traffic levels at the air
parcel location qualified by the daily traffic usage curve.
Emissions from area sources are also calculated as a function of
time. A constant diurnal usage curve is applied to grid system dis-
3
tributions of hourly pollutant emissions of NO, reactive hydrocarbor
and less reactive hydrocarbons (see Figures II.9, 11.10 and 11.11).
2
Appendix A, Development of a Simulation Mpdel for Estimating Ground
Level Concentrations of Photochemical Pollutants by P. J. W. Roberts,
P. M. Roth and C. L. Nelson; Report 71SAI?-6, March 1971, prepared by
Systems Applications, Inc.
3
Oxides of nitrogen taken as NO
-------
z
o
u
u
Q
u.
O
.10
.09
.08
.07
.06
.05
.04
.03
.02
.01
T I I I I I I I I I T
\
I I I I
I I I I I I
24
10 12 14 16 18 20 22 24
DAILY HOUR
Figure II.5. Temporal Distribution of Daily Street Traffic
(Interpreted from Data in Reference 2, pg. 11.13)
-------
CO
z
D
O
u
O
E'
K
O
u.
O
z
O
O
.10
.09
.08
.07
.06
.05
.04
.03
.02
.01
M
M
Ul
I I
:4 2
I
"4
I
6
I
8
I
10
1
12
1
14
1
16
1
18
I
20
I
22
I
24
DAILY HOUR
Figure II.6. Temporal Distribution of Daily Freeway Traffic
(Interpreted from Data in Reference 2, pg. 11.13)
-------
11.16
35
60
137
239
139
25
9
0
47
==
12
155
255
250
118
30
25
18
64
.
34
202
238
249
107
17
0
0
99
^*T- i
149
223
238
250
194
12
2
7
117
K
60
146
252
266
264
182
59
30
98
231
476
X
125
172
221
295
414
350
63
70
283
512
254
305
v
\
SO
135
152
251
328
268
121
32
403
346
338
255
266
261
^89
1\4
,\
1
X
V
v
43
85
233
276
280
155
260
549
459
261
255
413
296
255
215
360
239
121
49
"V
28
66
54
212
17
84
20
2X5
247)180
•
116 25
478 375
512 572
478
468
412J413
311 478
392 331
397
328
285
358
412 1322
324
304
188
231
249
266
28 |142
^25 J176
>
1
73
27
4
204
245
259
473
642
512
477
393
314
176
210
12.1
175
254
lfr
r
5
0 0
7 i 2
157 50
i
0. 0
«; o
131 58 91 ! 113
204 21
242 191
164 116
226 1 217
579
328
364
273
254
265
121
79
94
157
290
324
292
322
292
293
173 ! 241
162 318
I
1901236
229
285
199 186
293 J264
273 j299
305 J265
255 ;386
257 1298
270 312 301
204 288 i430
i
I |
283 192 282
413 | 386
%£fyr
317
•5.A
2
0
21
212
308
227
243
75
195
193
253
263
241
93
204
0; 0
1! 0
1
156
241
236
250
108
127
185
190
229
230
96
138
165 ! 155
162
175
75
60
X<.48
1
2
91
218
185
242
173
43
69
295
234
135
104
186
173
116
125
114
24
X
0
°
0
57
183
160
114
131
39
21
119
181
114
270
305
259
209
212
195
133
126
X,
1 0
0
0
1
77
130
190
180
227
82
42
240
62
179
242
228
220
200
116
135
107
104
^
0
0
0! o
0
0
6
126
155
280
142
121
42
22
156
2
2
• OJ 2
i i .. 2
0
0
0
0
o! o
1
0' O
'6.2 o i 0
149 ; 151
232 i 143
244 i 71
18 48
45 | "53
i
90 i 33
1 37
103 , 21
1291114
290
323
304
95
197 82
241 113
169 123
255 i 149 168
i
1 i
219 | 249 [266
166 203
116
248
277
s218
74
215
94
196
190
80
100
114
52
119
76, 32
66! 77
lo' 99
50' 73
27.' 5
12' 0
i! 5
9J 7
57
48
4
30
3l| 6
109: 11
163
182
98
11
29
10
0
18
28
32
19
26
15
0
0
0
Figure II.7. Distribution of Street Traffic by Grid Area
(thousands of vehicle miles per day)
[From Reference 2, pg. 11.13]
-------
11.17
59
169
«>•»••.
271
-•M^
311
^»,
383
\
!I66
94
155
242
222
890
352
356
242
101
N
119
68
456
137
473
29S
\
\
144
ISC
126
426
141
511
295
\
\
\
/
\
"^
158
17
494
.140
333
51
160
404
348
351
-'^->
109
104
178
275
215
426
32
391
^
150
395
194
205
466
326
-x
190
423
290
436
795
531
391
366
305
257
217
625
131
143
tf
r
51
335
620
452
449
343
45
¥
13
193
257
585
623
1S9
241
236
545
107
*tr
37
BO
109
257
518
576
283
281
302
131
340
— -^
28
297
224
332
129
56
323
306
79
444
527
314
383
278
195
529
X
311
234
205
265
33
30
289
102
349
\
41
254
285
263
104
243
59
213
181
\
74
362
147
156
253
83
144
112
153
\
66
291
70
12
330
183
78
US
s.
65
214
59
251
145
25
150
160
80
V18
43
203
42
15
138
68
398
163
97
6
"^
11.
no
21
108
156
257
67
148
95
\
,
173
32
3
1
207;
137
161
131
169
61
159
.!<
1!
16
85
77
91
59
Figure II.8. Distribution of Freeway Traffic by Grid Area
(thousands of vehicle miles per day)
[From Reference 2, pg. II.13J
-------
11.18
0
0
8
8
8
0
0
0
8
— ••—
0
I
9
9
8
0
0
0
8
1=1=
0
1
9
13
12
0
0.
0
8 '
0
1
18
18
14
0
1
1
9
\
9
10
il-
ls
12
4
3
6
.19
9
X
9
10
15
15
8
4
7
10
19
18
19
S'
\
9
12
15
17
8
4
9
10
19
20
21
25
29
\28
\
*
9
.12
• is
15
8
8
18
10
. 20
20
21
25
29
29
21
21
0 1) 19
/
V
V
4
18
18
6
1
11
12
12
9
9
19
11
20
20
20
22
25
25
27
27
24
21
21
^
1
9
1
7
10
10
19
12
21
19
19
20
22
22
34
34
29
23
23
20
^
0
7
1
4
10
10
19
12
21
19
20
20
22
22
34
34
29
23
t
fc
0
2
8
8
9
9
19
12
21
37
38
31
24
24
25
25
30
35
¥
0
0
4
8
1
9
19
12
21
37
38
31
24
24
25
25
30
35
^
0
0
1
8
9
18
18
11
20
29
29
26
24
24
24
24
26
28
"i>
0
0
0
8
10
16
18
10
19
21
21
22
23
23
23
23
22
21
^
0
0
0
3
10
8
18
10
19
21
21
22
23
23
23
23
22
21
13
^
0
0
0
0
16
16
19
13
22
9
19
22
27
27
19
19
18
11
11
11
\
0
0
0
0
19
18
19
13
22
1
5
18
27
27
19
11
11
11
11
11
11
\
0
0
0
0
16
18
20
12
21
6
3
7
23
23
11
11
11
11
11
11
11
11
X
0
0
0
0
14
14
20
11
20
15
9
0
11
11
11
11
11
11
11
11
11
11
11
K
0
0
0
0
13
13
20
11
20
19
2
0
11
11
11
11
11
11
11
11
11
11
11
11
\
1
0
0
0
0
9
9
17
9
1
18
1
0
5
11
11
11
11
11
11
11
0
11
11
11
^.
0
0
0
0
9
9
17
9
1
IB
1
0
0
8
11
11
11
11
11
11
0
0
0
0
0
V,
0
0
0
0
9.
7
9
1
1
9
0
0
0
0
2.
s
8
11
11
0
0
0
.0
0
0
0
0
0
0
8
0
9
10
10
8
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Figure II.9. Distribution of Reactive Hydrocarbon Emissions
(average kilograms/hour) by Grid Area
[From Reference 2, pg. 11.13]
-------
11.19
0
0
30
30
30
0
0
0
30
0
0
31
31
31
1
0
0
30
0
0
31
31
31
1
0
0
30
i
]
1
|
I
1
3
35
35
30
0
0
5
35
\J
-
32
36
39
39
30
15
5
24
39
40
X
32
36
39
39
30
15
15
39
39
40
40
\3°
\
31
43
53
53
34
19
23
33
33
45
45
51
56
V51
\6
6\
1
/*
Vs
i>
31
42
53
53
34
34
33
33
33
45
45
51
56
56
31
31
31
30
30
<^^—
15
1
40
48
48
38
38
36
35
35
47
47
»
56
56
46
46
39
32
32 .
^
i
i
0
37
13
33
42
42
39
• 37.
37
49
49
51
56
56
61
61
47
33
33
31
5s1
0
27
13
23
42
42
39
37
37
49
49
51
56
56
61
61
47
33
of
III
0
10
31
31
36
36
48
60
60
110
110
90
70
70
36
36
34
33
%
0
0
16
31
6
36
48
60
60
110
110
90
70
70
36
36
34
33
>/*n
0
0
5
30
35
35
44
53
53
82
82
69
57
57
36
36
34
33
-18-
0
0
0
25
35
35
40
45
45
56
56
50
45
45
36
36
34
33
^
0
0
0
15
35
35
40
45
45
56
56
50
45
45
36
36
34
33
14.
"X
o
0
o
0
37
37
49
62
62
34
34
35
36
36
34
34
32
-- ^
U
11
11
\
o
°
0
0
•
37
37
49
62
62
4
19
35
36
36
34
15
13
»
11
11
11
X
0
0
0
0
39
39
43
49
49
22
12
20
33
33
13
13
11
11
11
11
11
11
V
0
0
0
0
41
41
36
36
36
39
34
4
11
11
11
11
11
11
11
11
11
11
11
\
0
0
0
0
41
41
38
36
36
39
9
4
11
11
11
11
11
11
11
11
11
11
11
10
0
0
0
0
30
30
31
31
1
32
2
0
S
11
11
11
11
11
11
11
0
11
11
11
•>-
c
0
0
0
15
30
31
31
1
32
2
0
0
6
11
11
11
11
11
11
0
0
0
0
\°
0
0
0
0
0
20
32
4
4
1
1
0
0
0
2
5
10
11
11
0
0
o'
. 0
0
0
0
0
0
0
0
10
30
30
30
0
' 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Figure 11.10. Distribution of Oxides of Nitrogen Emissions Taken
as Nitric Oxide (average kilograms/hour) by Grid Area
[From Reference 2, pg. 11.13]
-------
11.20
0
0
60
60
79
0
0
0
79
ess
0
3
84
84
81
2
1
0
79
t=aa
0
3
84
84
81
2
1
0
79
-• —
6
8
105
105
81
2
11
20
99
X
91
108
125
125
80
41
21
79
118
120
X
10
91
108
125
125
80
41
41
118
118
121
120
\"
X
85
133
180
180
96
57
74
90
90
145
145
189
233
V"
V
24\
3
/*
\60
11
85
133
180
180
96
96
93
90
90
145
145
189
233
233
96
96
78
79
79
-^^~
26
3
101
158
158
113
113
105
98
98
150
150
179
214
214
173
173
130
96
96
><72
0
107
56
116
129
129
118
106
106
156
156
175
193
193
251
251
182
111
111
95
>
0
83
56
76
129
129
118
106
106
156
156
175
193
193
251
251
182
111
ft
r
0
22
81
81
103
103
153
204
204
426
426
333
245
245
131
131
144
156
*
*
0
2
52
81
24
103
153
204
204
426
426
333
245
245
131
131
144
156
>^T
0
1
11
80
102
102
136
174
174
307
307
267
197
197
120
120
122
125
"3T-
0
0
0
70
101
101
122
143
143
189
189
170
150
150
108
108
101
94
^
0
0
0
30
101
101
122
143
143
189
189
170
150
150
108
103
101
94
55
^k
0
0
0
0
108
108
159
212
212
98
98
118
137
137
94
94
86
44
44
44
X
0
0
0
0
108
108
159
212
212
19
59
118
137
137
- 94
59
51
44
44
44
44
\
0
0
0
0
118
118
137
158
158
69
99
68
109
109
51
51
44
44
44
44
44
44
X
0
0
0
0
129
128
117
104
104
117
108
19
-14
44
44
44
44
44
44
44
44
44
44
X
0
0
0
0
129
128
117
104
104
117
38
19
44
44
44
44
44
44
44
44
44
44
44
.40
0
0
0
0
80
80
82
85
6
86
7
1
23
45
44
44
44
44
44
44
0
44
44
44
"&-
0
0
0
0
41
80
82
85
6
86
7
1
1
31
44
44
44
44
44
44
0
0
0
0
N°
!
! 0
0
0
0
2
62
90
21
21
4
4
1
0
0
5
20
40
44
44
0
0
0
0
0
0
0
0
0
0
1
21
97
115
115
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Figure 11.11. Distribution of Less Reactive Hydrocarbon
Emissions (average kilograms/hour) by Grid Area
[From Reference 2, pg. 11.13]
-------
TIME.- 0.380199370 03 .MINUTES
STEP = 66
PQSITION= 58.4039. 40.3573
STEP SIZE- A.IQOQOOOOD qo._ PAGE.. is..
EXPECTE.D RATE OF CHANGE OF CONCENTRATIONS DUE TO EMISSIONS FROM RE.GJONA.L SOURCES.
.... .. „ PP.HN/_M1N .._._ . ._. __ _._..-GRAM-MOLES/MJN "
._•. PRO.P.YLkE.Ni ._ Nfl CO ; -_.DUMHC PRQPYIENE NO CO DUHHC
-STRE1T_T_R.AFF.I_C 0.36600-02 0 , 102 50-01.... 0.2 I960. 00 0., *.?.! SO-01 ._.... 0.5274E-06 0.,A.477E-p5 . ...Q_._3165Eift* 0.7089E-05
TRAffJC 0...1*4 60-02 0.40490-02 .0 . 86J^7D-0 1, _.q._l_944D-qj. .Qi.2Pl*EM>6_-°.I*36E-p_6_ _0.jj25_l.L-P_i 0.2801E-05
Pil75.2D.-Pl-.-Q_.J,5A7D-_Q2. __. 0.72520-01 0^2 525 E-Q6__ 0.1_230E-06. 0.1045E-04
.: __...___ __-gjjj-jjg MUING" ~ """
TEHPERATURE HlJ^IpJjrY ELEVATION HEIGHT DEPTH
1CENT.) (METERS) (METERS) (METERS)
27.2~l"37 "" 0~.2»7'f Y»1V3918 " 757.9106 57" " '~'
19J!f.OUND___CONC_PPHM CHANGED rPHH/niH_ _
I N02 0.10570770 02 -0.15016660-01
I -NO __ 0.'7_1?6^9^D 00 .-O.J883168P-J2
"303 0."13083660"02 0.27856620-01
4 C3H6 0.582904AD 00 -0.46061620-02
5 HN03 " 0.29533170 fcl 0.11190490-01
^ CH_3CHP PO"13.!?3?!? PJ _~°^55*^?'l?5"0*
"7 H'CHQ "0.34705030 01" -0.70338840-62 """ "
? £2 (PPin 0.4JS4J978D ^l -0.3806332pT02_ _
9'CH3ffN02 "0.39845050 00 -6.15815400-03 " "
10 C2H3N05 0.55473800 00 0.41342320-03
11 F~IXED~02 0."6 0.0" " " "
12 OUMJHC 0^11756080 03 -0.33399040-01 _
13 " " 6.6" " -- -
14 H ^..1*162*60-07
15 HH ~ 0.1244696"6-04 " ~" ~ "
16 H02 0.52582540-02
17 :H3 0.68931480-08 '
18 CH30 0.40259910-04
19 "CH302 0.38551340-02
20 C2H3Q _0^ 14202760-06
21 C2H302 " 0.795388"60-03
2.2 C2.H4P2 0.10959920-05
2-3 Q. ' 0.230,38470-0-5 _ ' "
24 N03 0.1 116lo70-05 - --- -..--
25 C2H303 • 0.57995750-03
Figure 11.12. Sample REM Output
-------
11.22
8. Output Routine
This program prints output data at various times along the tra-
jectory route. The user may specify the times at which the printouts
are to occur and the routine checks to see if a requested output time
has been reached. A sample output is shown in Figure 11.12. The
primary data categories printed include:
• Time of output and trajectory position;
• Step size;
• Rate of change of concentrations from source routine
in pphm/minute and in gram-moles/minute;
• Meteorological values at the trajectory point including
mixing depth;
• Pollutant concentrations;
• Rates of change of accumulating (primary) pollutants; and
• Rates of the elementary reactions.
9. Plotting Routine
At the end of the simulation run, this program outputs three digi-
tal plots containing the history of each accumulating pollutant repre-
sented in time vs. concentration curves. An example of this output is
shown in Figure 11.13.
10. Reverse Trajectory Program
The reverse trajectory program calculates wind trajectories and
interpolates initial concentration values for use by the simulation
model. Its primary inputs consist of hourly average wind data moni-
tored at a network of meteorological stations, and hourly average
pollutant data recorded at a group of pollutant monitoring stations.
The coordinate positions of these stations are also required.
In addition, the user must specify a coordinate position and the
length of time for which the simulated trajectory is to be calculated.
The program then considers the given coordinate position to be the
-------
non o
r'.cir f
0.2
7.9
15.'"'
73.2
1C. 9
11.5
'• ') . 7
53.'-)
61. ft
69.7
76.9
84.6
92.7
99 -T
107.6
115.2
122.9
110.6
1.18.2
145.9
153.6
161.3
163.9
176.6
134.3
199.6
707.3
214.9
772.6
23'"1. 1
217.9
245.6
253.3
761.0
268.6
?7£. 1
764. 1
291.6
799. 1
107.-
314.6
322.3
335-2..
317.6
345.1
351.0
16C.6
361 .'
176.0
313.7
1 "UN SI^ML *T IPN 38=C.93C*1 02 PRHT (3>oQ. 1000D 00 PRMT (4 I »0. 10000-01
NVIPOKPNTU 5"V1CFS, INC, — PHP.TOCHE »IC»L MODEL («E1> — VESSION CP
•"^ 1f".-|fj<; flT 0700 HOUSS ON SEPTFMRFR 11, 1969
TO -\ l^.Tnqv TH OCUNTrWN LOS ANC.FLS AT 1330 HOURS
53 X4
53 241 _
51 241
s -> 7 *- 1
5 < 24 1
1
i
I
:•> 3 z 4 1
15 3 2 * I
1523 4 1
15234 1
j
|
1
152 43 1
157 43 1
157 4 3 1
t 5 2 4 3 1
[
1
I .....
• i 4 31
I 25 4 31 _
1 25 4 31
1
1
I 7 /. 5 11
I ?4 5 13
I 24 5 13
|
I ....
I
1X5 1 3
1 ?>• 5 1 1
24 5" 11
1245 13
I
I
I
1245 31
1745 31
1745 3 1
'245 3 1
1
I
1
1245 3
0.20
8.00
16.85
25.55
34.25
60.95
69.80
78.80
87.80
114.80
123.80
137.8?
141.80
168.80
177.80
186. CO
222. 3C
231.80
240.80
276.15
285.13
293. 9C
302.90
329.90
338.45
347.45
156.45
1^3.45
1." 7.15 4.30 6.44 ".59 10.74 . 12.89 15.03 17.18 19.31 21.46
to
CO
Figure 11.13. Digital Plot of Primary Pollutant Time vs. Concentration Histories
Vertical axis represents time in minutes; horizontal axis depicts pollutant concentrations in
pphm. The species plotted are as follows: 1 - N02, 2 - NO, 3 - 0^, 4 - C^, and 5 -
-------
11.24
terminus of the trajectory, and computes consecutive points along the
trajectory in a direction opposite to that of the wind. At the con-
clusion of the trajectory (the determination of the starting point), an
interpolation routine calculates initial concentrations for nitrogen
dioxide, nitric oxide, ozone and carbon monoxide from the.pollutant
data.
Although its primary use has been for calculating reverse trajec-
tories, the program may also be used for computing forward trajectories.
In this case, the pollutant interpolation routine would yield concen-
tration values which could be considered as validation points.
Theoretically, the program could operate with one meteorological
and one pollutant station. However, the greater the number of stations
utilized, the more realistic and accurate will be the results.
-------
III.l
Chapter III
SIMULATION OF IRRADIATION CHAMBER EXPERIMENTS
A. INTRODUCTION
An important aspect of the evaluation of the model was the explora-
tion of the adequacy of the chemical mechanism in terms of available
information from irradiation chamber experiments. Since rates of
elementary reactions in the gas phase are, to a good first approxi-
mation, determined only by the concentrations of the reacting species
(and the temperature of the system), the chemical mechanism can be
applied to atmospheric simulation with greater confidence if it can be
shown to be consistent with laboratory observations on similar systems.
For this purpose the contracting agency furnished kinetic data,
from twenty-six experiments in which air containing hydrocarbons and
oxides of nitrogen was irradiated and time-profiles of the concentra-
tions of various constituents were determined. These data were to be
used to test the simulation capability of the mechanism under conditions
i
uncomplicated by variations in weather variables and emission rates.
The twenty-six experiments fell into four distinct systems,
differing in their hydrocarbon constituents: six using propylene and
ethane; ten using only toluene; six using toluene and n-butane; and
four using auto exhaust gases containing small proportions of a large
number of different hydrocarbons.
The mechanism incorporated in the atmospheric simulation model
had been designed to test the possibility of simulating propylene
photo-oxidation kinetics accurately, and of simulating other systems
as if their hydrocarbon constituents consisted solely of propylene and
a second, hypothetical, hydrocarbon of less reactivity than propylene.
The simulation of the propylene-ethane system was appropriate for, ap-
-------
III. 2
preaching the first of these objectives, while simulating the auto ex-
haust systems could be expected to furnish information more directly
relevant to the atmospheric simulation problem, especially for the
required validation on Los Angeles data. Simulation of the toluene-
containing systems further tested the flexibility of the approach.
These tests showed that all four systems could be simulated reason-
ably well with only two rate factors reserved as adjustable parameters.
The values selected are shown in Table III-l and discussed in detail
in following sections.
A problem arises in evaluating the success of simulation of the
laboratory data, in that these data are subject to uncertainty due to
random error introduced in preparation, irradiation, sampling and
analysis of the chamber contents. Such random errors may have pertur-
bing effects of two types. First, time-profiles for individual con-
taminants may appear irregular, as random errors affect successive
analytical results to different degrees or in different directions.
Second, time-profiles for the same cons-tituent in different experi-
ments may be subject to systematic distortion due to unrecognized or
undocumented variation in the conditions which may occur in the interim
between the experiments (e.g., change in irradiance of the lamps).
Errors of the first sort may be roughly estimated by examining
the degree of variation between duplicate analyses. Errors of the
second sort could be investigated by systematic study of the agree-
ment between replicate experiments. No assessment of errors of either
type was furnished by the contracting agency; however, some relevant
evidence was contained in the data.
The data for oxides of nitrogen included many sets of duplicate
determinations for total NO , as well as a smaller number of duplicate
sets for N09. NO was determined by subtracting N09 from NO and was
£ fa X
therefore subject, presumably, to somewhat larger resultant errors.
For NO , differences between duplicates were on several occasions as
large as 0.4 ppm, when the total concentration was 2 ppm or less. The
root-mean-square difference between duplicates, for NO in the range
X
-------
III. 3
Table III-l
a
Parameters Used for Simulation of Various Experiments
Hydrocarbon System
Propylene + ethane
Auto exhaust
Toluene
Toluene + n-butane
Exp'
329,
318,
321,
222,
224,
250,
251,
t No.
325
336
334
233
231
etc.
etc.
K31
3.5
3.5
3.5
7.0
3.5
7.0
7.0
Z2
0.0
0.3
0.25
0.0
0.1
0.0
0.0
a Effects of variation of Zl and Z2 are illustrated
in Figures 111-10 to 111-17. All other rate
factors were as listed in Table III-4.
-------
III.4
1 to 1.5 ppm, was 0.19 ppm; this corresponds to 95% confidence limits
at about 0.27 ppm above and below the value from each individual deter-
mination.
Regarding the individual oxides of nitrogen, since nitric oxide
was determined by difference, the 95% confidence interval cannot be
less than that for NO and is presumably larger (ca. 0.3 ppm). The
precision of replicate analyses for N0« appears to have been substan-
tially better, giving 95% limits at 0.1 ppm (or less) on each side of
a given determination.
Duplicate analyses were not reported for other constituents of
the system; therefore, corresponding estimates of error in the analy-
tical process for these constituents are not available. However, a
related error estimate can be derived from the deviations of the obser-
ved values from a smoothed time-profile. A cursory review of devia-
tions of this sort indicated that the 95% confidence limits surrounding
individual determinations of propylene and ethylene are of the order
of magnitude of 5 pphm or less.
Systematic errors are more difficult to detect, but evidence of
such errors is seen in some of the chamber data. For example, Figure
III-l shows the ethane data for experiment 321, plotted against time
of sampling. The points for the first 100 minutes and the last 100
minutes together determine a smooth, self-consistent time-profile for
ethane, but all the intervening points fall below that profile with
deviations of 8 to 12 pphm.
Other evidence pointing to the possibility of systematic errors
is apparent in the comparison of results of separate experiments having
similar initial conditions. Since there was no exact replication of
initial conditions in these experiments, we cannot accurately evaluate
the variance between replicates for this type of error. However, the
comparison between nitrogen dioxide time-profiles for experiments 222
and 233 (see Figure III-2) strongly suggests that this type of error
may be important. In this case an increase of only 20 percent in
-------
EXPERIMENT 321
350
250
230
300
D
D
D
"V
60 120 180 240 300 360
TIME, MINUTES
Figure III. 1. Concentration of ethane related to time of sampling. Experiment 321 (pro-
pylene, ethane, and oxides of nitrogen system).
-------
e
Q.
x
o
UJ
C3
O
DC
200
150
100
50
60
120
180
TIME, MINUTES
240
300
360
Figure III.2.
Time profiles of nitrogen dioxide concentrations. Experiments 222 and 233
(diluted auto exhaust system). Smoothed experimental profiles. Experiment
222: ——; Experiment 233: — —.
-------
III. 7
initial NO appears to cause a fourfold increase in the rate of initial
oxidation of nitric oxide and an increase of more than 75 percent in
the total dosage of nitrogen dioxide over the six-hour irradiation
period. Unless due to systematic error, this indicates a very unexpected
sensitivity to the initial nitrogen dioxide concentration.
B. SIMULATION OF PROPYLENE-ETHANE-OXIDES OF NITROGEN SYSTEM
1. By Original Mechanism, Ml
The propylene-ethane-oxides of nitrogen system was represented
in the experimental data by six irradiation chamber runs. These are
listed in Table III-2, with summary information about their initial
conditions and the findings which resulted.
Ihe first simulation tests utilized mechanism Ml, shown in
Table III-3, which was substantially the mechanism reported in the
final report of Contract CPA 70-151, submitted to EPA in January, 1971.
In these tests, the most difficult part of the task of simulating
observed time profiles of concentrations from chamber runs appeared to
be the prediction of appropriate levels of ozone. Since only two of
the experimental irradiations for this sytem had generated any sub-
stantial ozone levels, the initial simulation tests were directed at
these runs (experiments 329 and 325).
Results achieved using mechanism Ml are illustrated in Figures
III-3 and III-4, showing time profiles for Experiments 329 and 325
respectively. With the best choice of rate factors resulting from
these tests, discrepancies of 20 percent or less (between observed
and calculated values) could be maintained for more than an hour for
ozone and propylene concentrations, and for more than 100 minutes
(the entire length of the simulation) for nitric oxide and nitrogen
-------
III. 8
Table III-2
Irradiation Chamber Experiments in the
Propylene-Ethane-Oxides of Nitrogen System
Exp't
No.
329
325
318
336a
321
324a
Concentrations ,
Initial
N02 NO C3H6
4
2
6
10
11
9
30
34
115
110
130
125
24
45
51
61
28
26
pphm
C2H6
322
269
289
328
350
273
Comparisons
N02
12/7
9/4
78/88
78/62
89/58
89/49
360 minutes
NO C,H,
— 3 o
0/0
0/0
2/1
2/14
12/35
12/34
0/2
0/4
2/8
2/10
8/6
8/5
(est/obs) at
°3
41/42
50/46
43/0
43/3
7/0
7/0
maximum
N02
28/27
30/27
102/89
102/83
89/60
89/54
a. Runs 336 and 324 were not separately simulated; they are compared
with simulations for 318 and 321, respectively.
-------
III. 9
Table III-3
Original Mechanism (Ml) for the Chemical Dynamics
Module (with Rate Factors for Exp. 329 and 325)
1
2
•3,
4
5
6
7
0
9
10
11
12
13
14
15
16
17
18
19
20
i 21
22
! 23
24
25
26
27
28
29
30
31
32
33
Nr»? + HV
QK12 + M
03 + NO
N02*-03
02 S
N03+NO
C2H402+02 •=
. Nflf-HQZ
H+Q2+M
' DUMHC + 02 S
CH300H + HV
NO 2+ OH
OH + 03
OH+r.o
CH3+02+M
CH3Q2+NO
CH30+02+MD
C2H30+02+NO
C2H302+NO
C2H402+NO =
CH30+02
C3H6+Q
C3H6-HB
C3H6+0+02
• C3H6 + 02S
C3H6+H02
C3H6+MF02=C
C2H30+M
CH30+N02
C2H30+02+NO
OUMHC+0
HCHO + HV =
= ,NO + 0
=03 + M
= NO?>02 S
=N03f02
= 02
= 2 NO 2
CH3CHO+03
=N02+OH
=H02+M
= CH300H
= CH30 *• OH
= HN03
=H02 + 02
=H+C02
=CH302+M
= CH30 + NC)2
=CH302+N02
=C2H302+N02
=C2H3U+N02
CH3CHO+N02 .
=HCHO+H02
=CH3+C2H30
=HCHO+C2H402
= HCHO*-C2H4Q2
=CH30 + C2H30
=f. H30 + CH3CHO
H3+MFO+C2H30
= CH3+CO>M
=CH30N02
2=C2H303N02
=CH3+C2H30
H + H + CO
N02 +N03 +H20 = HN03+ "
0.270000000 00
0.140000000 07
0.500000000-01
0. 100000000-0?.
0.100000000 0.*
0.245000000 03
0.100000000 02
O.I 0000 0000 02
0.100000000 07
0.200000000 02
0.. 150000000-01
0.100000000 04
0.100000000 03
0.300000000 01
0.140000000 07
o.iooooooon 03
0.100000000 02
0.100000000 02
0.100000000 03
0.100000000 03
O.I 00000000 03
0.350000000 02
0.500000000-04
0.17500000D 02
0.10000000D 01
0.200000000 01
0.100000000 00
0.100000000 02
0.100000000 01
0.200000000 01
0.700000000 01
0.500000000-02
0. 100000000 02
-------
Ill.10
EXPERIMENT 329
D
A ;NOZ
!NO
O j.
I EXPERIMENTALJSMpOjHED)
— —• —— SIMULATION
100
Figure III.3
Time profiles of contaminant concentrations. Experiment 329 (propylene,
ethane and oxides of nitrogen system). Simulation results using original
i mechanism (Ml), compared with experimental data points and smoothed
j profiles. Experimental points, C; C3H6; A : N02;V: NO;O: '03 Smoothed
profiles: ; simulation results:— — .
-------
III.11
EXPERIMENT 325
pphtn
a c3H6
A NO2
V NO
°3
EXPERIMENTAL (SMOOTHED)
- — — SIMULATION
Figure III.4
Time profiles of contaminant concentrations, Experiment 325 (propylene,
ethane and oxides of nitrogen system). Simulation results using original
mechanism (Ml), compared with experimental data points and smoothed
profiles. Experimental points, D ; 03^; A : NC^; V: NO; O: 03 Smoothed
profiles: —— ; simulation results: — — .
-------
III.12
dioxide. In both runs, simulation of propylene concentration was reason-
ably good for the first hour, but thereafter the simulated propylene
decreased too rapidly and virtually vanished after 100 minutes. Ozone
simulation was within experimental error for the entire 100 minutes for
experiment 329, but the rapid accumulation of ozone observed after about
50 minutes in experiment 325 was not successfully simulated*.
Apparently the underestimation of ozone in experiment 325 was
associated with the overestimation of the propylene consumption rate.
As described below, this sort of difficulty was largely overcome by a
revision of the mechanism. The revision, approved by the project
technical officer, Dr. Sklarew, incorporated several suggestions by
Dr. Dodge of the Division of Physics and Chemistry. j
2. By Revised Mechanism. M2
Satisfactory simulation of experiments 329 and 325 was soon achieved
using the revised mechanism M2, whose most essential new feature was
reaction q£ propylene with hydroxyl radicals. Mechanism M2 is set
forth in Table III-4, together with the appropriate rate parameters as
optimized for experiments 329 and 325.
In mechanism M2, the following revisions were incorporated:
a. All reactions attributed to singlet molecular oxygen were
eliminated.
b. Reaction 7 (consumption of propylene by hydroxyl radicals) was
added.
c. Reactions previously written as termolecular, involving molecular
oxygen and acetyl radicals, were eliminated in favor of accounting for
the adduct, peroxyacetyl radical. Thus reactions 18 and 30 of Ml were
dropped, and new reactions 10, 18, 25 and 30 incorporated in M2.
* In simulating exp. 325, the same parameters were used as for exp. 329,
inasmuch as the objective was to find a single set of parameters
uniformly applicable for this system.
-------
III.13
Table III-4
Revised Mechanism (M2) for the Chemical Dynamics
Module (with Rate Factors for Exp. 329 and 325)
1
2
3
4
5
6
7
3
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
, 28
29
; 30
! 31
i 32
| 33
N02+HV =NO+Q
0+02+M = 03+M
03+NO = N02+02
N02+03 = N03+02
C2H303 + 02= C2H302 + 03
NO 3* NO = 2,M02
•C3H6+OH = CH3CHQ+CH3
NO+H02 = N02+OH
H-t-02+M =H02 + M
C2H30 + 02 = C2H303
N02+OH =HN03 x
OH+03 =H02+02
OH+CO =H+CD2
CH3+02*M =CH302+M '
CH302 + NO =CH30-«-N02
CH30+02+NO =CH302+N02
C2H303+NO = C2H302+N02
C2H302+NO =C2H30+N02
C2H402+NC =CH3CHO+N02
CH 30+02 =HCHO+H02
C3H6+0 =CH3+C2H30
•C3H6+03 =HCHO+C2H402
C3H6+0+02 =HCHCn-C2H402
C2H402+02 = C2H303+OH
C3H6+H02 =CH30+CH3CHO
C3H6+ME02=CH3*-MEO+C2H30
C2H30+N1 =CH3+CO+M
CH30+N02 =CH30N02
C2H303+N02 = C2H.303N02
OUMHC+0 =CH3+C2H30
HCHO + HV = H + H + CO
N02 +N03 +H20 = HN03+ "
0.40000000D 00
0.14000000D 07 ;
0.40000DOOD 00
0.10000000D-04
0.20000000D 02
0.24500000D 03
O.iOOOOOOOD 04
0.100000COD 02
0. 10000000D 07
O.IOOOOOOOD 06
0.0
0.100000COD 03
O.IOOOOOOOD 02
0.300000000 01
0.14000000D 07
0. 20000000D 01 ;
O.IOOOOOOOD 02
0.25000000D 01
0.20300000D 02
O.IOOOOOOOD 03
0.200000000 03
0.35000000D 02
0. 500000000-04
0.17500000D 02 :
0. 30000000D 03
O.IOOOOOOOD 01
0. lOOOGOOOD-01
O.IOOOOOOOD 02
0.10000000D-01 :
" 0.20000000D 00 i
0.35000000D 01 •
0. 300000000-02 •
0. 10000000D 03
-------
III.14
d. To allow for ozone accumulation to exceed initial oxides of
nitrogen input, the new reaction 5,
C2H3°3 + °2 ' C2H3°2
was incorporated.
At the same time that these changes were made, the chemical dynamics
module was also adjusted to allow for use of the chain termination rate
factor, Z, as an input parameter, so that its effect on the simulation
could be more easily explored.
Although the reaction set shown in Table III-4 was not further
altered in the course of the project, there were subsequently two
further adjustments of the chemical dynamics module. We shall designate
the module in the state just described as M2a. The further alterations,
giving modules M2b and M2c, were the following:
M2b was derived from M2a by allowing for radical chain termination
by reaction of radicals with nitric oxide. The rate factor Z thereupon
became Zl, and the new rate factor was made Z2.
M2c was derived from M2b by incorporating a term for chamber dilu-
tion (at a rate of .007 min ) in the differential equation for each of
the non-steady-state species involved in the numerical integration
procedure.
a. Module M2a
Figure III-5 shows the simulation results achieved for Experiment
329 by module M2a, using the rate factors listed in Table III-4 and
with Z taken to be 100 pphm min . The initial concentration of
ethane was used as the input value of dummy hydrocarbon (DUMHC) and
the rate factor for reaction 31 (attack on dummy hydrocarbon by oxygen
atoms) was taken as 3.5 pphm min . Concentrations predicted over
the six-hour period of irradiation were within 0.08 ppm of those indi-
cated by the experiment for the four main reacting species, nitrogen
dioxide, nitric oxide, propylene and ozone.
The largest discrepancies were for nitrogen dioxide, and corres-
pond to a consistent overestimation. A rough correction for chamber
-------
EXPERIMENT 329
50
40
SMOOTHED DATA
SIMULATION 855
NO2 CURVE CORRECTED FOR DILUTION
Q.
a
o
5
0
1
120
180
240
300
360
TIME, minutes
Figure III.5.
Time profiles of contaminant concentrations, Experiment 329 (propylene,
ethane and oxides of nitrogen system). Simulation results using revised
mechanism M2a, compared with smoothed experimental profiles. Smoothed
profiles: ; simulation results:^— -^—.
-------
III.16
dilution* (not incorporated in M2a) indicated, as shown in the figure,
that dilution could approximately account for the major part of these
discrepancies.
Ozone concentrations predicted were within 0.06 ppm of observed
values throughout the irradiation period. The largest discrepancies
arose from the fact that a buildup of 0.06 ppm was estimated for the
first hour of reaction, as compared with only 0.01 ppm observed.
After three hours, the difference between predicted and observed values
became appreciably less.
Nitric oxide concentrations were consistently estimated within
0.04 ppm of the observed (smoothed-curve) values, and propylene within
0.03 ppm. There was, however, a marked difference between the estimated
initial rate of consumption of propylene and the initial rate indicated
by the smoothed data curves, the estimated initial rate appearing too
low by a factor of 4 or 5. After the first twenty minutes of simulation,
the estimated and observed rates were approximately equal.
Figure III-6 shows simulation results for experiment 325 by M2a,
using the same rate factors and other conditions specified in the pre-
ceding paragraphs for experiment 329. In this case the discrepancies
between estimated and observed values of nitrogen dioxide are higher—
up to 0.16 ppm before correcting for dilution, and up to 0.12 ppm
after correction. In other respects, however, the results are quite
similar to those just discussed.
b. Module M2b
Simulations of experiments 329 and 325 by module M2a, as shown in
the preceding discussion, were for the most part within experimental
error of the data furnished. However, when applied to simulation of
other experiments of the propylene-ethane-oxides of nitrogen system,
* This correction was based on a simple reduction of nitrogen dioxide
concentration, proportional to the loss of material from the chamber
assuming steady dilution at a relative rate of 0.007 min
-------
EXPERIMENT 325
E
_c
a
a
<
oc
LU
o
z
o
o
SMOOTHED DATA
SIMULATION 856
CURVE CORRECTED FOR DILUTION
20
10
120
180
TIME, minutes
240
300
360
Figure III.6.
Time profiles of contaminant concentrations. Experiment 325 (propylene,
ethane and oxides of nitrogen system). Simulation results using revised
mechanism M2a, compared with smoothed experimental profiles. Smoothed
profiles: —— ; simulation results:— —
-------
III.18
the same set of rate factors predicted ozone levels which were seriously
in error, indicating premature accumulation of ozone, when in fact
very little oxidant was found within the six-hour irradiation period
in those.experiments. Module M2b allowed us to test the effect of a
hypothetical chain-terminating reaction in which nitric oxide could
react with any chain-carrying radical to remove it from the system.
The rate constant for this reaction, it was found, could not be greater
—1 -1
than 0.5 pphm min without unduly retarding the overall progress of
photo-oxidation in the simulations.
A series of simulations based on experiment 329 showed that the
effects of variation of Zl, the rate factor for radical recombination,
was small relative to the effect of introducing Z2, the rate factor for
chain termination by nitric oxide. Figure III-7 shows nitrogen dioxide
time-profiles as calculated for Z2 = 0.3 pphm min with values of
—1 —1
Zl ranging from 15 to 300 pphm min ; the N0« profile from Figure
III-5 is also included for comparison. The major effect of introducing
Z2 at this value is a drastic reduction in the initial rate of nitrogen
dioxide accumulation due to photo-oxidation of nitric oxide. Both
the time required for half-conversion and the time required to attain
maximum nitrogen dioxide concentration were extended by more than an
hour by this change, while the effect of changing Zl was relatively
slight. Figure III-8 shows the ozone time-profiles for the same set of
simulations from which it is seen that the onset of accumulation of
ozone is substantially delayed by the assumption of Z2 = 0.3.
Figure III-9 shows the application of these parameters (Zl = 100,
Z2 = 0.3) to the simulation of experiment 318. Here, the initial con-
centration of nitric oxide was much larger than in the experiments
previously discussed (329 and 325), and larger discrepancies between
estimated and observed time-profiles emerge as a consequence. Estimated
concentrations of nitrogen dioxide, nitric oxide, and propylene were
within 0.25 ppm of the experimental smoothed-profile values throughout
the six-hour irradiation. However, the simulation predicted substantial
accumulation of ozone after about three hours, an effect clearly absent
-------
EXPERIMENT 329 SIMULATIONS
30
Q.
a
uf
Q
X
g
a
z
LLI
O
O
tr
20
10
I
Z2 = 0
^••^ —™
Z1 = 100
Z2 = 0.3
- /
Z1 = 300
100
30
15
60
120
180
240
300
TIME, MINUTES
Figure. IT 1. 7 . Time profiles of nitrogen dioxide concentration simulating Experiment 329
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2a and M2b, with various values of chain-terminating para-
meters Z1 and Z2.
-------
s
o
EXPERIMENT 329
OXIDANT
60
120
180
240
300
Figure III .8. Time profiles of oxidant (ozone) concentration simulating Experiment 329
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2a and M2b, with various values of chain-terminating para-
meters Z1 and Z2.
-------
EXPERIMENT 318
120
SMOOTHED DATA
SIMULATION 8129
360
Figure III.9.
i Time profiles of contaminant concentrations. Experiment 318 (propylene,
ethane and oxides of nitrogen system). Simulation results using revised
mechanism M2b, compared with smoothed experimental profiles. Smoothed
profiles:——.; simulation results: _ — .
-------
III.22
in the experiment. Furthermore, 'the predicted rates of change of
concentrations of the other contaminants at the start of the irradiation
were much smaller than indicated by the chamber data.
Any further suppression of ozone (in the simulations) by further
increasing Z2 would evidently lead to greater discrepancies in these
initial rates of change. Therefore, it appears that, in order to
account for the lack of ozone in experiment 318, more extensive modi-
fications in M2 should be explored.
Further detail regarding the effect of varying Z2 on the simula-
tion of nitrogen dioxide profiles for experiment 318 is shown in Figure
111-10. Shown are estimated profiles for the series of values Z2 = 0,
0.05, 0.15, 0.3, 0.4, 0.5. The maximum concentration of nitrogen
dioxide is not much affected by these variations of Z2, even though
the time of reaching the peak is delayed from about 75 minutes for the
first to around 300 minutes for the last.
Figure III-ll shows the corresponding series of ozone time-profiles.
It is interesting to note that these curves all rise steeply, and with
approximately the same slope, after an initial delay which is longer,
the larger the value of Z2. This mechanism evidently implies an inhi-
bitory effect of initial nitric oxide, in at least qualitative agree-
ment with many experimental observations in hydrocarbon-oxides of nitro-
gen photo-oxidation systems.
c.
Module M2c
The creation of module. M2c made available a convenient means of
estimating the total effect of dilution on the progress of photochem-
istry in the chamber. That this effect is, as expected, small, is
shown by the curves in Figures 111-12 through 111-15, which present time-
profiles for various contaminants in experiment 318, corresponding to
various assumptions as to Z2, dilution, the rate factor K31, and the
relation of ethane to dummy hydrocarbon concentration.
In each of these figures, curve 1 corresponds to the simulation
shown in Figure III-9 and discussed above. In this run Z2 was taken
-------
III.23
SIMULATION, EXPERIMENT 318 NITROGEN DIOXIDE
Q
X
g
Q
LU
C3
O
tc.
100
40
20
120 180
TIME, minutes
240
300
Figure III. 10. Time profiles of nitrogen dioxide concentration simulating Experiment 318
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2a and M2b, with various values of chain-terminating para-
meter 22.
-------
III.24
SIMULATION, EXPERIMENT 318
s
O
TIME, minutes
Figure III. 11. Time profiles of oxidant (ozone) concentration simulating Experiment 318
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2a and M2b, with various values of chain-terminating para-
meter Z2.
-------
SIMULATION, EXPERIMENT 318 (AND 336J NITROGEN DIOXIDE
100
a
a
o
x
g
o
2
LU
O
O
CC
#1: Z2 = 0.3; no dilution; DUMHC = 289; K31 = 3.5
#2: 22 = 0.3; dilution; DUMHC = 51, K31 = 7
#3: 2.2 = 0.25; dilution; DUMHC = 51, K31 = 7 (NO2(i) = 12)
#4: 22 = 0.1; dilution; DUMHC = 51, K31 = 7
20
Ui
120
180
240
300
360
Figure III.12.
TIME, MINUTES
Time profiles of nitrogen dioxide concentration simulating Experiment 318
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2b and M2c, with various values of chain-terminating para-
meters and parameters for unreactive hydrocarbon.
-------
SIMULATION, EXPERIMENT 31.8 OXIDANT
Q.
Q.
o
NJ
O
#1: Z2 = 0.3; no dilution; DUMHC = 289; K31 = 3.5
#2: Z2 = 0.3; dilution; DUMHC = 51, K31 = 7
#3: Z2 = 0.25; dilution; DUMHC = 51, K31 = 7 (NC>2(i) = 12)
#4: Z2 = 0.1; dilution; DUMHC = 51, K31 = 7
120
180
240
300
TIME. MINUTES
Figure III.13.
Time profiles for oxidant (ozone) concentration simulating Experiment 318
(propylene, ethane and oxides of nitrogen system). Results using revised
mechanism M2b and M2c, with various values of chain-terminating para-
meters and parameters for unreactive hydrocarbon.
-------
SIMULATION, EXPERIMENT 318 (AND 336) NITRIC OXIDE
120
Q.
Q.
Ill"
O
X
O
O
cc
I-
100
#1 Z2 = 0.3 NO DILUTION
DUMHC = 289
K31 = 3.5
#4 Z2 = 0.1 WITH DILUTION
DUMHC = 51
K31 = 7
#2 22 = 0.3
#3 22 = 0.25 NO2(i} = 12
EXPERIMENTAL
120
180
TIME, MINUTES
240
300
360
M
M
M
CO
Figure III. 14. Time profiles of nitric oxide concentration simulating Experiment 318 (pro-
pylene, ethane and oxides of nitrogen system). Results using revised mech-
anism M2b and M2c, with various values of chain-terminating parameters
and parameters for unreactive hydrocarbon.
-------
SIMULATION. EXPERIMENT 318 (AND 336)
PROPYLENE
50
40
I
LU
_l
a.
O
EC
a.
30
20
10
\\ X.
M
M
M
•
S3
00
#1: 22 = 0.3; no dilution; OUMHC - 289; K31 = 3.5
#2: Z2 = 0.3; dilution; DUMHC = 51, K31 = 7
#3: 22 = 0.25; dilution; DUMHC = 51, K31 = 7 (N02
-------
III.29
as 0.3 pphm mln ; K31 was 3.5; each mole of ethane was assumed
equivalent to one mole of dummy hydrocarbon; and dilution was ignored.
Curve 2 was from a simulation with Z2 = 0.3; K31 = 7; one mole
ethane = 0.18 mole DUMHC; and dilution assumed.
Curve 3 was from a simulation having Z2 = 0.25, K31 =• 7; one ethane
= 0.18 DUMHC; dilution assumed; and initial NO taken as 12 pphm (twice
the amount specified in the experimental data) .
Curve 4 was for Z2 = 0.1; K31 = 7; one ethane = 0.18 DUMHC; and
dilution assumed.
Zl was taken at 100 pphm min throughout.
-Comparison of the curves with the experimental'smoothed profiles
also shown in the figures indicates that the conditions which permit
optimum simulation are not the same for all contaminants. Thus, the
nitrogen dioxide and nitric oxide data are perhaps best simulated by
curve 3; the ozone data, by curve 4, and the propylene data, by curve 2.
It is clear that, unless further modifications to the mechanism are
explored and substantiated, any choices for the values of parameters
to be used in simulation of the propylene-ethane-oxides of nitrogen
system will have to be in the nature of a reasonable compromise, in
view of the requirements of the particular case.
Although the set of rate factors used for the simulation of
experiment 318 (shown in Figure III-9) was inadequate to predict sup-
pression of ozone in that experiment, it was somewhat better relative
to experiment 321, as shown in Figure 111-16. In this case the predic-
ted ozone after six hours was less than 0.1 ppm. In spite of the use of
module M2c, which includes the adjustment for chamber dilution, the
concentration of nitrogen dioxide at the end of the experiment was
overestimated by about 0.3 ppm, and the nitric oxide underestimated by
about 0.2 ppm. For the first three hours, however, there were no dis-
crepancies above 0.05 ppm; and for propylene, the discrepancies were
less than 0.05 ppm throughout.
-------
SIMULATION. EXPERIMENT 321 (AND 324)
140
Figure III. 16. Time profiles of contaminant concentrations. Experiment 321 (propylene,
ethane and oxides of nitrogen system). Simulation results using revised mech-
anism M2c, compared with smoothed experimental profiles. Smoothed pro-
files:——, simulation results:—.— .
-------
III.31
C. SIMULATION OF AUTO EXHAUST PHOTO-OXIDATION SYSTEMS
Diluted auto exhaust gases were irradiated in four chamber experi-
ments, under the conditions indicated in Table III-5.
Figures 111-17, 111-18, 111-19 and 111-20 show the results of simu-
lation of the auto exhaust experiments, using module M2b. The rate
factors used in these simulations were the same as those shown in
Table III-4, with the exception that in the first two, K31 was increased
by a factor of two, to 1 pphm min .
For convenience in discussion, we consider these chamber tests in
two pairs: experiments 222 and 233, in which high levels of ozone
were produced, and experiments 224 and 231, in which ozone was essen-
tially absent.
Since the intention in the atmospheric model was to simulate auto
exhaust as a mixture of propylene and an arbitrary less reactive hydro-
carbon, it was desirable to adjust the relative amounts of these hypo-
thetical surrogate species so as to optimize the simulation of NO, NCL,
and 0. without attempting to simulate the hydrocarbon concentration
profiles. Degrees of freedom for the simulation were, therefore, the
initial concentrations of propylene and of the dummy hydrocarbon, and
the rate factor for reaction 31 (attack on dummy hydrocarbon by oxygen
atoms). We chose to set the initial concentration of propylene equal
to that found for propylene in the experiment in each case, and to let
the dummy hydrocarbon be higher by an order of magnitude, to simulate
the remainder of the hydrocarbon portion of the exhaust gases. The
parameter Z2, for rate of chain termination by nitric oxide, was also
allowed to vary.
The effect of variation of Z2 is seen in Figure 111-17, where
experiment 222 is simulated by three sets of curves, with successive
values 0, 0.05 and 0.1 for Z2. The main effect is to produce a delay
in the early progress of photo-oxidation, so that the simulated time
for oxidation of half the nitric oxide is increased from about 20 to
about 40 minutes, and the production of N09 and ozone are correspondingly
-------
III.32
Table II1-5
Irradiation Chamber Experiments with
Automobile Exhaust Gases
Exp't
No.
222
233
224
.231
Concentrations,
pphm
Initial
NO,
10
22
20
20
NO
195
220 .
320
285
C3H6
21
25
12
10
DUMHC3
480
600
200
150
Comparisons (estimated/observed) at
360 Minutes
N02
59/4
68/32
210/140
180/95
NO
0/0
0/0
22/20
45/65
C3H6
0/0
0/0
3/2
3/2
°3
85/91
85/102
7/4
5/3
Maximum
NO,
165/140
200/185
205/140
180/105
Concentrations of DUMHC were taken from the total hydrocarbon concen-
trations, assuming a carbon number of 2.9. The^rate factor for reaction
of DUMHC with oxygen atoms was taken as 7 pphm min'1 for Experiments
222 and 233, 3.5 pphm"1 tain"1 for Experiments 224 and 231.
-------
EXPERIMENT 222
250
t
a
O
O
O
200
150
SMOOTHED DATA
SIMULATIONS 823-5
100
360
Figure III.17.
Time profiles of contaminant concentrations. Experiment 222 (diluted auto
exhaust system). Simulation results using revised mechanism M2b, with
various values of chain-terminating parameters Z2, compared with smoothed
experimental profiles. Smoothed profiles: «_»•; simulation results— _ .
-------
EXPERIMENT 233
300 r
250
200
a
a
O
O
z
o
O
150 —
SMOOTHED DATA
SIMULATION 8227
NO, CORRECTED FOR DILUTION
100 —
180
TIME, minutes
240
300
u>
360
Figure III. 18. Time profiles of contaminant concentrations, Experiment 233 (diluted auto
exhaust system). Simulation results using revised mechanism M2b, compared
with smoothed experimental profiles. Smoothed profiles: ; simulation
results: — — .
-------
EXPERIMENT 231
350
300
250
200
LU
o
O 150
o
100
50
SMOOTHED DATA
SIMULATION 8272
CORRECTED FOR DILUTION
NO,
u>
Ul
60
120
180
240
300
360
Figure III.19. Time profiles of contaminant concentrations, Experiment 231 (diluted auto
exhaust system). Simulation results using revised mechanism M2b, compared
with smoothed experimental profiles. Smoothed profiles: ; simulation
results:——.
-------
EXPERIMENT 224
300
SMOOTHED DATA
SIMULATION 8211
_CORRECTED FOR DILUTION
M
360
Figure III.20.: Time profiles of contaminant concentrations, Experiment 224 (diluted auto
exhaust system). Simulation results using revised mechanism M2b, compared
with smoothed experimental profiles. Smoothed profiles: ; simulation
results: — —.
-------
III.37
delayed. Apparently, for the most accurate simulation of the first half
of experiment 222, a value of about 0.2 for Z2 would be appropriate;
however, this would result in increased over-estimation of nitrogen
dioxide and under-estimation of ozone at the six-hour mark.
Figure 111-18 shows the simulation of experiment 233, using Z2 =
0. As in 222, the simulation predicts faster progress than is observed
in the early part of the irradiation, but slower increase of ozone
and decrease of N0_ than is observed in the latter half. The ozone
predicted, however, is within 0.25 ppm of the observed value over the
last 200 minutes of the test.
Simulations for the auto exhaust experiments in which ozone was
' negligible are'shown in Figures 111-19 and 111-20. For optimum results
under these circumstances, the rate factor for the consumption of
dummy hydrocarbons was reduced to half its previous value, and Z2 was
taken at 0.1. Figures 111-19 and 111-20 show, beside the computed
simulation curves, other curves for NO and N0_ corrected by hand cal-
culation to account for dilution at the rate reported for the chamber.
Both corrected and uncorrected curves give good approximations to the
experimental results in the first part of the test, but the estimated
nitrogen dioxide after six hours is higher than that observed.
D. SIMULATION OF TOLUENE-CONTAINING SYSTEMS
The chemical dynamics module of REM is designed for application
to photo-oxidation in mixtures containing olefins as the principal
reactive hydrocarbon component. However, the contract required us to
use it to simulate systems which contained no olefins (the only hydro-
carbons present being toluene or toluene and n-butane).
For this purpose we postulated the (chemically absurd) assumption
that toluene consists of a mixture of propylene and dummy hydrocarbons,
such that the total concentration of these species is the same as the
specified total concentration of hydrocarbons in the system to be
simulated.
-------
III.38
1. The Toluene-Oxides of Nitrogen System
Table III-6 presents salient features of the results of simulations
for the toluene-oxides of nitrogen system, compared with the experi-
mental observations. Listed are initial concentrations for each experi-
ment, postulated initial values for the simulation, and both predicted
and observed values for N0~, NO and 0, at 360 minutes, as well as pre-
dicted and observed values of the maximum concentration of N0?.
(Experiments 309, 272 and 305 were not separately simulated because
their initial conditions were nearly the same as those in experiments
which were simulated, namely, experiments 250, 263 and 300, respectively)
An initial run done to simulate Experiment 250 was. based on an
assumed equivalence (for reaction rate purposes) of one mole of pro-
pylene to five of toluene. Results showed that the simulated course of
reaction was substantially too rapid in this case; maximum NO. was
reached in less than 200 minutes, compared to 300 in the experiment;
and although N0_ at 360 minutes was very close to that measured, the
simulated oxidant was twelve times too large and the simulated nitric
oxide was six times too small.
A second simulation run was therefore made, assuming one mole of
propylene equivalent to twenty-five of toluene. This gave results for
ozone and both oxides of nitrogen which were un^ormly within 1/4 ppm
^^
of the experimental curves (smoothed from EPA data). This assumption
was therefore adopted as standard for toluene photo-oxidation, and the
remaining toluene experiments were simulated on this basis. Figure III-
21 shows the effect of this assumption on nitrogen oxides in Experiment
250.
Rate factors used in these simulations were the same as those used
for the auto exhaust experiments in which ozone was produced, experi-
ments 222 and 231; that is, they were the same as those used in all the
simulations with the exception of K31, which was taken as 7 pphm
min in these experiments.
-------
Ill.39
Table III-6
Irradiation Chamber Experiments in the
Toluene-Oxides of Nitrogen System
Experiment
Number
250 (a)*
(b)
309
263
272
259
271
258
277
300
305
Concentrations, pphm
NO
2
7
5
4
2
4
4
3
3
8
Initial
N0_ C
123
131
55
55
33
31
32
30
128
124
7*8
153
166
171
150
164
120
288
320
322
314
Postulated
Initial
C3H6
30
6
6.8
6.7
4.8
12.0
14.4
14.4
DUMHC
123
147
164
157
115
276
306
308
Resulting
Conc'ns (sim/obs)
360 Minutes
NO NO
L ~~~
69/70 2/13
77/70 12/13
77/72 12/0
35/2 2/4
35/3 2/2
19/4 1/3
20/4 2/3
16/4 1/2
14/4 1/3
19/4 1/3
19/3 1/5
36/3
7/3 ,
7/10
17/31
17/31
15/26
10/18
36/24
33/18
15/26
15/39
Maximum
N02
92/70
77/70
77/91
41/21
41/39
24/17
22/21
27/26
25/21
98/89
98/91
* Run 250 was simulated twice (see text). Simulation (b) is based on a
propylene equivalence of 1 mole to 25 moles of toluene, as are all the
other simulations tabulated, except Simulation (a), which was based on
1 to 5.
** Runs 309, 272 and 305 were not separately simulated; they are compared
with simulations for 250, 263 and 300, respectively.
-------
EXPERIMENT 250
5
_c
a
a.
<
-------
III.41
2. The Toluene-Butane-Oxides of Nitrogen System
Since the experimental data indicated that a mixture of toluene
with n-butane was more reactive in the photo-oxidation system than was
toluene alone, our chemical dynamics model required a representation of
the additional component, n-butane°, in terms of the surrogate hydro-
carbons, propylene and DUMHC. For this purpose we postulated the new,
chemically absurd assumption that n-butane consists of a mixture of
propylene and dummy hydrocarbons, such that the total concentration of
these species is the same as the specific total concentration of hydro-
carbons in the system under study.
Since n-butane is known to be substantially less reactive than
toluene, we assumed that one mole of propylene would be equivalent to
a hundred moles of n-butane.
Table III-7 shows the comparison of the simulation results, cal-
culated on this basis, against the experimental observations at 360
»
minutes. (Parameters of the simulation were the same as for toluene
alone). Experiments 270 and 280 were not separately simulated, inasmuch
as their initial concentrations closely resembled Experiments 253 and
257, respectively.
3. Discussion of Results with Toluene-Containing Systems
The accuracy of simulation, represented by Tables III-6 and III-7,
in the face of the rather elementary stage of development of the reac-
tion mechanism used in our chemical dynamics module, is extraordinary.
Predicted ozone values were within a factor of two of those observed
(at 360 minutes) in fourteen cases out of sixteen. Predicted nitric
oxide values were within 2 parts per hundred million of the observed
in thirteen cases out of sixteen.
Although concentrations of nitrogen dioxide at 360 minutes were
usually overestimated in the simulation, the maximum concentration of
nitrogen dioxide during the run as predicted by the simulation was
within a factor of two of that observed in every case, and was within
a factor of 1.5 in fifteen cases out of sixteen.
-------
III.42
Table III-7
Irradiation Chamber Experiments in the
Toluene-Butane-Oxides of Nitrogen System
Exp't
No.
251
260
253
270
257
280
Con cent rat ions , pphm
NO
11
10
11
3
7
3
Initial
NO C7Hg
111 160
110 142
50 143
57 163
27 129
30 165
C4H10
302
273
342
278
297
305
Postulated
Initial
C3H6
9.4
8.4
9.2
8.1
DUMHC
453
407
454'
439
Resulting Conc'ns (sim/obs)*
360 Minutes
N02
48/36
46/42
24/5
24/2
14/13
14/4
NO
1/3
1/3
1/2
1/3
1/1
1/1
°3
51/44
49/32
46/55
46/62
34/52
34/40
Maximum
N02
94/89
84/66
47/54
47/43
25/28
25/24
* Runs 270 and 280 were not separately simulated; they are compared with
simulations for 253 and 257, respectively.
-------
III. 43
It is interesting to note that these results, indicating a sub-
stantial predictive capability especially in the qualitative sense,
were achieved with only a single set of rate factors and with uniform
assumptions as to the reactivity of the hydrocarbons involved, and
that the rate factors were substantially the same set that also served
for simulation of the auto exhaust experiments and the propylene-ethane
system. This degree of success, obtained in spite of strictures for-
bidding the application of advanced understanding of the chemical
systems involved, must augur well for the improvement to be expected
when such strictures are eventually lifted.
E. SIMULATION OF PHOTO-OXIDATION PRODUCTS IN THE IRRADIATION EXPERIMENTS
Data for the products of photo-oxidation were rather fragmentary,
being confined to occasional determinations on formaldehyde, acetalde-
hyde, peroxyacetyl nitrate, and methyl nitrate. Observed concentrations
of methyl nitrate were never larger than 0.01 ppm, so this product will
not be further discussed. Table III-8 compares estimated with observed
concentrations for the other compounds mentioned.
Observed concentrations of peroxyacetyl nitrate (PAN) were in
some cases of the order of magnitude of 0.2 ppm, especially in experi-
ments 325 and 329 of the propylene-ethane system, and 222 and 233 of
the auto exhaust series. It may be noted that the higher concentrations
of PAN seem to be well correlated with high levels of oxidant, as
expected on chemical grounds. Simulated values of PAN ranged from 0.02
to 0.56 ppm, and were more often overestimated than underestimated.
Observed concentrations of the aldehydes ranged up to about 0.50
ppm for each. Simulated values ranged up to about 0.2 for formaldehyde
and 0.4 for acetaldehyde. For both of these compounds, the estimates
seemed predominantly lower than the observed values.
F. SUMMARY AND CONCLUSIONS
Tables III-2, III-5, III-6 and III-7 present the results of the simu-
lations of the irradiation chamber experiments in terms of the comparison
-------
III.44
Table III-8
Comparison of Estimated vs. Observed Concentrations
for Formaldehyde, Acetaldehyde, and Peroxyacetyl
Nitrate (after 6 hours irradiation)
Experiment
Concentrations, pphm (estimated/observed)
P ropy lene-e thane
Auto exhaust
Toluene-butane
329
325
318
336*
321
324*
222
233
224
231
251
260
253
270*
257
280*
Formaldehyde
10/40
IB/—
19/20
19/33
6/17
6/20
15/42
17/50
6/39
4/20
Ti-
ll—
5/11
51—
3/—
3/14
Acetaldehyde
22/1
41/55
30/27
30/31
10/13
10/46
17/22
20/18
6/16
4/11
6/34
6/23
7/32
7/24
6/25
6/22
P.A.N.
23/11
3/17
9/0
9/1
2/0
21—
36/14
56/19
12 /—
6/0
14'/10
12/4
7/8
7/4
3/7
3/2
* Estimates for experiments 336, 324, 270, and 280 are from simulations
for experiments 318, 321, 253 and 257, respectively.
-------
III.45
of concentrations estimated with concentrations observed, using the best
simulation achieved in each system. Table III-l shows the extent of
modification to the parameter set which was required to achieve this
degree of simulation.
A review of these results points to the following conclusions:
1. Under most of the conditions represented by these experiments,
the maximum levels of nitrogen- dioxide and of ozone can be predicted
within experimental error using the mechanism given here (M2), with
correction for rate of dilution in the chamber (module M2c).
2. The principal deficiencies in this mechanism are indicated by
serious overestimation, in certain circumstances, of the ozone concen-
tration after six hours of irradiation, and simultaneous underestimation
of the time required for nitrogen dioxide to reach a maximum and decline
from it. This is seen especially when initial nitric oxide levels are
very high.
3. The mechanism should be susceptible of marked improvement with
relatively minor changes, by systematic exploration of the effects of
plausible additional reactions.
-------
IV.1
Chapter IV
SIMULATION OF PHOTOCHEMICAL AIR POLLUTION EPISODES
A. INTRODUCTION
The most direct and demonstrative testing of the atmospheric simu-
lation model, under the current contract, involved the use of weather
data and contaminant data from six reasonably typical episodes of
photochemical air pollution as observed in Los Angeles in the summer
and fall of 1969. Data furnished by the Environmental Protection
Agency for this purpose included hourly average values of wind direc-
tion, wind speed, temperature and relative .humidity at each of twenty-
five meteorological stations in the area of the Los Angeles Basin.
Radiosonde measurements of the vertical temperature profile at the Los
Angeles International Airport for each of the days chosen were obtained
from the Los Angeles County Air Pollution Control District.
These data, together with information regarding the location and
elevation of the various stations, were sufficient to permit the cal-
culation of air parcel trajectories with associated temperature,
humidity, and mixing depth, according to the principles and methods
described in Chapter II. Since the objective of the evaluation was to
simulate air quality at places where measurements existed, the locations
of various air monitoring stations were taken as terminal points for
trajectories and their starting points (as well as a series of points
of the intervening locus) were estimated with the reverse trajectory
routine (cf. Ch. II).
Initial values of the concentrations of contaminants in air par-
cels, at the starting points thus determined, were estimated by inter-
polation from data for the Los Angeles air monitoring network, also
furnished by EPA. These data consisted of hourly average values for
-------
IV. 2
various contaminants, including carbon monoxide, oxides of nitrogen,
oxidants, and sulfur dioxide, for a network of ten to twelve stations.
For estimating the concentrations of contaminants expected in the
air parcels upon arrival at their specified terminal points, the amounts
acquired while en route are needed. They are approximated by inter-
polation of emission rates from reference values, specified as a
function of location by the emission inventory grid maps, and modified
as a function of time of day in accord with an assumed curve of diurnal
variation.
The atmospheric simulation model is especially designed for the
estimation of chemically reactive contaminants and secondary contami-
nants which are products of chemical reactions. For such contaminants,
ordinary air quality models (based on the assumption of Gaussian
dispersion) are not adequate. The PES model, however, is equally
applicable to reactive and non-reactive contaminants. Carbon monoxide
concentrations seem to be little affected by atmospheric chemical
reaction on the time scale involved in photochemical smog. It is use-
ful, therefore, to evaluate the performance of the model in estimating
carbon monoxide concentrations, so that the complications which can be
introduced by deficiencies in the chemical reaction system can be
avoided.
The discussion below, therefore, is addressed first to the simu-
lation of carbon monoxide levels, after which the results for reactive
contaminants and reaction products are evaluated. Pertinent data for
each trajectory run are tabulated in Tables A-l and A-2, Appen-
dix A, and corresponding trajectories are shown on the maps in Figu-
res A-l through A-6.
B. CARBON MONOXIDE SIMULATION
To facilitate a broader evaluation of the reliability of the
atmospheric simulation with respect to non-reactive contaminants, the
-------
IV.3
numerical integration routine (DIFSUB) was applied for carbon monoxide
as the sole contaminant, and the kinetics routine (DIFFUN) was modified
with an option to bypass the chemistry.
The model was then run for six days specified by EPA: September
11, 29 and 30, October 29 and '30, and November 4. For each of four
receptor locations (corresponding to the air monitoring stations for
Downtown Los Angeles, Whittier, Azusa and Burbank), three six-hour
trajectories were computed for each day; these trajectories were ar-
ranged to arrive at the stations at hours 0930, 1130 and 1330. Initial
time for these trajectories was ordinarily 0700. Thus, travel times
for the simulated air. parcels, before reaching the specified receptor
locations were, respectively, 2.5, 4.5 and 6.5 hours.
Of the 72 trajectories thus planned, 67 were successfully completed.
This included all trajectories to the Los Angeles and Whittier receptor
locations, sixteen to Azusa, and thirteen to Burbank. Two of the
scheduled trajectories to Azusa for November 4 could not be used because
their points of origin were outside the limits of the simulated terri-
tory. Two trajectories to Burbank on September 11 and all three on
November 4 failed in computation, apparently as a result of blanks in
the data base. A set of three trajectories to the Pasadena receptor
location for November 4 was computed and substituted for the Azusa
trajectories of that date for purposes of evaluation, since Pasadena
is the most northeasterly station other'than Azusa.
A graphic comparison of the results of these calculations against
hourly average values obtained from the air monitoring records is
shown in Figures IV-1 to IV-6. For the period 0800 to 1400 PST each
day, six hourly average values of carbon monoxide concentration were
on record for each station. Estimated values for these same receptor
points were available for hours 0930, 1130, and 1330; these are com-
pared with the averages for hours beginning at 0900, 1100 and 1300
respectively. For hours beginning at 0800, 1000 and 1200, calculated
-------
IV.4
1 1
DOWNTOWN LOS ANGELES
9/11/69
a MEASURED CO
PES PREDICTED CO
a n a
1 1 1 1 1
10 11
a
WHITTIER
9/11/69
MEASURED CO
PES PREDICTED CO
1 1
8 9 10 L- 11 12 13
L-
I I I I I I
AZUSA
9/1 1/69
MEASURED CO
PES PREDICTED CO
ni a ® • a a
1
10 11 12 13
1 1 1
8URBANK
9/11/69
MEASURED CO
PES PREDICTED CO
a a a
10 _| 11
Figure IV.1. Comparison of calculated vs. observed values
of carbon monoxide at air monitoring stations
on September 11, 1969.
-------
IV. 5
DOWNTOWN LOS ANGELES
9/29/69
III MEASURED CO
— PES PREDICTED CO
10 _ 11 12 13
d
WHITTIER
9/29/69
MEASURED CO
PES PREDICTED CO
b "
AZUSA
9/29/69
MEASURED CO
PES PREDICTED CO
I I l I
10 _ 11 12 13
BURBANK
9/29/69
MEASURED CO
PES PREDICTED CO
9 10 /-I 11 12 13
Figure IV.2. Comparison of Calculated vs. Observed Values
of Carbon Monoxide at Air Monitoring Stations
on September 29, 1969.
-------
IV. 6
DOWNTOWN LOS ANGELES
MEASURED CO
PES PREDICTED CO
10 _ 11 12 13
a
WHITTIER
9/30/69
ffl MEASURED CO
— PES PREDICTED CO
9 10 I 11 12 13
AZUSA
9/30/69
l«l MEASURED CO
— PES PREDICTED CO
10 _ 11
I I
BURBANK
9/30/69
MEASURED CO
PES PREDICTED CO
J L
10 -J 11 12 13
Figure IV.3. Comparison of Calculated vs. Observed Values
of Carbon Monoxide at Air Monitoring Stations
on September 30, 1969.
-------
IV. 7
— IB
DOWNTOWN LOS ANGELES
10/29/69
51 MEASURED CO
— PES PREDICTED CO
1 I
— IS
WHITTIER
10/29/69
© MEASURED CO
•—• PES PREDICTED CO
I I I
8 9 10 11 12 13
8 9 10 L- 11 12 13
I I I
AZUSA
10/29/69
® MEASURED CO
~— PES PREDICTED Cp
10 _ 11 12 13
BURBANK
10/29/69
SI MEASURED CO
— PES PREDICTED CO
10 I 11 12 13
Figure IV,4. Comparison of Calculated vs. Observed Values
of Carbon Monoxide at Air Monitoring Stations
on October 29, 1969.
-------
IV. 8
DOWNTOWN LOS ANGELES
10/30/69
B MEASURED CO
— PES PREDICTED CO
WHITTIER
10/30/69
MEASURED CO
PES PREDICTED CO
I II I I
9 10 -\ 11 12 13
10 |_ 11 12 13
i 1 1 1 i r
AZUSA
10/30/69
IS MEASURED CO
— PES PREDICTED CO
III! I
10 11
d "
Figure IV.5. Comparison of Calculated vs. Observed
Values of Carbon Monoxide at Air
Monitoring Stations on October 30, 1969.
-------
IV.9
t 1 T
DOWNTOWN LOS ANGELES
11/4/69
B MEASURED CO
— PES PREDICTED CO
I I I I I I
10 _ 11
d
—m
PASADENA
11/4/69
ffl MEASURED CO
— PES PREDICTED CO
10 11 12 13
c
Figure IV.6. Comparison of Calculated vs. Observed
Values of Carbon Monoxide at Air
Monitoring Stations on November 4, 1969,
-------
IV. 10
CO values from the trajectories did not correspond exactly to the
receptor locations, but they are included in the comparisons for all
cases when the trajectory locus at mid-hour was within a five mile
radius of the designated receptor point. Thus the broken lines shown
in the figures furnish an approximate representation of the estimated
air quality at the various stations during the hours 0800 to 1400 PST.
To compare the simulated and observed concentrations on an overall
basis: the mean of all calculated values is 8.6 ppm, while the mean of
observed values is 9.2 ppm. Thus, on the whole, the estimates appear
low by about 7 percent, corresponding to an average undercalculation
of 0.6 ppm. The root mean square deviation (uncorrected for the dif-
ference of means) is 4.5 ppm, or about half the mean value.
To compare the calculated values with the hourly averages on a
relative basis, the number of estimates which proved to be more than
twice the observed value of the hourly average, or less than half that
value, were counted. Over the entire set of 135 comparisons shown in
the Figures, the estimates fell outside these limits in 19 cases, or
approximately 15 percent.
Examination of the frequency distribution of individual deviations
(by absolute value) showed that more than half are 2 ppm or less;
about two thirds are 3 ppm or less, and the 95 percent confidence
limits are at about 7 ppm above and below the observed hourly average.
In three cases of the 135, deviations larger than 9 ppm were found; the
rate of incidence of such deviations appears, therefore, to be less
than one in forty.
It is interesting to note, however, that the frequency distribu-
tion of individual deviations is not independent of the time of day.
Significantly fewer deviations greater than 3 ppm occur on the compari-
sons for the last three hours of the observation period (i.e., 1100 to
1400 PST) than during the first 3 hours. This is in accord with expec-
-------
IV. 11
tations (assuming that the probable error of the simulation estimates
is generally related to the magnitude of the concentration being simu-
lated) , since carbon monoxide levels usually decline as the hours pass
after 0900.
A separate calculation using only"the estimates for 0930, 1130
and 1330, when the trajectory locations-were very near the receptor
locations, showed that the average undercalculation for these cases
was less than 0.1 ppm. The root mean square deviation however, was
reduced only slightly, to 4.4 ppm; therefore, the complete set of
comparisons was retained for further evaluation and review.
From Figures IV-1 ff. it is evident that there were substantial
differences in the levels of carbon monoxide at the various monitoring
stations. In fact, the overall average of measured values for the six
days ranged from 6.3 ppm to.12.3 ppm, as shown in Table IV-1, while the
average of estimated values ranged from 5.8 to 10.5 ppm. The average
estimated for the Burbank location was low by 2.1 ppm, (23%), while
that for Whittier was high by 1.6 ppm (22%). The fraction outside limits
of one half to twice the hourly average ranged from 4 of 36 (11%).for
Los Angeles and Whittier to 5 of 27 (19%) for Burbank.
The root mean square deviation ranged from a low of 2.9 ppm
(Whittier) to a high of 6.0 ppm (Burbank). It appeared to be more or
less parallel to the mean observed concentrations, leading to a rela-
tively uniform coefficient of variation, ranging from about.0.40 to 0.53.
Since it would be expected that the reliability of simulation might
be different for various days, the data were examined also from this
point of view with the results shown in Table IV-2. The carbon monoxide
level averaged over all stations for each day ranged from 5.9 ppm
(October 30) to 11.3 ppm (November 4), while the average of estimated
values ranged from 6.8 ppm (September 11) to 11.3 ppm (November 4).
Carbon monoxide levels were overestimated for all stations on October 30,
-------
IV. 12
Table IV-1
Comparison of Estimated Mid-Hour Concentrations
to Observed Hourly Average Concentrations for
Carbon Monoxide, for Six Smog Days
at Different Receptor Locations
Estimated ave., ppm
Observed ave., ppm
Deviation, ave., ppm
Deviation, r.m.s., ppm
Deviation, % of obs.
LA
10.5
12.3
-1.8
5.5
52
Wh
8.9
7.3
1.6
2.9
40
Aza
5.6
6.3
-0.5
3.2
51
Bub
9.2
11.3
-2.1
6.0
53
All
8.6
9.2
-0.6
4.5
4.9
Q
One trajectory for Pasadena included as a substitute for one
Azusa trajectory, November 4.
Only 0930 trajectory available for Burbank on September 11; none
available November 4.
-------
IV. 13
Table IV-2
Comparisons of Estimated Mid-Hour Concentrations
to Observed Hourly Average Concentrations for
Carbon Monoxide, for Four Receptor Locations
on Different Smog Days
9/11 9/29 9/30 10/29 10/30 11/4
Estimated ave., ppm 6.8 8.4 8.1 8.9 8.4 11.3
Observed ave., ppm 8.8 9.0 10.3 10.0 5.9 11.3
Deviation, ave., ppm -2.0 -0.6 -2.2 -1.1 +2.5 0.0
Deviation, r.m.s., ppm 3.8 4.0 5.5 2.9 5.0 5.1
Deviation, % of obs. 43 44 53 29 87 45
Only 0930 trajectory available for Burbank, 9/11.
No trajectory available for Burbank; Pasadena substituted for Azusa
trajectory, 11/4.
-------
IV. 14
by an average 2.5 ppm (43%) for September 30, they were underestimated
by an average 2.3 ppm (22%). The fraction of estimates outside a
factor of two of the observed hourly average ranged from one in 21 or
5% (September 11) to 7 in 24 or 29% (September 29).
For various days, the root mean square deviation ranged from 2.9
ppm (October 29) to 5.5 ppm (September 30). The coefficient of vari-
ation was 0.55 or less for all days except October 30, on which it reached
0.87. (October 30 was the day having the lowest observed carbon monoxide
levels.)
C. COMPLETE ATMOSPHERIC SIMULATIONS
In accordance with contract requirements, atmospheric simulation
runs were performed for each of the six episode days specified by EPA,
for four receptor locations. For each receptor a 390 minute trajectory
was taken, arriving at the receptor at 1330 PST. In addition, for
October 29 only, two additional trajectories to each receptor location
were computed, with arrival times 0930 and 1130 PST and travel times
150 and 270 minutes, respectively.
Concentrations of nitrogen dioxide, nitric oxide, ozone and carbon
monoxide were calculated following each simulated 30-minute period of
travel for each of these thirty-two hypothetical air parcels. Initial
concentrations were estimated for the initial locations of the parcels
(at 0700 PST) by spatial interpolation of the concentrations observed
at the various air monitoring stations. A minimum value of 50 meters
was imposed, to override the calculated value of mixing depth wherever
the calculated value might be less than that figure.
The calculated concentrations for mid-hour trajectory locations
which were within a five-mile radius of the receptor location are
tabluated, with the associated observed hourly average values for
comparison in Table A-2 . Air monitoring stations chosen for these
-------
IV. 15
calculations were those at Downtown Los Angeles, Whittier, Azusa and
Pomona, with the Pasadena location as an alternative to Azusa on
September 11 and as an alternative to Pomona on September 29 and
November 4.
1. Simulation by Trajectories for 1330 PST Arrival
For evaluating success in simulation of photochemical air pollution,
the 1300 PST information is especially relevant, inasmuch as this is
the modal hour for occurrence of maximum daily oxidant readings; it
represents, in a sense, the most likely peak hour for this kind of smog.
Table IV-3 summarizes the information for four contaminants, ob-
tained by comparing the estimated 1230 and 1330 PST concentrations
against the hourly averages for 1200 to 1300 and 1300 to 1400 PST,
respectively.
Data for carbon monoxide form essentially a subset of the more
extensive data discussed in the previous section. The estimates used
in this comparison were generated by the standard (DIFFUN) routine,
with complete chemical kinetics included; however, inspection confirmed
that the values predicted were substantially identical with those
previously obtained, in which the chemistry was bypassed. The results
for this subset, as shown in Table IV-3, confirm that the 1330 PST
trajectories yield carbon monoxide estimates with characteristics
reasonably similar to those of the whole set. It may be noted that
the carbon monoxide concentrations as observed for these hours are, on
the average, smaller than those found for the whole set (5.9 ppm, as
compared with 9.2). At the same time, the predictions for the subset
have a correspondingly reduced r.m.s. deviation (2.5 ppm vs. 5.5), and
there is no bias toward either overestimation or underestimation
evident in these data.
Nitrogen dioxide, on the other hand, is appreciably overestimated,
as shown in the table; the average estimate is about 45% higher than
the average value observed, and of 37 individual comparisons, 30 showed
-------
IV. 16
Table IV-3
Comparison of Estimated Mid-Hour Concentrations
to Observed Hourly Average Concentrations for
Carbon Monoxide, Nitrogen Dioxide, Nitric Oxide
andjOzone, for 1330 PST Trajectories3
CO N00 NO 0,
— £ J
Estimated ave.i ppm 5.9 0.145 0.013 0.138
Observed ave.,|ppm 5.9 .099 0.014 .173
Deviation, ave;,, ppm 0.0 .046 0.001 -.035
Deviation, r.m.s., ppm 2.5 .084 0.008 .089
Deviation, % of obs. . 42 85 57 51
a Estimates for 1230 PST are included, where within 5 miles of
receptor station.
-------
IV. 17
overestimates. The r.m.s. deviation, accordingly, was 85 percent of the
mean observed value; however, if translated to equal means, the r.m.s.
deviation would be reduced to .07 ppm or less.
For nitric oxide, again there was no appreciable bias in the esti-
mation. However, most nitric oxide observations at these hours were so
small (0.01 ppm or less) that the comparison really shows only that
the estimated values were of the appropriate order of magnitude.
Ozone is somewhat underestimated by this program, in contrast to
nitrogen dioxide. Thus the average estimate was 0.138 ppm, about 20
percent less than the average value observed. The r.m.s. deviation was
about half as large as the observed average. Thus, it appears from
these comparisons that ozone is predicted somewhat more accurately than
nitrogen dioxide, but less accurately than carbon monoxide.
2. Extended Simulation of a Single Episode
The degree to which the model can simulate the progress of a single
episode of photochemical smog, in relation to the changes of concentra-
tions with time of day, was further investigated, using the episode of
October 29 for the comparison.
Predicted mid-hour concentrations for carbon monoxide, nitrogen
dioxide, nitric oxide, and ozone are graphically compared with the
corresponding hourly averages as observed at the air monitoring sta-
tions (located at Los Angeles, Whittier, Azusa and Pomona) in Figures
IV-7 to IV-10. (Data for comparison of oxides of nitrogen for Down-
town Los Angeles were incomplete; the only values available were for
0800 to 0900 PST: nitrogen dioxide 20 pphm and nitric oxide 24 pphm.)
A review of these Figures confirms that the model yields an ap-
propriate indication of the general trend of contaminant concentration
with time; for both carbon monoxide and ozone, for all four receptor
locations. For nitrogen dioxide, the trend indicated by the model is
similar to that observed for Whittier, but rather dissimilar for Azusa
-------
IV. 18
DOWNTOWN LOS ANGELES
10/29/69
BMEASUREDO,
— PES PREDICTED O,
10 11 12 13
d
i r
1 T
DOWNTOWN LOS ANGELES
10/29/69
IS MEASURED CO
' i PES MEASURED CO
J I
U 11 12 .13
Figure IV.7. Comparison of Calculated vs. Observed Values of Total
Oxidant and Carbon Monoxide at Downtown Lo9 Angeles
on October 29, 1969.
-------
IV. 19
AZUSA
10/29/69
ffi MEASURED NO2
— PES PREDICTED NO,
a a
10 -, 11 12 13
d
AZUSA
10/29/69
91 MEASURED NO
— PES PREDICTED NO
T T T T—f
8 9 10 I 11 12 13
AZUSA
10/29/69
HI MEASURED 0,
— PESPREDICTED 0,
10 '11 12 13
AZUSA
10/29/69
ffl MEASURED CO
— PES MEASURED CO
I I I I I I
8 9 10 I 11 12 13
Figure IV.8. Comparison of Calculated vs. Observed Values
of Nitrogen Dioxide, Nitric Oxide, Total
Oxidant and Carbon Monoxide at Azusa on
October 29, 1969.
-------
IV. 20
2V
10 —
WHITTIER
10/29/69
ffl MEASURED NO,
"•••" PES PREDICTED NO,
I I I I
9 10 — 11 12 13
a
— B
I I
I I
WHITTIER
10/29/69
HI MEASURED NO
— PES PREDICTED NO
'11
10 I 11 1Z 13
WHITTIER
10/29/69
Si MEASURED CO
— PES MEASURED CO
10 _J 11 12 13
Figure IV.9. Comparison of Calculated vs. Observed Values
of Nitrogen Dioxide, Nitric Oxide, Total Oxidant
and Carbon Monoxide at Whittier on October 29, 1969.
-------
IV. 21
POMONA
10/29/69
ffl MEASURED NO,
— PES PREDICTED NO,
9 10 _ 11 12 13
1 I I
I T
POMONA
10/29/69
ffl MEASURED NO
— PES PREDICTED NO
I I I I I T
9 10 I 11 12 13
I I
POMONA
10/29/69
8 MEASURED O,
^ PES PREDICTED O,
10 11 12 13
I I
POMONA
10/29/69
9 MEASURED CO
• PES MEASURED CO
a a
a a
I I I
10 I 11 12 13
Figure IV.10. Comparison of Calculated vs. Observed Values
of Nitrogen Dioxide, Nitric Oxide, Total Oxidant
and Carbon Monoxide at Pomona on October 29, 1969.
-------
IV.22
and Pomona. For nitric oxide, again, the trend indicated for Whittier
is similar to that observed, but the predicted values for Azusa and
Pomona were too low to reflect the observed trends.
An evaluation of the quantitative agreement between predicted and
observed episode levels can be appropriately made in terms of dosage.
Table IV-4 presents predicted and observed values of dosage (i.e.,
integrated concentration times duration) for the six-hour period 0800
to 1400 PST, for each receptor location, and for the average of all
four. (The Table also includes a column for NO , which is defined as
X
the total of nitrogen dioxide and nitric oxide.)
For carbon monoxide there appears to be reasonable agreement for
all locations except Pomona, where the dosage estimated by the model
is only 42 percent of that estimated from the air monitoring record.
Since the model estimates for the other contaminants are also substan-
tially lower than the estimates based on air monitoring, it appears
that there may have been & systematic source of error affecting con-
taminant values for the trajectories to Pomona on October 29, such as,
e.g., an underestimate of traffic emissions or oyerestimation of ground
temperature in the extreme eastern part of the region.
The only other striking discrepancy found in Table IV-4 is the low
model estimate'of nitric oxide dosage for Whittier. Together with the'
underestimated single-hour value for Los Angeles and the 3-station
average of 7 pphm-hr against 20, this suggests a systematic tendency
for the model to underestimate nitric oxide, especially in the earlier
morning traffic hours, when nitric oxide concentrations are relatively
high. This might imply that the model, with its currently incorporated
reaction mechanism and set of rate factors, predicts too rapid con-
version of nitric oxide to nitrogen dioxide. In accord with this
suggestion, it is seen that NO dosage estimates show reasonable
X
agreement, except at the Pomona location, while the model estimates for
nitrogen dioxide tend to be higher than those based on air monitoring
(again, except for Pomona).
-------
IV. 23
Table IV-4
Comparison of Dosage of Various Contaminants
as Estimated from Simulation Model to Dosage
Estimated from Air Monitoring Data,
Four Locations, Six Hours on October 29, 1969
Station
Los Angeles
\
Whittier\
Azusa
Pomona
Average
\
CO
ppm-hr
89/87
67/56
31/34
17/41
51/55
N02
pphm-hr
(42/20)a
145/137
42/32
31/66
73/78
N0_
pphm-hr
(8/24)a
10/23
6/9
6/27
7/20
NO
•x.
pphm-hr
(50/44)3
155/160
48/41
37/93
80/98
°3
pphm-hr
98/72
104/92
53/68
35/52
73/71
These values for the Los Angeles location represent only the hour
0800-0900 PST.
-------
IV. 24
Ozone dosages as estimated from the model calculations are in
reasonable agreement with those from air monitoring data for all four
locations. They range from about 35 percent higher (at Los Angeles)
to about 32 percent lower (at Pomona). This suggests that whatever
inaccuracies exist in the estimation of nitrogen dioxide and nitric
oxide in the model, these inaccuracies are not reflected proportionally
in the prediction of oxidant (ozone) levels.
3. General Accuracy of Atmospheric Simulations
Figures IV-11, IV-12 and IV-13 present scatterplots for the com-
parison of calculated and observed levels for carbon monoxide, nitrogen
dioxide and ozone (or oxidant) respectively, combining data from all
the atmospheric simulations discussed above.
a. Carbon Monoxide
With respect to carbon monoxide, the relation of calculated values
to observed hourly averages is not much different from the relation
found from comparisons with 67 trajectories in which only carbon
monoxide was estimated, as discussed in Section B, above. In this more
restricted set of data, the average observed carbon monoxide level
was 7.2 ppm, while the average estimated was 7.0 ppm, for an average
underestimate of 3 percent.
Figure IV-11 shows that, of sixty-one data pairs, only ten (or
16 percent) gave ratios (calculated/observed) greater than 2 or less
than 0.5. It is interesting to note, also, that eight of these were
attributable to the Pomona location, which yielded anomalously low
estimates for all contaminants (v.s., Section C). Thirty-seven, or
sixty percent of the sixty-one data pairs, showed deviations of 2 ppm
or less.
The correlation coefficient between calculated and observed values
is 0.82, accounting for nearly 70 percent of the variance in these
-------
IV. 25
20
I
of
O
o
o
UJ
x
o
o
1
EC
15
10
10
15
20
HOURLY AVERAGE CARBON MONOXIDE OBSERVED, ppm
Figure IV. 11. Scatter Plot of. Calculated vs. Observed Carbon
Monoxide for the Atmospheric Simulations.
-------
IV. 26
data sets, and corresponding to a value of 11 for Student's _t. These
values give evidence of good simulation, considering the present state
of the art of air quality modeling, even with the use of well established
procedures such as Gaussian plume diffusion models. The correlation
would certainly be even further improved if the data for Pomona tra-
jectories, which appear to be subject to some sort of systematic
underestimation, were to be adjusted by an appropriate correction;
howeveri no attempt has been made to develop such a correction.
b. Nitrogen Dioxide
Data for a total of 56 comparisons of estimated midhour nitrogen
dioxide levels with observed hourly averages are shown in Figure IV-12.
For nitrogen dioxide, the scatter of the data pairs is evidently some-
what greater than for carbon monoxide, and the coefficient of correlation
is smaller: 0.66, as compared with 0.82. (This coefficient, however,
corresponds to a t-value of more than 6, so that its statistical sig-
nificance is unquestionable; the criterion value of t for p = 0.1%
is only about 3.5).
The average estimate of nitrogen dioxide was higher than the
corresponding average of observations by 25 percent; this is substan-
tially less than the 45 percent overestimation found for the 1330
trajectories only (v.s., Section C-l). This apparent improvement
results partly from the fact that the estimates for the earlier hours
are relatively more accurate, and partly from the inclusion of data
for the Pomona trajectories, in which there were several substantial
underestimates.
In Figure IV-12, the broken lines indicate upper and lower cri-
terion levels such that points falling between them have ratios of
estimated to observed levels between 0.5 and 2, or have differences of
5 pphm or less (a reasonable allowance for experimental error for this
contaminant). Six points fall below the lower criterion line, all of
these corresponding to Pomona trajectory estimates. Eleven points
-------
IV.27
e
O
O
D
X
O
5
z
m
O
O
cc
10 20 30
HOURLY AVERAGE NITROGEN DIOXIDE OBSERVED, pphm
40
Figure IV.12. Scatter Plot of Calculated vs. Observed Nitrogen
Dioxide for the Atmospheric Simulations.
-------
IV. 28
fall above the upper criterion line, and these are not confined to any
one location.
These observations suggest that accuracy in the simulation of
nitrogen dioxide concentrations by the atmospheric model would perhaps
be improved by modifications in the chemical dynamics module designed
to reduce the simulated rate of conversion of nitric oxide to nitrogen
dioxide, or to increase the simulated rate of consumption of nitrogen
dioxide to produce nitric acid and organic nitrates.
c. Nitric Oxide
Observed levels of nitric oxide were, for the most part, too low to
be of any consequence relative to air quality. Of 56 observations,
only five were greater than 5 pphm; of these, all were underestimated.
by the model calculations (comparisons are shown in Figures IV-8b,
IV-9b and IV-lOb).
This relation between estimated and observed values for nitric
oxide is consistent with the hypothesis that the chemical dynamics
module, as used in these simulations, overestimates the rate of con-
version of nitric oxide to nitrogen dioxide.
d. Ozone (vs. Measured Oxidant)
Figure IV-13 presents data for 61 comparisons of estimated mid-
hour ozone concentrations with corresponding hourly average values of
i
total oxidant concentration. The correlation coefficient for this
collection of data is 0.83, which is very near ,that found for carbon
monoxide,
The average observed total oxidant was 14.1 ppm, while the average
of the estimated ozone values was 11.7, about 17 percent smaller. Six
of the estimates were more than twice as large ;as the corresponding
hourly averages, while eleven were less than half as large; the total
of 17 points outside these limits amounts to 28 percent of the entire
collection.
-------
IV. 29
HOURLY AVERAGE TOTAL OX1DANT OBSERVED, pphm
Figure IV.13. Scatter Plot of Calculated vs. Observed Total
Oxidant for the Atmospheric Simulations
-------
IV. 30
It is especially interesting that the accuracy of simulation of
ozone, an ultimate product of the photo-oxidation system in photochemi-
cal smog, appears to be greater than that of nitrogen dioxide, a
contaminant which is reasonably regarded as a precursor of the ozone.
At least, such is the case when the mechanism assumed for this project
is applied to the data for the chosen episode days. This circumstance
suggests, although it does not prove, that the ozone estimates are
less sensitive to uncertainties in certain rate factors, or other
parameters of the chemical reaction mechanism, than are the estimates
of the oxides of nitrogen. If true, this would be a valuable property
of the atmospheric simulation model, in connection with forecasting
of ozone emergency levels, or with evaluation of emission control
strategy for abatement of oxidaht air pollution problems.
D. CONCLUSIONS
The most important conclusion justified by these results is that
REM can successfully simulate observed behavior of photochemical
contaminants in the vicinity of stationary receptor sites, despite
the Lagrangian structure of the model. Although the same conclusion
has been indicated by earlier studies, the evaluation reported here
has been considerably more extensive than the previous work. For
data representing six episodes of photochemical smog in the Los Angeles
Basin, the model gave reasonable estimates of the major contaminants
over a six-hour period for each of four widely separated receptor
locations. Comparison with hourly average concentrations, as recorded
at air monitoring stations of the Los Angeles County network, confirmed
the magnitude of the concentrations calculated as instantaneous mid-
hour values, using the model.
Nevertheless, the evaluation indicated that agreement between the
model estimates and the air monitoring observations might, in certain
particulars, be improved by further attention to details of the postu-
lated distribution of contaminant emissions, by the incorporation of
-------
IV. 31
more complete information regarding topography and the variation of
vertical temperature profiles; and by optimization of the chemical
reaction mechanism, and the associated rate parameters, to be incor-
porated in the chemical dynamics module of REM.
More detailed, specific conclusions are the following:
1. Simulation of carbon monoxide concentrations in the atmosphere
by REM is no less accurate, on the whole, than by existing air quality
models which do not account for effects of chemical reactions. For
most of the conditions simulated by REM with the complete chemistry
module, carbon monoxide estimates showed a bias of less than 20 per-
cent. On a total of 61 direct comparisons, the correlation coef-
ficient was 0.82. On a total of 150 comparisons with estimates
obtained by REM with a special option for carbon monoxide alone, the
correlation coefficient was 0.70, and biases for various days and
receptor locations were mainly less than 25 percent. The average bias
for the whole body of data was a slight underestimation.
2. Simulation of ozone levels in the atmosphere gives values
which are, eminently reasonable when compared with the recorded data
for total .oxidant concentration. Relative to recorded total oxidant,
the ozone values were, on the average, low by less than 20 percent.
On 61 direct comparisons, the coefficient of correlation was 0.82.
3. Simulation of nitrogen dioxide levels was adequate to reflect
major trends in the observed concentrations as a function of time, but
was somewhat less accurate than simulation of carbon monoxide and
ozone. Estimates of nitrogen dioxide were high by 25 percent on the
average, with the degree of overestimation apparently increasing for
the later hours. The coefficient of correlation between mid-hour
estimates and recorded hourly averages, for 56 comparisons, was 0.66.
4. Observed concentrations of nitric oxide during the period of
interest in these smog episodes were, for the most part, too low for
-------
IV. 32
useful simulation. Of five observations which were larger than 0.05 ppm,
the corresponding model estimates were all less than 0.05 ppm. It
therefore appears likely that the model, with its presently incorpora-
ted chemistry, tends to underestimate nitric oxide by overestimating
the rate of its oxidation to nitrogen dioxide.
-------
A.I
APPENDIX
RESULTS OF REM SIMULATIONS
-------
A. 2
Table A-l
Carbon Monoxide Simulations
Trajectory
Run Termination
Number Point
Start
Time
Starting Other Points
Value End Within 5 Mi.
(PPM) Time of Station
Calculated/
Observed
(PPM)
September 11, 1969
101
102
103
104
105
106
107
108
109
110
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
Whittier
Whittier
Whittier
Azusa
Azusa
Azusa
Burbank
0700
0700
0930
0700'
0700
0700
0700
0700
0700
0700
12
0930
9
1130
9
1330
5
0930
7
1130
8
1330
8
0930
10
1130
5
1330
12
0930
0830
1030
1030
1230
1230
0830
1030
1030
1230
0830
1030
1030
1230
1230 '
0830
1030
11/16
9/16
7/16
9/16
9/12
. 11/11
9/11
9/10
5/5
4/6
4/6
6/6
7/5
8/4
8/5*
5/7*
4/7
3/7
5/7
6/7
4/7
5/7
5/8
8/12
6/12
5/12
Missing data; value interpolated from other hourly observations at monitoring
station.
-------
A. 3
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory
Run Termination
Number Point
Start
Time
Starting Other Points
Value End Within 5 Mi.
(PPM) Time of Station
Calculated/
Observed
(PPM)
September 29, 1969
111
112
113
114
115
116
117 :
118
119
120
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
Whittier
Whittier
Whittier
Azusa
Azusa
Azusa
Ryrbank
0700
0700
0930
0700
0700
0700
0700
0700
0700
0700
15 0830
0930
1030
15 1030
1130
14 1330
14 0830
0930
1030
15 1030
1130
9 1230
1330
10 0830
0930
1030
14 1030
1130
16 1330
18 0830
0930
1030
12/18
13/14*
10/10*
9/10*
8/6
7/3
11/11
9/11
6/13
9/13
7/7*
6/2
6/1
6/11
3/10
2/8
6/8
4/6
4/9
23/17
14/9
8/11
Missing data; value interpolated from other hourly observations at monitoring
station.
-------
A. 4
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory
Run
Number
121
122
Termination
Point
Burbank
Burbank
Start
Time
September
0700
0700
Starting
Value
(PPM)
29, 1969
16
17
September 30,
123
124
125
126
127
128
129
130
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
Whittier
Whittier
Whittier
Azusa
Azusa
0700
0700
0700
0700
0700
0830
0700
0700
20
18
17
12
17
18
12
14
End
Time
(cont'd)
1130
1330
1969
0930
1130
1330
0930
1130
1330
0930
1130
Other Points
Within 5 Mi.
of Station
1030
1230
0830
1030
0830
1030
1030
1230
0830
1030
1030
Calculated/
Observed
(PPM)
6/11
3/10
10/11
10/7
12/32
12/18
12/15*
9/12
4/4
7/7
7/6
6/5*
8/5*
7/5*
9/4
10/4
9/9*
6/7
4/6
4/6
4/6
Missing data; value interpolated from other hourly observations at monitoring
station.
-------
A.5
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory Starting Other Points
Run Termination Start Value End Within 5 Ml.
Number Point Time (PPM) Time of Station
131
132
133
134
135
136
137
138
139
140
Azusa
Burbank
Burbank
Burbank
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
Whittier
Whittier
Whittier
September 30, 1969 (cont'd)
0700 16 1330
0700 22
0930
0700 20
1130
0700 14
1330
October 29, 1969
0700 17
0930
0700 10
1130
0700 10
1330
0700 14
0930
0700 21
1130
0700 11
1330
0830
1030
1230
0830
1030
1030
1230
1230
0830
1030
1030
1230
1230
Calculated/
Observed
(PPM)
5/5
11/17
9/14
13/21
11/18
8/13
8/8
18/20
20/22
19/17
12/17
12/11
11/9
10/9
9/8
11/10
11/9
9/10
17/10
15/12
13/8
8/8
8/7
-------
A.6
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory
Run
Number
141
142
143
144
145
146
Termination
Point
Azusa
Azusa
Azusa
Burbank
Burbank
Burbank
Start
Time
October
0700
0700
0700
0700
0700
0830
Starting
Value
(PPM)
29, 1969
10
11
12
14
14
14
End
Time
(cont'd)
0930
1130
1330
0930
1130
1330
Other Points
Within 5 Mi.
of Station
0830
1030
1030
1230
1230
0830
1030
1030
1230
1230
Calculated/
Observed
(PPM)
7/11
5/6
4/3
6/3
5/4*
4/5
5/5
5/6
8/14
6/12
5/11
4/11
4/9
4/8
4/8
5/7
October 30, 1969
147
148
149
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
0700
0700
0700
16
16
16
0930
1130
1330
0830
1030
1030
1230
1230
17/8
10/7*
7/6
5/6
5/6
5/5
6/5
5/5
Missing data; value interpolated from other hourly observations at monitoring
station.
-------
A. 7
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory
Run
Number
150
151
152
153
154
155
156
157
158
Termination
Point
Whittier
Whittier
Whittier
Azusa
Azusa
Azusa
Burbank
Burbank
Burbank
Start
Time
October
0700
0700
0700
0700
0700
0700
0700
0700
0700
Starting
Value
(PPM)
30, 1969
18
9
15
8
8
7
18
17
17
End
Time
(cont'd)
0930
1130
1330
0930
1130
1330
0930
1130
1330
Other Points
Within 5 Mi.
of Station
0830
1030
1030
1230
0830
1030
1030
1230
1230
0830
1030
1030
1230
1230
Calculated/
Observed
(PPM)
22/19
17/9
9/4
5/4
3/3
4/2
4/3
8/3
8/3
4/2
4/2
3/1
2/1
2/1
1/1
27/21
30/12
12/8
5/8
4/5
3/5
3/5
3/2
-------
A. 8
Table A-l
Carbon Monoxide Simulations (cont'd)
Trajectory
Run
Nunber
Termination
Point
Start
Time
Starting
Value End
(PPM) Time
Other Points
Within 5 Mi.
of Station
Calculated/
Observed
(PPM)
November 4, 1969
159
160
161
162
163
164
165
166
167
Downtown
L.A.
Downtown
L.A.
Downtown
L.A.
Whittier
Whittier
Whittier
Pasadena
Pasadena
Pasadena
0700
0700
0700
0700
0700
0700
0700
0700
0700
15
0930
11
1130
12
1330
14
0930
13
1130
12
1330
16
0930
23
1130
10
1330
0830
1030
1030
1230
1230
0830
1030
1030
1230
0830
.1030
1030
1230
1230
15/30
15/22
16/16
9/16
8/12*
8/9
7/9
7/8*
14/11
13/14
12/12
12/12
12/9
10/7
9/5
16/10
10/7
5/7
19/7
15/6
11/11
9/11
8/8
Missing data; value.interpolated from other hourly observations at monitoring
station.
-------
Table A-2
Photochemical Pollutant Simulation Trajectories
Trajectory
Run Termination
Number Point
1 Downtown
L.A.
2 Whittier
3 Pomona
4 Pasadena
Start
Time
0700
0700
0700
0700
Starting Values
PPHM PPM
N00 NO
12 14
12 15
11 9
9 13
0,
CO
j
September
2
1
11
2
14
8
6
11
September
5 Downtown
L.A.
6 Whittier
7 Azusa
8 Pasadena
0700
0700
0700
0700
11 27
10 21
15 31
9 28
2
2
2
2
14
9
16
15
Other Points
End Within 5 Mi.
Time of Station
11, 1969
1230
1330
1230
1330
1330
1230
1330
29, 1969
1330
1330
1330
1230
1330
Calculated/Observed
PPHM PPHM PPHM
N0_
17/7*
22/5
30/23
33/20*
3/6*
29/20*
38/13
8/6*
17/10
8/5*
15/9*
12/5
NO
1/2*
2/1
2/2
2/1*
1/1*
1/3*
2/2
5/3*
1/1
1/1*
1/1*
1/1
0,
15/7
10/11
14/4
18/18
2/23
18/28
17/36
2/7
12/19
12/38
15/32
15/20
PPM
CO"
7/11
8/10
7/4
8/5*
1/7
11/7
13/8
3/3
5/1
3/9
5/~-
Missing data; value interpolated from other hourly observations at monitoring station.
Missing data; no interpolation possible.
-------
Table A-2
Photochemical Pollutant Simulation Trajectories (cont'd)
Trajectory
Run Termination
Number Point
Starting Values
Start PPHM PPM
Time NO,, NO 0^
CO
September
9 Downtown
L.A.
10 Whittier
11 Azusa
12 Pomona
0700 9 36 2
0700 10 38 2
0700 11 37 2
0700 11 36 2
17
18
16
17
October
13 Downtown
L.A.
14 Downtown
L.A.
15 Downtown
L.A.
16 Whittier
0700 14 27 3
0700 16 15 2
0700 16 14 2
0700 13 16 2
17
10
10
14
Other Points
End Within 5 Mi.
Time of Station
30, 1969
1330
1330
1330
1330
29, 1969
0830
0930
1030
1030
1130
1230
1230
1330
0830
0930
1030
Calculated/Observed
PPHM PPHM PPHM PPM
NO.
15/8*
11/24*
12/6*
11/13*
42/20
50/-
48/-
35/-
32/-
29/-
26 /-
24/-
24/20
25/25
22/26
NO
4/5*
1/1*
1/1*
1/1*
8/24
21-
21-
l/-
l/-
l/-
l/-
4/10
2/6
1/3
0.
3
4/17
8/5
14/27
16/15
4/5
13/8
23/11
19/11
20/14
22/16
22/16
20/19
4/3
11/6
17/10
CO
5/4
4/4
4/5
4/5
18/20
20/22
20/17
12/17
12/11
11/9
10/9
9/8
12/10
12/9
10/10
Missing dataj value interpolated from other hourly observations at monitoring station.
Missing data; no interpolation possible.
-------
Table A-2
Photochemical Pollutant Simulation Trajectories (cont'd)
Trajectory
Starting Values Other Points
Calculated/Observed
Run . Termination Start PPHM
Number Point Time N00 NO
0_
<- ~>
October
17 Whittier 0700 14 33
18 Whittier 0700 14 17
19 Azusa 0700 8 3
20 Azusa 0700 10 7
21 Azusa 0700 13 10
22 Pomona 0700 13 6
3
2
1.
2
3
2
PPM
CO
29,
21
11
10
11
12
10
End Within 5 Mi.
Time of Station
1969 (cont
1130
1330
0930
1130
1330
0930
'd)
1030
1230
1230
0830
1030
1030
"1230
1230
0830
1030
PPHM
NO
i
37/26
31/32
25/22*
20/22*
18/12
6/10
5/10
4/4
9/4
9/3
7/3*
12/3*
11/2*
11/15
3/15
2/12
PPHM
NO
1/3
1/2
1/1*
1/1*
1/1
1/4
1/1
1/1
1/1
1/1
1/1*
1/1*
1/1*
1/10
1/9
1/3
PPHM
°V
J
26/10
31/19
33/25
20/25
21/29
5/3
6/8
6/10
9/10
11/12
12/17
12/17
13/19
5/3
4/4
3/8*
PPM
~CO~
17/10
15/12
13/8
8/8
8/7
7/11
5/6
4/3
5/3
5/4*
4/5
' 5/5
5/6
6/14
2/9*
1/5
Missing data; value interpolated from other hourly observations at monitoring station.
-------
Table A-2
Photochemical Pollutant Simulation Trajectories (cont'd)
Trajectory Starting Values
Run Termination Start PPHM PPM
Number Point Time NO,, NO 00
' ' £. ' "J
October
23 Pomona 0700 10 5 1
24 Pomona 0700 84 1
CO
29,
10
10
October
25 Downtown 0700 7 42 1
L.A.
26 Whittier 0700 9 20 2
27 Azusa 0700 4 16 2
28 Pomona 0700 5 21 2
16
15
7
8
Other Points
End Within 5 Mi.
Time of Station
1969 (cont
1130
1330
30, 1969
1330
1330
1330
1330
'd)
1030
1230
1230
1230
1230
1230
1230
Calculated/Observed
PPHM PPHM PPHM
NOT"
.._ 2,
4/12
3/9
2/8*
7/8*
3/6*
11/9
10/11*
8/9*,
8/8
1/2*
1/1*
1/5*
1/5*
NO
1/3
1/2
1/2*
1/2*
1/1*
1/1
1/1*
1/2*
1/3
1/1*
1/1*
1/3*
1/2*
0,
J
5/8*
6/12
5/12
9/12
6/13
12/13
11/15
11/10
8/10
2/7
2/9
2/4
2/6
PPM
CO
2/5
2/5
2/4
3/4
2/4 >
i-i
S3
4/5
4/5
3/2
3/3
1/1*
1/1
1/4
1/4
Missing data; value interpolated from other hourly observations at monitoring station.
-------
Table A-2
Photochemical Pollutant Simulation Trajectories (cont'd)
Trajectory
Run Termination
Number Point
Starting Values
Start PPHM PPM
Time NO,, NO 0,
CO
t. j
November
29 Downtown
L.A.
30 Whittier
31 Azusa
32 Pasadena
0700 12 24 2
0700 12 25 2
0700 12 27 2
0700 11 28 2
12
16
14
10
Other Points
End Within 5 Mi.
Time of Station
4, 1969
1230
1330
1230
1330
1330
1230
1330
Calculated/Observed
PPHM PPHM PPHM PPM
NO
1 /
20/18
15/15
27/16*
25/10
17/7*
25/12*
23/12
NO
1/1*
1/1
1/1*
1/1
1/2*
1/1*
1/1
22/10
12/16
22/10
23/11
26/23
22/21
25/21
CO
10/9 s
6/8*
10/7
9/5
8/7
9/11
8/8
Missing data; value interpolated from other hourly observations at monitoring station.
-------
A.14
1 T
OHESI
POMONA
15 ZO 25 30
40 45 50 55
(X-Axii in Milei)
65 70
Figure A.I
Trajectories for September 11, 1969
-------
A.15
0«
WEST LOSANGELE
122 A
o
LOSANGELESV.^""'•———^, >»•
INTERNATIONAL ,,CA ^«^
O LONG BEACH
J I
O'
REM Simulation Trajectories, S«pl 29.1969
LEGEND
STATION ARRIVAL TIMES
930
— — ••••1130
^_._1330
A -CARBON MONOXIDE
B - ALL CONTAMINANTS
Q- RECEPTOR POINTS
• - TRAJECTORY START POINTS
J I
J I
15 20 25 30
35 40
50 55
70 75
IX-Axil in Mital
Figure A.2
Trajectories for Septetnber 29, 1969
-------
A. 16
PASADENA
o
> I A\
LOS ANGELES ft\ ^ I L
IATIONAL AIRPORT Vl/ ^X I _ *.
• 134A
EL MONTE O/ / /'
/ I /
... ..... . ,. A
y^ LOS ANGELES // ^
o I// / \
1 /
! ^
/
CITV OF COMMERCE
Simulation Trajoctorin. S«pt 30, 1969
LEGEND
STATION ARRIVAL TIMES
930
——1130
1330
A - CARBON MONOXIDE
8 -ALL CONTAMINANTS
- RECEPTOR POINTS
~ TRAJECTORY START POINTS
(X-AxiiinMiln)
Figure A.3
Trajectories for September 30, 1969
-------
A.17
\ I I46A« {08301
Figure A.4
Trajectories for October 29, 1969
-------
A.18
70 75
Figure A.5
Trajectories for October 30, 1969
-------
A. 19
Q- RECEPTOR POINTS
• - TRAJECTORY START POINTS
I I I I I I
20 25
Figure A.6
Trajectories for November 4, 1969
* U.S. Government Printing Office: 1973"7'<6-771/'il88 Region No.
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
EPA-R4-73-013a
3. Recipient's Accession No.
4. Title and Subtitle
Controlled Evaluation of the Reactive Environmental
Simulation Model (REM) Volume 1: Final Report
S. Report Date
February 1973
6.
7. Author(s)
L.G. Wayne, A. Kokin, and M.I. Weisburd
8. Performing Organization Kept.
No.
9. Performing Organization Name and Address
Pacific Environmental Services, Inc.
2932 Wilshire Boulevard (Suite 202)
Santa Monica, California 90403
10. Project/Task/Work Unit No.
11. Contract/Grant No.
68-02-0345
12. Sponsoring Organization Name and Address
ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Monitoring
Washington, D. C. 20460
13. Type of Report At Period
Covered
Final
14.
15. Supplementary Notes
16. Abstracts The development and validation of an operational version of the Reactive En-
vironmental Simulation Model (REM) were completed. REM was specifically designed to
handle large chemical mechanisms to assess the impact on air quality of air pollution
control devices, fuels, propulsion systems, stationary sources, and transportation sys-
tems where thorough evaluation of emissions, emission constituents and reaction rates
are required. This completed version of the model contains a mechanism involving 32
reactions, 12 accumulating species and 12 non-accumulating species. REM contains such
user features as reverse and forward trajectory routines; automatic and objective inter
polation from input emission inventory, meteorological and air quality data bases; a
chemical dynamics routine capable of accommodating mechanisms based on elementary chemv
cal reactions: and automatic estimation of mixing depth and solar irradiance based on
input of local weather and sun angle data. REM has attained a running time which makes
it cost-effective for practical use. The program also is user-oriented in that it pro-
vides simple input procedures, user documentation, receptor point and time-of-day selec'
tivity, flexibility in treating specific problems, and ability to select any of an in-
finite number of trajectories on anv number of davs of interest.
17. Key Words and Document Analysis. 17o. Descriptors
Exhaust emissions
Carbon monoxide
Trajectories
Air pollution
Photochemical reactions
Mathematical models
Simulation
Chemical reactions
Kinetics
Turbulence
Irradiation
Chambers
Nitrogen oxides
17b. Identifiers /Open-Ended Terms
Reactive Environmental Simulation Model (REM)
Air pollution control
Environmental model
Photo-oxidation
Air pollution episodes
17c. COSATI Fie Id /Group
18. Availability Statement
Unlimited
19..Security Class (This
Report)
UNCLASSIFIED
207 Security Class (This
Page
UNCLASSIFIED
21. No.'of Pages
175
22. Price
FORM NT1S-35 (REV. 3-72)
USCOMM-DC I40S2-P72
-------
INSTRUCTIONS FOR COMPLETING FORM NTIS-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 bound 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-87 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 (he 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 Dote. F.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(«). 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 Organization 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/Tosk/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 Nome 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 . . .
16. Abstroct. Include a brief (200 words or less) factual summary of the most significant information contained in the report.
If the report contains a significant bibliography or literature survey, mention it here.
17. Key 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 196? COSATI Subject Category List.
Since the majority of documents are multidisciplinary in nature, (he primary Field/Group assignment(s) will be the specific
discipline, area of human endeavor, or type of physical object. The application(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.
FORM NTIS-35 IREV. 3-72) USCOMM-DC 14«B2-P72
------- |