EPA-600/3-77-118
November 1977 Ecological Research Series
INTERNATIONAL CONFERENCE ON OXIDANTS,
1976 • ANALYSIS OF EVIDENCE
AND VIEWPOINTS
Part VI. The Issue of Air Quality
Simulation Model Utility
vS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal spe-
cies, and materials. Problems are assessed for their long- and short-term influ-
ences. Investigations include formation, transport, and pathway studies to deter-
mine the fate of pollutants and their effects. This work provides the technical basis
for setting standards to minimize undesirable changes in living organisms in the
aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-77-118
November 1977
INTERNATIONAL CONFERENCE ON OXIDANTS, 1976 -
ANALYSIS OF EVIDENCE AND VIEWPOINTS
Part VI. The Issue of Air Quality Simulation Model Utility
J.H. Seinfeld
California Institute of Technology
Pasadena, California
Contract No. DA-7-2143A
K.R. Wilson
University of California at San Diego
La Jolla, California
Contract No. DA-7-219U
Project Officer
Basil Dimitriades
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or recom-
mendation for use.
In general, the texts of papers included in this report have been repro-
duced in the form submitted by the authors.
11
-------
ABSTRACT
In recognition of the important and somewhat controversial nature of the
oxidant control problem, the U.S. Environmental Protection Agency (EPA)
organized and conducted a 5-day International Conference in September 1976.
The more than one hundred presentations and discussions at the Conference
revealed the existence of several issues and prompted the EPA to sponsor a
followup review/analysis effort. The followup effort was designed to review
carefully and impartially, to analyze relevant evidence and viewpoints report-
ed at the International Conference (and elsewhere), and to attempt to resolve
some of the oxidant-related scientific issues. The review/analysis was con-
ducted by experts (who did not work for the EPA or for industry) of widely
recognized competence and experience in the area of photochemical pollution
occurrence and control.
John H. Seinfeld, California Institute of Technology, and Kent R. Wilson,
University of California at San Diego, review the issue of Air Quality Simu-
lation Model (AQSM) utility. The strengths and weaknesses of the various
modeling techniques are discussed, and the authors offer their recommendations
on future studies.
111
-------
CONTENTS
ABSTRACT iii
FIGURES vi
TABLES vii
INTRODUCTION
B. Dimitriades and A. P. Altshuller 1
THE ISSUE OF AIR QUALITY SIMULATION MODEL UTILITY
B. Dimitriades and A. p. Altshuller 3
REVIEW AND ANALYSIS
J. H. Seinfeld 7
Introduction 7
Fundamentals of AQSMs 8
Currently Available AQSMs for Photochemical Oxidant 16
Sources of Uncertainty in AQSMs 28
Assessment of Accuracy of AQSMs for Photochemical Oxidant 40
Conclusions and Recommendations 56
Appendix 58
REVIEW AND ANALYSIS
K. R. Wilson 77
Abstract 77
Introduction 78
Alternative Emissions to Oxidant Models 78
Sources of Error 83
Utility and Areas of Application 87
Directions for Improvement 94
Conclusions 99
REFERENCES 101
BIBLIOGRAPHY 115
V
-------
FIGURES
REVIEW AND ANALYSIS - J.H. Seinfeld
Number Page
Denver air quality model validation comparisons
for 29 July 1975 49
Denver air quality model validation computed/observed
comparison at various stations for 28 July 1976 50
Denver air quality model validation computed/observed
comparison at various stations 3 August 1976 51
The variation of averages over all stations of observations
and predictions 52
Denver air quality model validation computed/observed
correlations (data for 3 days, 9 stations,
daylight hours) 53
REVIEW AND ANALYSIS - K.R. Wilson
Uses and utilities of PCMs 89
VI
-------
TABLES
REVIEW AND ANALYSIS - J.H. Seinfeld
Number Page
1 Sources of Invalidity and Inaccuracy in AQSMs 13
2 Currently Available AQSMs for Photochemical Oxidant 17
Comparison of Lumped Chemical Kinetic Mechanisms for
Photochemical Smog 20
The Treatment of Meteorological Variables in AQSMs for
Photochemical Oxidant 24
Source Inventories Required for AQSMs for Photochemical
Oxidant 29
A Summary of the Sensitivity of the SAI Model to Variations
in Input Parameters 43
7 Sensitivity Studies with the Bell Laboratories Model .... 44
8 Validation Studies with AQSMs 46
A-l Selected Inorganic Reactions 59
VI1
-------
ACKNOWLEDGMENTS
These contracts were jointly funded by the Office of Research and Develop-
ment (Environmental Sciences Research Laboratory) and the Office of Air Quality
Planning and Standards.
The assistance of the technical editorial staff of Northrop Services,
Inc. (under contract 68-02-2566) in preparing these reports is gratefully
acknowledged.
Vlll
-------
INTRODUCTION
Basil Dimitriades and A. Paul Altshuller
In recognition of the important and somewhat controversial nature of the
oxidant control problem, the U.S. Environmental Protection Agency (EPA) organ-
ized and conducted a 5-day International Conference in September 1976. The
one hundred or so presentations and discussions at the Conference revealed the
existence of several issues and prompted EPA to sponsor a followup review/
analysis effort. Specifically, this followup effort is to review carefully
and impartially and analyze relevant evidence and viewpoints reported at the
International Conference (and elsewhere) and to attempt to resolve some of the
oxidant-related scientific issues. This review/analysis effort has been
contracted by EPA to scientists (who do not work for EPA or industry) with
extensive experience and expertise in the area of photochemical pollution
occurrence and control. The first part of the overall effort, performed by
the EPA Project Officer and reported in a scientific journal (1), was an
explanatory analysis of the problem and definition of key issues, as viewed
within the research component of EPA. The reports of the contractor expert/
reviewer groups offering either resolutions of those issues or recommendations
for additional research needed to achieve such resolutions are presented in
the volumes composing this series.
This report presents the reviews/analyses prepared by the contractor
experts on the issue of air quality simulation model utility. In the interest
of completeness the report will include also an introductory discussion of the
issue, taken from Part I. The reviews/analyses prepared by the contractor
experts follow.
-------
THE ISSUE OF AIR QUALITY SIMULATION MODEL UTILITY
Basil Dimitriades and A. Paul Altshuller
Relationships between emissions and oxidant-related air quality are being
pursued following three distinct approaches that differ mainly in degree of
empiricism. In order of decreasing empiricism these approaches are:
1. Empirical Approach. The approach entails statistically or non-
statistically associating ambient oxidant-related air quality data
either with ambient concentrations of precursors or with precursor
emissions rates. These associations clearly are not of cause-effect
nature, and their intended use is not to predict absolute air quality;
rather, it is to predict changes in air quality resulting from
changes in emission rates.
2. Smog Chamber Approach. The approach entails deriving cause-effect
relationships between oxidant and precursors through laboratory
testing. This approach could be characterized as semiempirical
because the relationships are derived from laboratory observations
alone; they are not the product of theoretical derivations. Further,
as in the preceding case, this approach is intended to predict only
changes in air quality resulting from changes in emission rates.
3. Mathematical Modeling (or AQSM) Approach. The approach entails
deriving the requisite air quality-emission relationships entirely
from theory. Its intended use is to predict both absolute levels of
and changes in air quality from given emission rate and meteorologi-
cal data.
-------
To be usable, all methods of relating air quality to emissions, regardless of
approach, must be validated and evaluated for accuracy. The distinction
between "validation" and "accuracy-evaluation" follows. Validation refers to
the agreement between model-predictions and observations when the input
information fed into the model is perfectly accurate; thus, validation is the
process of checking the validity of the principle underlying the method.
Accuracy-evaluation refers to the agreement between model-predictions and
observations for a model based on a perfectly sound principle; thus, accuracy-
evaluation is an assessment of the error introduced by inaccuracies of the
input information. Another term often used in connection with model evalua-
tion is "verification," referring to the agreement between predictions and
observations for the specific case in which the observations used for verifi-
cation were taken from the same pool of data used to develop the model. This
is the case, for example, of development and verification of AQSMs from the
St. Louis RAPS data. In the discussion here, verification will be considered
to be a limited form of validation.
Between validation and accuracy-evaluation, the latter appears to be
relatively simpler, especially for the empirical and semi-empirical methods
and for the relatively simple AQSM methods. Thus, useful accuracy evaluations
can be made from estimates of the errors associated with the input information
and from numerical sensitivity tests to determine the impacts of such errors
upon model-predictions. Unlike accuracy-evaluation, complete and direct
validation of a model is extremely difficult — if at all possible — to accom-
plish for the main reason that the requisite "real world" data - on air quality
and emissions — are either not available or not easy to obtain. Thus, for the
empirical and semiempirical models relating emission changes to air quality
changes, data on such changes either do not exist, or, if they do exist, as
for the Los Angeles basin, they are useful only for verification of a "local-
use" model. For the AQSM methods, intended to relate absolute levels of air
quality to emissions rates, validation hinges upon solution of several prob-
lems, one of which is the definition of absolute air quality in terms of
commonly obtained air quality monitoring data. At the present time, these
problems in validating models are considered to be prohibitive by some inves-
-------
tigators, but not insurmountable by others. It is this latter disagreement
among investigators that constitutes the issue to be examined here.
More specifically, the question at issue here is:
• At the present time, are there any air quality simulation models
sufficiently validated/evaluated and appropriate for use in designing
urban oxidant control strategies'
The Environmental Protection Agency has not issued nor does it have plans
for immediate issuance of strategy design guidelines (for oxidant control)
involving use of AQSMs. Furthermore, EPA is conducting an extensive study
(RAPS project) to first verify and subsequently further validate and evaluate
several of the presently available AQSMs. Obviously, therefore, EPA does not
feel that at this time there are AQSMs ready for use. Contrary to this EPA
viewpoint, some investigators have suggested that there is one sufficiently
validated air quality simulation model (Bell Lab). The Bell Lab Model was
discussed at the International Conference; however, the more detailed descrip-
tion of the model and justification of the Bell Lab viewpoint are to be found
in a subsequently published journal article (2).
There have been numerous reports, presented at the International Confer-
ence and elsewhere, on the principle, status of validation, and intended
utility of several AQSMs. It is important, however, to make the distinction
between true model applications and model exercises that merely demonstrate
the intended utility of a model. The judgment called for here is on whether
there can be true model applications, that is, on whether there are models
presently available which can and should be used immediately in designing
urban oxidant control strategies.
-------
REVIEW AND ANALYSIS
John H. Seinfeld
INTRODUCTION
The purpose of this report is to address the question:
• At the present time are there any air quality simulation models*
sufficiently validated/evaluated and appropriate for use in de-
signing urban oxidant control strategies?
The following section is devoted to a brief summary of the basic mathe-
matical relationships underlying air quality simulation models (hereafter de-
noted by AQSMs). The mathematical bases and assumptions inherent in all common
AQSMs have been covered in detail elsewhere (3), and we choose not to reiterate
that material here. We present only those concepts necessary to discuss the
validity and accuracy-evaluation of AQSMs.
In the third section, we review the elements of several of the currently
available AQSMs. We have chosen not to review all available AQSMs for photochemical
oxidant. Only those AQSMs are reviewed that, in the author's opinion, are can-
didates for use by EPA in designing urban oxidant control strategies. The pur-
pose of the review is to compare carefully the different manners of treatment
of key physical and chemical processes in the principal AQSMs.
The fourth section is devoted to a discussion of the sources of uncertainty
in AQSMs. In particular, we focus on those elements of AQSMs that have substantial
*As used here, air quality simulation model will be taken to represent a
physicochemical model based on theoretical treatments of atmospheric
chemistry and meteorology.
-------
bearing on oxidant prediction and for which there are elements of some uncer-
tainty, arising either from lack of knowledge of the fundamental physics and
chemistry or from inaccuracies in input information.
The fifth section contains an assessment of the accuracy of AQSMs. In this
section we seek to answer the question posed at the outset.
In the sixth and final section, we provide conclusions and recommendations.
FUNDAMENTALS OF AQSMs
All conventional atmospheric diffusion models are based on the
equation of conservation of mass:
3c. 3c. 9c. 3c. /32c. 32c. 32c.
(Eq.
where c. is the concentration of species i; u, v, and w are the fluid veloci-
ties in the three coordinate directions; D. is the molecular diffusivity of spe-
cies i in air; R. is the rate of generation (or the negative of the rate of
disappearance) of species i by chemical reactions at temperature T; and S. is
the rate of injection of species i into the fluid from sources.
Substituting the usual mean and fluctuating terms into Equation 1 and
averaging the resulting equation over the ensemble of flows, we obtain the
equation governing the ensemble average concentration c.. In atmospheric
applications, the molecular diffusion term is negligible when compared to
that representing advective transport. Thus, neglecting the contribution of
molecular diffusion, the equation for c. is
3c. 9c. 3c. 3c. „
+ u -— + v -— w -— + —- u'c! + — v'c! +7— We! = R.(c,,...,c ,T)
^ ' ^ f\ lvf\ w f\ ^ , ^ *~ • -\
3t ^x 3y 3z 3* i 3y
+ Si(x,y,z,t)
(Eq. 2)
-------
Equation 2 is a rigorously valid equation for c. (neglecting, of
course, molecular diffusion); and, if the variables u'c., v'c., w'c., and
any of those arising from R. are known as functions of space and time, it can
be solved in principle to yield c.. Unfortunately u'c. and so on cannot be
measured at all points in an atmospheric flow and cannot be predicted ex-
actly because of the closure problem of turblent flow. Thus, we must resort
to models for these terms. The model employed in virtually all cases in
which atmospheric flows are involved is taat based on the concept of eddy
diffusivities:
3Cj
u'c! = - K i v'c! = - K i -r-r = - ^, i (Eq. 3)
1 H -r 1 H -r w'c! V -r
dx oy i 3z
The eddy diffusivities K and 1C are postulated to be functions of space and
H V
time (and not of c. or any of its gradients).
Although there has been some study on the nature of terms of the form
•~\
c arising from turbulent chemical reactions, no atmospheric diffusion models
for chemically reactive pollutants currently include expressions for these
terms. All models neglect the contribution of turbulent concentration fluc-
tuations to the mean reaction rate and employ the approximation,
(Eq. 4)
The result of using Equations 3 and 4 in Equation 2 is the so-called
atmospheric diffusion equation (ADE):
3c. 3c. 3c. 3c.
Si(x,y,z,t)
(Eq. 5)
-------
Because the governing equations are nonlinear, they must be solved
numerically. Futhermore, the use of numerical techniques generally
requires that the modeling region be subdivided into an array of grid
cells, where each cell may have horizontal and vertical dimensions on
the order of a few kilometers and several tens of meters, respectively.
Before the general mass continuity equation can be solved, it must be
"filtered" to remove all small scale variations that the grid cannot
resolve, both in the concentration field and in the independent para-
meters, such as the wind velocities and the eddy diffusivities. The
necessary filtering can be accomplished by averaging Equation 5 at each
point over a volume equivalent to that of a grid cell. This spatial
averaging will be denoted by the symbol < >. In addition, Equation 5
has been time-averaged over an interval equivalent to that used in each
step of the numerical solution procedure. Thus, the concentration
predictions obtained from Equation 5 represent spatially and temporally
averaged quantities.
As a result of spatial averaging of Equation 5 we obtain*
3 3 3 3
. . .
1 3/K i
(Eq. 6)
As in the case of Equation 4, all models employ the approximation,
Ri
(Eq. 7)
thereby neglecting the contribution of subgrid-scale concentration variations to
the mean reaction rate.
*We assume that u, v, and w represent spatially averaged quantities by virtue
of the manner in which they are determined.
10
-------
Employing Equation 7 we obtain the equation that is the basis of
all AQSMs,
9 9 9 .
i - i - i - i =
~~ + u ~~ v ~~ w ~~
(Eq. 8)
The validity of the atmospheric diffusion equation relates to how
closely the predicted mean concentration corresponds to the true
ensemble mean concentration. If the mean velocities, u, v, and w, and
the source emission rate S. are known precisely at all points as a
function of time, then, for an inert species, the only source of a
discrepancy between the predicted and true mean concentrations is the
eddy diffusivity model for the turbulent fluxes. If the true ensemble
mean velocities and concentrations are known for an atmospheric flow,
then it is relatively straightforward to assess the validity of Equa-
tion 8 for specified forms of K and K . Unfortunately, for any atmo-
spheric flow the ensemble mean velocities and concentrations can never
be computed since the atmosphere presents only one realization of the
flow at any time. (Of course, for a statistically stationary flow,
ensemble averages can be replaced by time averages. The atmosphere is,
however, seldom in a stationary condition for any appreciable period of
time.) Because the true mean velocities and source emission rates that
are required to solve Equation 8 and the true mean concentration with
which the solution of Equation 8 is to be compared are not available in
general, an unambiguous measure of the validity of Equation 8 for any
particular flow cannot be obtained.
Accuracy-evaluation refers to the agreement between model predic-
tions and observations for a model based on a perfectly sound principle;
thus, accuracy-evaluation is an assessment of the error introduced by
11
-------
inaccuracies of the input information.* Whereas an assessment o. model
validity is very difficult to obtain, accuracy-evaluations can be made
from estimates of the errors associated with the input information and
from numerical sensitivity tests to determine the impacts of such errors
on model predictions. Unlike verification and accuracy-evaluation,
direct determination of the validity of an AQSM is extremely difficult
to accomplish because the requisite exact data on emissions, meteoro-
logical variables, and air quality are neither available nor easy to
obtain. It is therefore necessary to rely on combinations of verifi-
cation and accuracy-evaluation studies to judge the adequacy of an AQSM.
By necessity, we adopt this approach here.
Table 1 summarizes the sources of invalidity and inaccuracy of
Equation 8. The sources of invalidity cannot be directly assessed for
the reasons just stated. The sources of inaccuracy, on the other hand,
can be assessed through verification and accuracy-evaluation studies.
The inputs needed to solve the atmospheric diffusion equation
together with possible sources of error in those inputs are given in
Table 1. In each instance unless the actual value of the input is
known, the level of error in that input can only be estimated. From the
standpoint of the effect of errors on the predictions of the equation,
joint consideration must be given to the level of uncertainty in each
input parameter and the sensitivity of the predicted concentrations to
the parameter. Uncertainty relates to the possible error in the pa-
rameter from its true value, and sensitivity refers to the effect that
variation in that parameter has on the solution of the equation. A
parameter may have a large uncertainty associated with it but have
*Another term often used in connection with model evaluation is "verification,"
referring to the agreement between predictions and observations for the spe-
cific case in which the observations used for verification were taken from
the same pool of data used to develop the input information for the model.
Verification contains elements, therefore, of both validation and accuracy-
evaluation. Henceforth, verification studies will be referred to as valida-
tion studies in keeping with the prevailing usage.
12
-------
tn
IS
w
CX
£2
H
>H
U
%
8
u
J2
H
Q
5!
£
H
Q
H
H
fa
O
w
w
0
&
D
O
,
rH
W
§
EH
4->
C
0)
5
0
0
£^1
jj
•H
TI
-H
H
(d
^
C
H
MH
0
w
0)
o
SH
3
0
SH
0
SH
SH
0)
MH
0
0)
u
SH
3
O
W
SH
0)
0
4-1
C
(U
g
0
0
SH
4-1
C C
,e 0) (d tp
u g o tn c
3 CU (d 0
c/) SH >, SH
•H SH tn 4-)
3 4-) 4-> en
• W en c
en cu -H cu o)
g j^j £ £ ^i
C cu cu CD
CD i — ! _C ^ C"!
ft 4-1 (d O -H 3
g
•H
SH
0)
MH
MH
O
rH
rH
•H
3
en
rH
CU
Ti
O
g
0)
SH
en
O
rH
U
\^
0)
Ti
SH
0
fa
cu
tn
•H
K
^
tn
0)
X
3
rH
MH
-p
C
CU
rH
3
X!
SH
3
4-1
cu
4J
MH
O
H
o
MH
0)
3
SH
EH
C 3
cu o 4-) tr1 en
en -H ecu c
cu 4-> 0) SH o
X! O 4-> -H
C SH (d
CU CU O 4-1 =
en tn MH 3 rH SH
0) SH ft CU 3
SH (d cu g T) o
ft H 4J O O O
cu (do go
SH o -H
4-i SH 0) 0) en
C ft tn H d)
•H Ti O SH (d U
(d S-i (d O SH
en 0) ft rH en 3
0) rH ft 10
-H (d 4-i • o en
4-i tn 3 cu SH
-H T! en X! en o 0)
> 0 H -H -H C
•H ,C 01 TI SH g -H
tn 4-> TI 0) (d = rH
30) O ft
MH E g O >i 0) T!
MH rH (d 0 C
•HO) cu cu E 3 (d
Ti SH rH > Ti
3 3 0) 0) O 4->
>i en tn T) > SH C
T! O 0 O 4J -H
T! rH rH 0) X! C O
CU O U XI id H ft
CO
4-1
B(d
SH
C SH
O MH C
•H ~ 0
4-i -* en -H
0 C 4-1
id C 00
cu o -H (d
SH -H 4-)CU
4-1 (d M
, — | nj rj _ s
(d 3 4-1 H r-
C u tJ1 o (d
3 -H W 3 u c
O E rH -H O
C 0) • MH g -H
,k x; tn a) 4-)
c u • c x;
-H 30 e e •
O 4-10) Q) -H 0)
0 rH 0 tn ^
3 3 tn c (d
rH 0) 0 M Tl
- MH C O CU 0)
-•H > 4-1
U 4-i 0) iu rfl o
C SH O 0)
> Q) (d rH rH
rH 4-> (d tn
3 W 0 -H 0)
-•H XI E 0) 4-) C
O SH C MH (d
30) MH ft w
3 EH 4J H en -H
4-1
C
0)
p
5
0
O
K**1
O
(0
ik-l
3
O
o
(d
c
H
MH
O
en
0)
u
SH
3
0
W
SH
O
^H
SH
0)
MH
0
CU
U
SH
3
O
0) 0)
x; o
-P g 3
O Ti
E SH cu
O MH S-i
SH
MH CU O
rH • 4-1
C ft rH
(0 -H 0) T5
CU U Ti 0)
ECO Ti
•H E 0)
0) SH 0)
3 ft 0) C
SH o
4-1 C C »
•HO) 0)
0) rH -H
x; T) 3 SH
4-> 0) X5 O
4-> SH 4->
cu m 3 c
C H 4-) 0)
•H 3 > •
gO rH C >1
rH (d -HO
o> (d o nj
4J 0 -H CM
0) C 03
TI 0' ro -HO
xi x; w o
o o w m
4-1 C 0) -H C
(d E g -H
>i o S
(d, TI MH
31 3 -H T! 0
3 0)
0, » H 1-10)
Cl > m -HO
(d M
V)*>< 4-13
•H| 3 o o) O
(d TJ tn
0) •» SH
SH (d 3 0) en
0) 4-) o SH -H
X! (d o ox;
EH Ti id S 4J
•
0)
TJ -P
cu SH m cu (0
SH 0) O 0) T! SH
(d Xj ft e 3 MH en
0),g 4-) oien -H oo SH
3l33C> 3 0 o
4-1 TI S-PCX; co o
§0) 0) (d -H 4-1 • -H -H td
4^ Sn-4-13 -Hen 4J MH
0-H 3CC 30) cn(d
c.^c enocuMH 3 -HO e
I > -H (TJ -rH tn O Ti O1 -H 0
CU MH 0) 4-1 0) 0) -H -iH4H -H
SH,» EOSnen 4JC w-H en
)4-> ,G-HrHO) 00) Oft E
H(d 4-ITilg O4-> -Htn-CU
, - rH e 0) en 4-i C
|>(d(d CTiOSH-enen uooc
34-1 -H C C 3 C (d -H C C -H -H
^cn(d (d eno en 3 4->
1 3 3 TJ en SH (d -H tn ^ MH SH (d en
— 0)TJOO)4-JO)i — 1 OOCU
en E -HO) E o -H id c O -H
Q)enO'4-)ci)CU CU4-1C OO)rH4-l
•HCSH— ~Cft4-lrHSHC(d -H4J C
4-i (d MH en -H en (d (d -H -H en (d 0) -H
OET!O4JT!O14J 4-lrH -rH3rH4-l
0 0)-HSHCO)(dTi>HO) EO3SH
HO)4->4-)0)-HTiacO)-H OOOO)
0)1 — i(d(dO3(dtn(dOM-i (deno
>xjrHuc c c cue c
g3OD M D OH D
C 0) 0 rH SH
(d en H 3
cu e (d MH o
S 0) 0 O 02
CU
3
-H
4-1
C
O
O
13
-------
4-1
G
CD
0
U
>i
--. 0
13 (8
0) !H
H 8
•H 0
4-1 03
G C
O H
U
— ' 4-4
O
rH W
CD
W U
pn jj
3! o
E~< e/3
^
0
SH
CD
4H
0
CU
o
*3
o
W
4H
O
C
0
•H
4J
3
rH
O
W
0)
SH
i-H
18
SH
O
ft
CU
4-)
O
G
SH •
0 C
0
0) -H
4J W
03 W
S-i -H
3 g
O CU
o
03
C
H
03 0)
4-4 SH X} •
0 O -P g en tn 13
-P U 03 13 C G
t« 4-1 CU .G O -H 03
T3 M C ft O rC SH SH
cu SH o cu en 4-> SH tn CD
13 0 SH g • CU CD -rH 4-i X) 4-1
CU SH 0) SH d) rG g 4-1 C g 0
GO 3 '(t) 5 rH 4-> X! 4-1 OS
•H TS in • 4H r8 d) u u c -P
(j}4-lCDOiSHSH4HO O4-) 03 CD 03
1) OS 4-> CU d) 0) O -H (8 CU C E 13
tn G oS g ,0 o ^ 4-*3 SH • O CU
W-HrH g g W 4-1 >< tj1 W SH CD
CDgO)rd(0034-l-H H 0) 4H i-H tn 3 >
U-rHrHCrGrGCcn 03 13 • O rH C W -H
OrH 03OUCDC GOJC 03 -H 03 CO
SH W !H gCD (8 C C7> G S SH CD G
ft d) .-H CD 13 d) -P -H-HO SH B • CD
rQ O rC G SH C •en en -H SH 3 CD 4-i
rH •gSH4->r33-Hg3Oa)4->d>U 13 SH X
03 d) 03 4-1 OeO d) O 4--^ T3 ft rQ U G 3 CU
O-PrCCCSHOSrOPS SH g o 03-P
•H 03 O 0 'H tn CUCencT'CUrHOOj OS CD
g SH I u ,VgoS>i-iH3CU[nrGeo i-HS-i M
CU3tT> OU tnXjTDXJ^^C OCU 0
pGOOOtNdOG g a'0 ° ^ft 6
(0 W -HO 4-1^1304-1 GCU SH
4-i S4 4n G SH -P -H SH -iH U G U • O -P O
OC04HOO-'HO34-IOCU OjtnOjWU 4-1
•r-IO X!03 CSH 4-1CDCU M
>i OWtnCU-H-HCUCDOGOSHOCUCU -P
1) -HG 4JrHCU4-iSHrd4-iCnOO3 03-PXj ft
3 PH O OS CD -H 03 -P 03 03 O ft 'rl i3 ' — 1 4H 03 H CU
4J -H 3^4-l3tnSHSHg 4-1003SH30) O
tn -P 4-i tr* CD -H tj^ -H SH 3 O SH ft SH o 3 tr1 rG X
03 03 cu i — I SH cu 13 -H u ,G 0 SH ft T-I en cu tj cu
'O^GO'Ci 313 Ul O g 13
d) -P -H (8 ftos oJG en cu oi ><
34-iG gG GOISrG G 13
Cd)-HH HH HS iflC <-) H CD
•H M -P g
4-1 3 C CD
G en oi SH
0 C 3
O d) D1 0
^
-P
tn c
G cu
0 -H
01 -H O
en SH -P -H
G 3 u cu 4n
>i 0 -P 0! ,G - 4H
C i-H -rtoi 0)4-ien3
O CUC-P'SH SH4-)eO
•H 4-1 -H oJencu 1 cu c
en oi G-PftGeo-H3-H
§g 0 tn rG 13 13
CU C 03 4-) 0 • 0 CD
g o -ri SH 4-) -P rH -H SH en 4-1 -H
01 US-ICUW CUU^ft-P 4H
03SH4->CrGrG Gd) -H
4H 3 CUO4->'-Pl3en*CU3 U
O 4-1 U T3 U -H G CUCUM-H13 d)
OU ^^Oft-HCUU ft
G GOrHOl 04-igU-P-Hg tn
O rti-POTC 3CUOS4-IW
•H encn 4-in34->fyd)rHftSH4H-rH >,
4-1 CDO)GSHGC3tnCDG rH
o! Oen d) oS3rC)4_| COOS- d)
u i3tn gG4-* OUOU^en -P
•H CU -rl O W S-I Cn -H -H U CD OS
4n gej SH-HCOCcn4H4->Ud) -r-i SH
•H tn O d) -P O -H tn -H U -H g T3 3
SH -HSn p^UUGftCUUoSSH 3 U
CD Cft XO! •HgGQ)Q)4-iQ)4-i U
>• OS 0)0)CUOS3CUftSHCU,Cen OS
Cn^irH SH4-l4->rH>tn g-P C
O d) U 03 C (t3 SH -H •> O C 'H 4-1
C-H d)U •rHUSHCDC-PO'-'HCO ^
tn g -H -H U -H o! 4-> tn rG -H -H eo en en
SHO g-en4H4HC-P • u +1 CG-H
OrH Cd)d)d)-HO3enGCUCU-iHWOS OOCU
0 OXJSH-rHU 0)CD> OCUO -H-rH£
-H -H 4-14-1
-PO P rGGftCd)Ud)-PtneoU4-i -H03C
03 rG cj CU ft -H eoOrCoSSnoSeo 03 -H 13 SH O
34-> (8 CO U] (8 •rH4-'3CirHO!rCirHrH C-P-rH
tfCU d)OO4-i4H4-)-r-iejld)CUrHC3d) OCeO
dJg SHrGgSnOOSCUCUSHSHUOJU^ UCUSH
13 4J-Pd)THrt3 U UCU
03 ' — I03USH03 03 r>^Cr>
G 034-iGoiC C SHOG
H UUCUD>H H 03UH
•H 0) .C 13
e rH P C
0) 4H 3
rG d) C O
CJ SH -rH ffl
14
-------
little influence on the solution. In such a case, effort at reducing
the uncertainty in the parameter value may be unwarranted. On the other
hand, small uncertainties in a parameter to which the solution is quite
sensitive may have a large impact on uncertainties in the predicted
concentrations. Thus, both uncertainty and sensitivity must be considered
when the accuracy of the atmospheric diffusion equation is evaluated.
Finally, we note that discrepancies between predicted and measured
concentrations may arise not only because of inaccuracies in input
variables but also because concentrations are measured at a point,
whereas the AQSM predicts spatially averaged concentrations. Measure-
ment errors may also, of course, contribute tc discrepancies between
model predictions and data.
Although the validity of the atmospheric diffusion equation cannot
be established without question, it is generally accepted that the
equation is essentially a valid description of atmospheric transport,
mixing, and chemical reaction processes. The major source of invalidity
is the eddy diffusivity representation of the turbulent fluxes. However,
as long as the eddy diffusivity functions used have been determined
empirically under similar conditions to those to which the equation is
applied, the equation should be considered valid. The principal problem,
therefore, lies with the question of accuracy, namely the effect of
uncertain specification of input parameters on the predictions of the
model. It is primarily this issue on which we will concentrate subsequently.
Therefore, we conclude that:
• The atmospheric diffusion equation is essentially a valid
description of atmospheric transport, mixing, and chemical
reaction processes. The principal problems associated with
AQSM utility are related to the accuracy of the equation,
namely the specification of input relationships and param-
eters .
15
-------
CURRENTLY AVAILABLE AQSMs FOR PHOTOCHEMICAL OXIDANT
In this section we summarize the salient features of several AQSMs
currently available. Since the question posed at the outset refers to
AQSMs sufficiently validated/evaluated and appropriate for use in designing
urban oxidant control strategies, we restricted our attention to AQSMs
currently available.
Summary of Available AQSMs for Photochemical Oxidant
Table 2 contains a summary of the four AQSMs to be evaluated in this
report. Each of the AQSMs in Table 2 can be viewed as evolving from
Equation 8. The SAI model is based directly on Equation 8 with no
further simplifying assumptions. The LIRAQ model is based on the
equation obtained by integrating Equation 8 with respect to the vertical
coordinate z from the ground to the base of an elevated inversion layer.
(Prior to this integration, assumed vertical concentration profiles are
used in Equation 8.) The DIFKIN model is that form of Equation 8 obtained
when one transforms to a coordinate system moving with the mean ground-
level wind velocity and neglects horizontal turbulent diffusion and wind
velocity changes with height (4). The Bell Laboratories model is obtained
by integrating Equation 8 over all three coordinate directions to produce
one (or more) well-mixed cells.
Chemical Kinetics
There are essentially two approaches that have been followed in
developing kinetic mechanisms for photochemical smog:
Lumped mechanisms - Mechanisms in which organic species are
grouped according to a common basis such as structure or
reactivity. Examples include the mechanisms of Hecht and
Seinfeld, Eschenroeder and Martinez, Hecht et al., MacCracken
and Sauter, and Whitten and Hogo (5-9).
16
-------
E-i
2
Q
H
X
o
^
rij
u
H
*3
W
33
C_)
P
O
33
P^
PH
o
fa
03
S
Of
w
f£
H
f£
£>
p-C
r^"*
[-3
EH
2
§
PH
D
U
CM
W
i-l
3
a
E-i
W
01
0
C
01
CTi
^ * — . ••
^ vO r°"- *•»•* ^-^
G) CTi CTi CTt 00
^ » ^ ^
o r-
CN kO
CM
O
r£j p^
0) O C ^ 1
to I>- rH"wwcnEfa'w§
CN OUCN XUCUU-CU
O < -rH 10 O C71 'H r-H *H *-* CU C ^
2 co -P cu CO i-3-H-P-PtngOrM'
33 C3 n3 M O rC tdWidCUO-PO O
-H g£3^ 33 gCg-PWCUPnO
K ij j O 3 W fc 'TJ O -H O 13 rX
p">MH&HC3 roCMM-tSHC"- •• -
o cd ifl"O o td(d>*-t(dci)cn'OcN(N
>H in ft njcricucoo
^•(tJ ^CJS^ ^WQJ^C) £^ ^O fC C^ 2J
fNJ P. [/} f^ f^ fSj C »^ (^J r^ r^ r^l K K
O £3>iU O-H-Hft-HOiw
2 - -H 43 O 2tW-P^---Pa>O~
CNMHO)^ cuuucu'd-HLncN
^O OTSt! »rHfdCN(dgrH-PO O
O CNi-HrH>i O ODCJCUO'dflCNO
2 33 0 fd ^3 2 ^"^ Vi 33 M co g 2 PH
03
-p
c
nj cu
4J 1 & CO II 1
1 34->4->g~ -P H -3 H
M IHU MiJUH C CUCIrHgldlHO) C 0)
g O 'O O "' i CU C! C CTi ,Q (3 -P ^i -P 'H 3 15 -P CD rH 0 td "H -P O CD
3 "H C^i C3 0 fd ' — 1 T3 O • 3 (d 03 ft *H O (d PH Q) -P r^3 CT1 -H
3 ' — ! W fd U pCl CU 03 O i — 1 ^ O CU r-H ,Q O 0 4-1 (d (d CU CU 'H
CMHCU-CUrH OJCU-P-P3^3 Old >id C7iUrH.O-HCU33X:T3
OMH-Htfl^rHT3gCJiCU(dMH-P ft "CCfiOH CUTIHSHM E-IQ
• rl MH -P Id (U 0) "OgCUMHCU rHCU-iHO-tJ'Cra>CU[dft CU g
T3T! C U-POO M-Hg MHidT) -H CUCUW Oi-H^3
CJ 'O QJ C/D J3 *rH Q ^«( -M 'oj 0 f^ C 'C 4-J *T3 f~^ tT' i™H 3 T3 M (0 -M * (IJ
CO o d S QJ ^O rQ -M ^\ ,JJ"^ C O ? Cl (ti QJ ^3 *^J f C Oi O 'O
10 -H -H Q) O -rl -H ^3DSH033(n
TScnfiO) OrOft-'H3CUiHU -H-H SHU grHCnCUOl-P 3-H ^
OOOgrH CUCUrH>a>T-| M^rHS-l CUCUrH -HrH ft-P T3
gg-rl CUMCU-PWOrO>2 idlrHCUM>-l CUTl'JH OCUidgCUTS
-P 03 rH t> CU SH 03 \ U3(d 1 I>OCU-POidCUCU ^rj-P-PCCO^O
'O (ddCUCU^rdl-P *• .^ ^U fa F^ P'OCCCO -H i — 1 pC
•rl 0)>rHld l£>irHj3'-~C! rH-P CU 03 CUOC3-r-(IC3CO-P
(HCUgcuirHcnrooid-PnnJ id ,C,C -P JI-H-HO oo-Hnso3cu
C^^-'HrH'O A! CUO-HCTiM MldO-P.C'-P'OSM>i'3< ^ g
-P-OlfigO 33-HS-^U O (fl ><(dH 0) OIH HUCU
H 1 T3 3 O HJ • M ftCW>!rH-PHCUrjH -H rdCUSnCO
fd4H CUCJ O-P-P-PCUCU03 g ^t g P|H 43CU3CU-PftftJ-Pg • i — I g ftJ *
CO CU3 M-P 034H t~] g ft ^ Q QJ K"rHU(d GfdMrH 'H M
O >HOCTiO O-P3CU-P -PCO'ft-rHrH rHCUO OrHg0103(d
•HS^3SH XigrH 2-P-H>, CCOO-PO^CUi-HTSM-iCU-H&iCCU
WOHtng Oldg 03SHM CUOrH-PuftO^-PWCUTHUOl'OOCJ
C-H OCUJ-I O O-PXi-HH CU 03 C-PC 0-H
QJ4J g^H^IIUHCOM 'rHCT^CO -P-POltJftGQJOOCUP^TJ "-Q-P4H
g3'OMH-P rH4Hg(dr-H-H (dQC3W^3 -P-Pg-H CUg\(dO
•rHrHCM CflrH 03C(dg 03S-I IdCU-PG -HgT!> fd -P EH O -H CUMO ^U3T3CO-H003OI (d 4H -H
CU i — 1 (d CD -P -H 03 T(J -H ,13 O C/3 T^ UCT1'PCUCUCU3-P^-PCU i — 1 r£3 C3 CT1 CO
M fd 3 K* '"H 03 -H ^H O Id ^ T3 CUC3 -P0303X ^J-Pf3 *' ' ' — 1 CJ '>H C3 JH
ri3UCT('HgrO-HMCUCUM3HC3 MOMHOidO'HtOM fdCCUCUCU-HCU
E-l-HCU>-lG)
03
c
O '^*» r>^
•H H CU M
-P X! H
>i ft C3 (d -H Id i-l
w
-------
f — s
•a
0)
3
C
•H
4J
G
0
U
' —
•
CN
w
%
EH
W
0)
O
c
0)
SH
0)
m
0)
w
OJ
•H
O
0)
ft
w
s
w
icjj
MH
O
C
O
•H
4-1
ft
•H
SH
0
w
0)
Q
ft
3
O
SH
O
, — *
^D
—
*•
M
0
*.
CN
O
*.
O
MH H
o fd
o
G c
O O
01
05 3
MH
G MH
0 -H
13
13
0) rH
U) 05
03 O
n -f-i
4-1
rH SH
0) tt)
13 t>
O
E 43
O
£_| ^(~]
0 15
O G
(U -H
•n
o3 SH
SH -H
EH 0)
i — |
fd T3
•P G
G rt3
Oi
£1 ^
G O
0 SH
M 05
•H 0)
> w
C 0)
W &
,-^
j — i
o
r-H
in
K c
t/1 O
CH ,-Q
O SH
in U
03 0
0 SH
0 13
H >,
15 43
£>S|
43 0)
[>
-rH
> 4-1
•rH U
4-1 03
U 1)
03 >H •
0) C O
SH 3 O
T3
0>
4-1
1 1 ^.^
•H SH
6 ft 05
(0
01 MH
C 0) O
O 43
4-1 g
U 0 3
03 -P r-H
0 G O
SH -H O
^i
tn
0 —
H &
0 H
43 r*i
O H
0) O
EH —
ft
01
4-1 0)
01 SH
1 05
rH 01
C
0
• -H
05 4-)
4-> 05
flj 0
13 D1
0)
13
G tn
•H G
S -H
G
0) SH
O 0)
05 >
MH O
SH O
3
01
•
E E
o 01
SH -H
li i (— i
T~l r-l
ftj
13 43
0) O
4-> 03
o5 E
H
O 0
ft -H
<. I I)
0) 0)
-P G
G -H
•rH ly*
H
OJ
•H
4-1
C
0)
0)
MH
MH
•H
13
^"1
M
fd
a
•H
t^
j^i
o
MH
O
E
M
O
a)
4_j
C
•H
c
0)
4J
4-1
•H
^_l
|5
G
0
•H
4-1
fd
6
•H
X
O
ft
ft
05
-
d)
13
03
rd
tn
•H
01
3
rQ
OJ
r-H
O
01
13
G
03
01
£
O •
•H 13
-P O
(0 43
3 4-1
a* CD
0) E
, — ,
CM
N —
1
0) 0) »
tn C 13 fl)
GO) H 13
•H rH 05 >1
3 ft 0) 0)
rH O O 13
O SH 05 rH
C ft 05
•H - G
(DO
» n 13 -H
01 o >i ft
0) 43 O
•H * o SH
O CN 13 ft
0) O rH
to go*
SH 13 CN
v£> O O >i O
•
1 H
(U 0
•H rH G -H 01 -rH U W
0) -H OS O 4-1 3 G
13 O 4-> 05 O H 0
G 13 0) rd O -H
•H g 0 T3 ^SHG-P
SOX 0) 0) -H 03
SH -H C 13 -P 3
MH g O >i G 01 W
43 -rH H 0)
01 |? >i 13 0) ~— O
0) O H 0) 13 W H
•H rH 0) 0) r-H O 03 •
SH MH 3 oS o3 • SH 'H *-~»
0) O 43 SO 0) -P O
01SHO) M
Q)0)GO)rHGo5 0)
3 ft -H -rH SH
TlrHOl OO ftt! 0301
H O03 G -
X K* O 4- ft SH SH 'H SH
•H -HO) ft 13 03
g ^ 01 G MH 0) SH CD
1 43 01 -H O 13 0) O O
r-t -H ,X G SH
i — 1 13 E K^I o5 MH tn I>i
0) a) w ft SH G 43
^ 4J CD 4-1 - G -H
G -P 01 0) 0) G t!
0)0) • to -H >d 0) in O
0)C/lr-H 1 E>lSO)>
S-l 0)r-iroO)43-P >rH
43SHCi)^4^ 0)0) OO
EH ftUrHOl343OW
1
SH
rQ
03
h3 W
0)
r-H -H
rH SH
0) O
ffl -P
•
G
0
•H
4-1
03
•H
X
O
CM
O
tn
G
•rH
O
•H
SH
O
w
-------
• Surrogate mechanisms - Mechanisms in which organic species in
a particular class, e.g., olefins, are represented by a single
member of that class, e.g., propylene. Examples include the
mechanisms of Niki et al., Demerjian et al., Dodge, and Gradel
-t al., (2, 10-12).
Of the four AQSMs in Table 2, three employ lumped mechanisms and one
employs a surrogate mechanism. In general, the surrogate mechanisms
tond to be more lengthy than lumped mechanisms because within a surrogate
mechanism each individual species is treated as a separate chemical
ontity. For this reason surrogate mechanisms have not found wide utility
:n AQSMs that have substantial meteorological treatments because of the
computational requirements associated with calculating simultaneous
chemistry and transport.
In this subsection we wish to compare the kinetic mechanisms in the
SAI, Livermore, and Bell Laboratories models. We do not consider the
mechanism in DIFKIN since it is merely a single hydrocarbon special case
of the other two lumped mechanisms. Our object here is to ascertain if
there are any substantial differences among the three kinetic mechanisms.
Table 3 presents a comparison of the original Hecht/Seinfeld/Dodge
mechanism, that contained in LIRAQ, and that contained in the present
version of the SAI model. In Table 3 we list all the reactions that
were included in the three mechanisms. If a particular reaction is
included in a mechanism, then the rate constant value used in the
mechanism is entered corresponding to the reaction in the column cor-
responding to the mechanism. The issue of most interest here is which
reactions are included in which mechanisms and not the particular rate
constant value adopted. (The rate constant values in all mechanisms
continually undergo revision as new measurements become available.)
From Table 3 it is clear that, aside from rate constant differences,
the three mechanisms are very similar, even though the interpretation of
the lumped organic species varies somewhat among the mechanisms. Dif-
ferences in rate constants are the result of choices from among avail-
19
-------
TABLE 3. COMPARISON OF LUMPED CHEMICAL KINETIC MECHANISMS FOR PHOTOCHEMICAL
SMOG
Reaction
NO + hv -»• NO + 0
2
0 + 0 +M-+0 +M
2 3
0 + NO -> N0_ + 00
3 22
0 + NO + M -> NO + M
0 + NO,,-* NO + 0-
2 2
0 + NO,, + M -* NO- + M
2 3
0., + NO- -* NO- + 00
3232
N00 + NO -* 2NO-
3 2
NO + NO,, -* N.,0.-
3 225
N-0C -* N0_ + NO,
25 2 3
N-0C + H-0 + -* 2HNO-
252 3
NO + HNO., -* HNO- + N0_
3 22
HN02 + HNO -> H20 + 2N02
NO + NO- + H-0 -> 2HNO-
22 2
NO^ + NO^ + H^O -> 2HNO^
232 3
2HNO -> NO + NO- + H-0
2 22
HNO + hv ^ OH + NO
OH + NO^ ->- HNO^
2 3
OH + NO + M -> HNO- + M
2
OH + CO -^ CO + HO,,
2 2
HO + NO -> OH + NO
HO^ + NO^ -> HNO,, + 0.,
Rate Constants
Hecht/Seinfeld/
Dodge
VAR
2.0xlO~5
20.8
.0035
13800
.0022
.046
15000
4500
15
-5
10
.00025
0.2
-6
2.1x10
4.5
VAR
15000
12000
250
700
—
@25 C (ppm, min units)
b c
LIRAQ SAI
VAR VAR
2.0xlO~ 2.0xlO~
23.6 25.2
.0034
13400 13400
.0039
.047 .05
13000 13000
5600
12
-6
6.8x10
— — — —
—
-9
2.2x10
-3
1.66x10
— _ __
VAR VAR
10000 9000
10000 9000
210 206
290 2000
44 20
(continued)
20
-------
TABLE 3. (continued)
Reaction
OH
H°2
HCl
HCl
HCl
HCl
HCl
HC2
HC2
HC2
HC2
HC3
HC3
HC4
HC4
HC4
HC4
HC4
HC4
HC4
HC
+ HNO -> HO + NO
+ OH -v H20 + 02
+ 0 -»• RO + aRCO +(l-a)HO
6. J ^-
+ 0 -> RCO + RO + HC4
+ 0 •> OZONIDE
+ OH -> ROO + HC4
+ OH -> ROO + HO
+ 0 -> ROO + OH
+ OH -+ ROO + HO
+ 0 -> HCO + HCHO + OH
+ NO ->• PROD
+ 0 -* ROO + OH
+ OH -> ROO + HO
+ hv -» 6ROO + (2-6) HO
3 22
+ OH -> CO + HO + HO
+ 0 -> OH + HO + CO
+ H02 + H202 + RC03
+ NO -> HNO + RCO
+ N03 ->- CO + HNO + HO
+ RO -»• ROOH + RCO.
Rate Constants @25 C (ppm, min units)
Hecht/Seinfeld/
Dodge3 LIRAQ SAI°
19
29000
6800 4400 5300 (a=l)
.016 .018 .01
(^HO +RO+HC )
.005
25000 23000 38000
14000
107 73 37
8000 2900 8000
.002
50
65 — 20
3800 — 1300
VAR VAR VAR
23000 14000 100000(6=1)
7300
290
.0048
.22
0.5
.004
(continued)
21
-------
TABLE 3. (continued)
Reaction
Rate Constants @25 C (ppui, min Units)
Hecht/Seinfeld/
, a
Dodge
LIRAQ
SAI
HC + hv -> CO + H
ROO + NO -> RO + NO,
RCO + NO -> ROO + N
RCO + NO ->- PAN
-J ^
RO + 0 -»• HO + HC4
RO + NO
RONO
RO + NO ->• RONO
HO + RO ->• RO + OH + 0
£. £
910
910
100
.024
490
250
5300
100
2700
1300
43
.0036
74
74
8200
98
VAR
2000
2000(R=H)
150
4000
4000
ROO + ROO ->2RO + 0,
100
98
PAN -
+ H02 + HC(0)OOH + 02
>• RCO + N02
10000
.02
ain the Hecht et al. mechanism HCl, HC2, HC3, and HC4, refer to olefins, alkanes
aromatics, and aldehydes, respectively a represents the fraction of the total
olefins containing terminal double bonds, and
of the total aldehydes.
is the formaldehyde fraction
In the LIRAQ mechanism HCl denotes olefins and highly reactive aromatics, HC2 -
paraffins, less reactive aromatics and some oxygenates, HC4-aldehydes, some
aromatics and ketones.
°In the SAI model, a 31-step mechanism, called the carbon-bond mechanism, has
been developed as a variation of the Heoht/Seinfeld/Dodge mechanism (9). Because
of the association of reactions and reactivities with carbon bonds, the range
of reactions and the range of rate constants in a kinetic mechanism can be
narrowed somewhat if each atom is treated according to its bond type. In this
mechanism, hydrocarbons are divided into four groups: single-bonded carbon
atoms, fast bouble bonds (i.e., relatively reactive double bends), slow double
bonds, and carbonyl bonds. Single—bonded carbon includes not only paraffin
molecules, but also the single-bonded carbon atoms of olefins, aromatics, and
aldehydes. Double bonds are treated as a pair of carbon atoms. An activated
aromatic ring is considered as three double bonds in the present formulation
of the mechanism, and because of a similarity in reactivities, aromatics are
lumped with the slow (ethylene) double bonds rather than with the fast double
bonds. In the mechanism HCl, HC2, HC3, arid HC4 represent fast double bonds,
slow double bonds, single-bonded carbon atoms, and carbonyl oor.ds, respectively.
22
-------
able rate constant values and more recent determinations as well as the
result of different lumping schemes. We conclude that:
With respect to the representation of chemical detail, the
reaction mechanisms in the current versions of the SAI and
LIRAQ models are essentially equivalent. The mechanisms in
DIPKIN can be viewed as a special case of those in the former
models.
The next point to consider is the comparison of the Bell Laboratories
mechanism with those in the SAI and LIRAQ models. As noted above, the
Bell Laboratories mechanism is a detailed mechanism based on the chemistry
of propylene and the three lowest aldehydes. For the purpose of this
section, our interest is to ascertain if the Bell Laboratories mechanism
contains any fundamental chemical details not present in the SAI and
LIRAQ mechanisms. In the inorganic chemistry, the Bell mechanism includes
0(1D) chemistry and free radical (primarily HO ) scavenging by aerosols,
not included in the other two mechanisms. (We discuss these two processes
subsequently.) With respect to the organic chemistry, appropriate combina-
tion of free radical reactions in the Bell mechanism can be shown to
lead to ths SAI and LIRAQ mechanisms. We conclude that:
• With respect to fundamental chemical detail, the Bell Labora-
tories mechanism, while more explicit than the SAI and LIRAQ
mechanisms, is equivalent in nature to the latter two mechanisms.
The Treatment of Meteorological Variables
Table 4 summarizes the treatment of meteorological variables in the
four AQSMs under consideration. In Table 4 the methods of treating
advection and turbulent diffusion are summarized separately.
Initial and Boundary Conditions
The initial condition for the atmospheric diffusion equation is the
concentration field at the time corresponding to the beginning of the
simulation, . Simulations are normally begun at night or
23
-------
H
2
Q
H
O
^
u
H
s
8
q
o
•r-t
in
3
4-1
4-t
•H
Q
4-1
C
•8
p
•rH
3
4-1 CJ
TJ rG <0
>i >O
T3 CJ X!
CJ in in
4-13-0
rH C CJ
O ~r"l CT> O
•H O
0) in
.C g -H
•O P O
14-10) M P
o 4-1 in in t, -H
3 0) -O rH -O
Q) ft 10 M 5
0
O O -H
H 4-1 SH
a
3
M MH M
G P
>•* >* ^-^ n5
O 43 S 0) -H
•H O i-l P
I
O
m a a) iH s o QJ
E >i CJ £ -M
3 CJ 3
(1) O
•-I ft -O
oj(d-H,c;
4-1
D1 >i -
C
in
C 4-) Q) C
O O O W 4-) O
W 4-1 "O
OTIXION
> M CM -O rH OCJCJtncIGCJCJ
4J
0)
(0
ft
4J
I
O
Pn
cx
2
H
w
H
>
U
H
O
o
§
o
w
p
ft
CJ
Q
tr>
•H
X
in a)
•rH -O
-H
-o s
r-1 I
CJ 10
•H CJ
ft G
03 -H
I
CJ
3 in p
in in P ft
3 "
3 CJ
P -d aJ
(0 G g
M O
cj en n
ft o a) o P
ft -a
•H <1)
-0
-a a
M
CJ CJ
. 3 ni c >.
!-l O ft •<-( (0
3 -H
•H P P CJ
4J OX
X ft O CJ 0 (0 O
•H O ft rH U P
S rH r4 -«H (0 in
qj 01 14-1 p -o
dJCUO-H
4J (U
CJ ft I
P
I CJ
CJ
C
4J
fi
o
T!
I
X
CJ CJ
CJ CJ
>-l I ft •
'—(DOT!
M ^ C rH CJ
&
§
w
E
H
H
O
•H
4-1
U
CJ
•o
OJ
•o
o
G in cnrO>CJft -rHO3in oois,Gtn S
CJron3(0OG UfdG-P U^N-om -H PCJ
•HCTlEO OJ54-4OUOfiCJrHlO^(nM4J?'tJ'IrO
r0 ,--1 ^_J rj^ 4-J J*-j -M *iH O rG Q^ ^ (Tj CJ G ?3 ^^ "^
OJ t-H>i 3 SrH4-lEHEW'ag>C)PP"-'CO
CJ^raa)-Hn3> OJOftCJ GOrdnO<0 ' -H
^4-l4-4p4HT3-HCJgWlOO •rHCJOrHI^ft-H>irHU5(U
0) -rH 4-> r-H (tj -O
O CJ CJ -d J3 -H
ft & i ?
CJ T) rH
ft CJ
10
0 -O CJ
ft in E (0
C
MWOO
TI
4-1
a)
(i)T3i4-lrdC0
- G rH 4-1 -H -r-l
G (0 rG
P
rH -H H 01
O rH O CJ CJ -H
O
> O rn 3 4H -H -H rH -H CJ CJ N
-H m
CjOrGU(04-l(l)4->^;tnft(04H
rHOWGPCJ C-rH
(0 -p >i V4-| (N(d P
4-1 CJ OJ MH O U -H
Enft
Q) C O
m c
,G>iOr4M-H,q(i)Cot7'
24
-------
o
in
MH
•H
o
•P
0)
>1
4->
-,H
a)
N
o
•rH
•r-l
M
(d
>
a
0)
G
(d
•iH
U) 4->
10
-H
3
u
(d
•rH .rH
H
0 • I I
MH 0) 0 ft W di
T3 0) -H ft 3 4->
0) C MH ui >i id
(d -rl >i
O 4->
43 «
O> H •
•H i-l
0>
DI !
•H -H M
i d -P a ,
M C O
4-> 0) O
> i-l
T3 0)
0) O 4J
•H (d ft c R ;>
>i IH id o
r>M 4-> id -H rH
W id G *d 4-) (d
O
0)
. u,»3
(d H 01 01
4-1 O 0 Id i-4
0)4->4->O3Cr4ftV4O
o) «; o) o) JQ o -rH
01 Id > -H O -H S
0) G i-H-Hg 4-10)14-4
nj 4-> -H id m a) o
•rH W ^ 4 1 fll i 1 ^ |
pCj y jj* p-| U) ^> QJ
-M 0 Q) d) <1) 'dO
(D (1) -H O O »C >-H
3
tn
•H
§
c
o
^3 "H
JH I J
3 0) U
o -d G
M 3 3
tr> 4-) 4-1
•P g C (d
14-4 MH g id
4-1 in
•H 0) rH 0) G
0 H ft W —
OM
M-)
^c
C-H
•H O
01 M
4JO)
4-1
0) 0) 0) -H 4->
4-^ O *^ O *l l
'd
(1)
c
•H
-P
o
u
0)
Q
0
0)
4JrH
^
M id
ft 4->
01
•H
MO) -HOiriOCEGO-P
•d id
•rH ft O -H
- G O 4->
•H Id
01 <4H >W 43 E _
0) M O O 4J O
4->ft01ftngMHWoE-iGigu
M-P G 0)3 (dgH
0) E 0)
> O 6
i40)0
iHM>ftO'OO)E H
rd(Oi-HlH>IS MH O-H
I 0) O G -d
0) ft U 0 0)
0) 01 -rH
(d o) (d > 0)
(d i-l 01 H
4-> O 4-> 0)
T3 )H 0) • (d
43 Id 0) D^ 01
W
•r)
X
•H
O M £j ** Qi ^4-1 QJ ^4
T3 O 1) O OJ *O 4J CO fl,}
CPGU3 3C01H 0)
G-Hid4-> ^-P-H3Q43i
* 6 O 43 4J
G T5 01 G
•H 0) 0) 0) O
I W
C 4-> -P
•rH 01 -H C
4-> 01 0) 0)
H G E > I
M-l
OM 4JM
fa3 ODj
01 0) ft
id 0) -n (d
o) id 0) g 4->
O en
G
•S
G -d
•H
S G
H 0)
id 43
C
id
HC
(0-H
CW
O3CXl
OM
4J3
W OO(S
4->
'w °
C 0) 4->
a>4J
0)
C
I -H O
(d MH 4->
H 0)
-d O
(d 01 -iH N 43 O
> O -rH -H 4-> 4J
4-> O 0) 0) 1-1
0) 0) G 3, 0) 43 T3
> T) O D1 -P tT> C
.C4JO)
01 0) 4-> 4^
G -H 43 01 G
O
4J
O
0)
T)
•H -rH (d
w id
G C
4JO)
43 O £ rH
T3 Q
00) —?
4JSC AT!<4-(0)a)CO
rd G O -H 01 4J 0)
0)0(d
w tr> 0) 3 -P -rH >
U 4J
•£,
•rH -H'
(da)G
ftg<
4J ft
MO)Hj5
34Ja)4->
tniH
0) rH M-l IH O tH
(UOIG'OO O
'a?4-)THO)Ofta)
X
0
ft UH -rH
ft 0)
(d T) 01 o) !H T) G
DCrH rHftO)4-)»H
aO> 0> 4-> 3
U> O rH G M -P
id -H o> id
G
4J 0)
•H g
0)
0) M
W 3
4->
id
0)
5
-Ho
O)T-l
o> 4-> id 0) ft
C34JO)-r-(Oa>GlH(dfl)>(T!
ft -n en 4-) id
Id G W rH
C H -H 3
O 4-> 0) T! O
•rH P G -H
4J 0) -H 4J
id > 0)
•H G 01
0)
T3
i
4J
C
H 0
< U
w ^
Ol
H
Q
25
-------
,-v
CD
D
C
•H
-P
G
0
O
*^-*
^
w
E-i
C
O
•H
tn
2
4-4
4-1
•rl
Q
^j
G
0)
rH
X!
M
|JH
.c
P
(U
Q
tr>
G
•H
X
-H
s
0
•H
P
O
0)
•O
, 1
0)
•o
o
s
T)
(0
>i
4->
-H
rH
-H
Q
rt)
p
w
•
t3
CD
C
•H
a
4->
(0
•^
O
G
•H
•d
CD
p
P
iJ
en
C
o
CD
irt
en
o
•rH
M
o
4-1
CD
10
z
H
NH
t.
H
Q
T!
0)
rQ
3
rH
U
q
•H
P
o
«
en
CD
• r-,
4->
^
0)
0
V)
CD
U
It)
4H
^1
3
>,
0
P
0
-j P Qt
H > M
P G 0) -rH
en -iH x; nj
• 0) P
03 J-l (D C
O Cl) ft O
•H G S 4H 4H
P -H 0
en S nJ -H c
t~] p -H
P P G .G 0
f^ "H tr* ni
0) 0) -H i-( •
g JM T3 Q) P P
0) 4-i eu § c
3 en 3 0) G g
en 0) o -H d)
10 en 0) C S >
1
G
ID
P
in
C
•H
en
•H
rH
, — 1
CL)
o
Ifl
G
•H
c
o
•rH
CO
3
4-1
4-1
•H
Q
T!
0)
10
M
>
(0
CO
"H
4J
QJ
0)
g1
•H
j^
•a
rG
^
(U
t~{
P
cn
0
10
o
rH
4H
t!
c
•H
|j
-
r^
^J
<
i-l
J
W
•
en
3
o
d)
c
m
4J
ti)
^
f~i
4J
rH
C
O
C
Ifl
r~H
rH
O
o
Ifl
Q)
>
0
•O
CD
g
en
ifl
en
en
rH
rH
U
0)
^fj
4J
ll 1
w
•H
•
T)
^1
•r(
&
0)
rH
S
(ti
W
tt)
rH
, — |
Ifl
•rl
ft
in
•H
^1
O
0
rH
O
•H
p
>H
>
•o
Q)
rH
•rl
(0
0)
H
'O 4-)
m o
c
's -p
M C
(U 0)
I) g
r$ 2
0) tn
M ftJ
0) 1)
P 10
(0 ^
rrj .,_)
rH
cu
f, 8
H >w
CO
G C
•H 0
•H
T3 P
10
^1 *H
O V4
ft >
Q) T3
CD
en cu
w
G
0 T3
3 !S
r-l
o
CO •
0) rH
M 0)
T!
"io i
II
C
•rl
G C
CU -H
•P X
0) g
O 4->
Ifl *H
0 5
Q) a
i— i .p
cti n]
0 G
•rl -H
aJ o
> 0
i
X!
Tl O
0)
G rH
•H (ti
4H G
rQ ^
Ifl
0) Q)
^H en
ifl
>i O
it)
13 en
C
A -H
P T3
CD
J_J ^
3
O 0)
A A
3
0 g
g
O
MH
T3
(D
•rl
p
^Q
0
(0
TD
P
"2
*
CD
rtj
it)
Ifl
0 >
ft t)
G
O -H
t! rH
O ft
26
-------
at sunrise, and the field at that time is constructed from the
1°
station readings. A ground-level interpolation routine and assumptions
regarding the vertical variation of the concentrations are required to
generate the full field from the station data. Because only
surface readings are generally available from which to construct a
field, we expect the most uncertainty in the initial conditions
io
aloft.
The boundary conditions for Equation 8 consist of the concentrations
upwind of the region, the pollutant fluxes at the ground (the source
emissions), and the flux condition at the upper vertical boundary of the
region. Concentrations upwind of the modeling region can be estimated
if monitoring stations exist at the upwind edge of the airshed. In such
a case uncertainties in these concentrations will be low when a previous
time is simulated. The major source of uncertainty in boundary conditions
generally arises at the upper vertical boundary. First, the temperature
structure, for example, the height of the base of an elevated inversion
layer, is not known precisely. Second, the pollutant flux condition at
the boundary is also not known precisely.
Thus, the major uncertainty in boundary conditions lies in specifying
the upper vertical boundary conditions, both the location of the boundary
and the species flux condition at the boundary.
For each of the four models in Table 2 it is necessary to specify
the initial concentrations of all species. For the grid models (SAI and
LIRAQ) boundary pollutant concentrations must be specified at each time
increment for all grid cells that lie along an inflow boundary from the
ground to the upper vertical boundary of the region. For a trajectory
model (DIFKIN) only initial concentrations need to be specified. The
usual approach is to initiate the trajectory in the vicinity of an air
monitoring station and use the observed measurements as initial concen-
trations in the lowest cell. As in all cases where vertical variations
are included, the initial vertical distribution of concentrations must
be assumed.
27
-------
Source Emission Inventory
Table 5 summarizes the source emission inventories required for the
four AQSMs by spatial, temporal, and species resolution.
SOURCES OF UNCERTAINTY IN AQSMs
In the previous section we summarized the key elements of four AQSMs
for photochemical oxidant. The object of that section was to delineate
the different treatments of the major physical and chemical processes.
Mow we proceed to consider the principal sources of uncertainty in the
specification of these processes. In the following section we will
attempt to estimate the levels of accuracy of the predictions of the
four models given the prevailing sources of uncertainty.
Chemical Kinetics
The assessment of the accuracy of chemical kinetic mechanisms for
photochemical smog has received a considerable amount of attention (7,
9, 10-13X In this subsection we consider the two most important aspects
of this assessment.
• What are the major sources of uncertainty in current mechanisms
(i.e., those in the four AQSMs of Table 2)?
• In reported mechanism validation studies can we draw general
conclusions about the performance of the current mechanisms
when simulating laboratory smog chamber data?
The levels of detail of the chemical mechanisms vary significantly
among the four AQSMs. Each of the four mechanisms has been shown to be
capable of representing the general chemical features of the photochemical
smog system. The 12-step mechanism in DIFKIN is, however, too simplified
28
-------
c-1
Z
C
o
EH
C
W
Oi
O
En
a
H
D
Oi
0
4J
3
tH
0
W
0)
01
D
-H
O
0)
ft
c
.3
4-1
9
05
a
T)
0)
- M
I I (00)
0) 01 rH G t)
OiO-P
u o
o
T) M
CT,^ D g •• • ,c M-I 4-1 a)
rH c >( M H ,G ,c
(0
C C
O O
•H -H
O rH 3 -H .
20 4-> ft
0) O
014-I »MH. C
O O XI 3
•H 01
4-» 13 iH
U .C 0) rH
O G
0)
Cn £)
C 0)
O .G
•H 4->
4->
•H
4->
O
fd
(U
(UMtO1*-!
-O C H -H
0) M O 4-> rH O *
tt)T33H>M
M
3
•H O rH -H
6* -P a) > 51
7>O
'01
M -P >i
aG C H . ,
0) -H M 0)
E -H 3 M M
0 E O G -H
'3
rH-P
4JC(a
O
I -H I
01 01 (1)
•H ft 10 -O
E 0) XI rH
a) 4-i rti
.-H.pC
•H 0) -H UH
MX 0)
-MO
01 W
0) o •
e E u -H 01
01 (D ffi 4J D
4-> 4-> <8 10
0) 42 T)
M 4-> •• C
(0 -H (0 Id
rH 3 01 0)
O M M (d
MM O W
IM CMC
01 a) OX
G O O O rH
O M CNM m
01 O >i «•
01 oi - fi (n
•H X 0)
E >l O MH C
01 M Z O 1 C - 01 rH
M o o
O
O
W
H
a
o
EH
§
K
U
D
O
W
G
O
•H
0)
10 rH
0)
O
4J 0)
O M 0)
10 M
•a
4)
•rH
IM
CtT"H
>1
I E D 4-) O <0
3 H M X IT! -HO
rH 0) O -H T3 10 4->
IT)
H
O
01
s
•ri
IT)
0) Ifl
(0 M O 4-> rH
C D O O
-H -H 0)
01 E T) 4-1 QJ
•H fN 4J 0) 3
H CN > 4-> ft
0)O>i4JO
O _ O X 0 M
DOHOCnOCtl-PrHCC
Q<
I
<
W
H
hH
29
-------
c
0
o
w
0)
w
0)
•H
O
0)
ft
cn
0)
0) 0) rH
rH rH rH
w ft S tt) >i id
4J E Q T3 4-1
tt) (3 -H 1-5 3 0) •»
43 0) rH
6 IH
id o>
o id G o)
W -H - N
C 0) G O
a) >< o id ft id a)
E M - M o T)
0 O T! O » •• -H
o) c
^-accc
(N -rl Id4-l >i 0) 4J 0)
C -H MH -H - o id 01
a>oco>o)(d >
•H T-l
M E
w
-p 4-1 Id >i M 4-1
•H U 43 43 0) U
0) -
rH O T3
>1 I.
w
o o
43
o
id - o) tn o
in o ft tn s
2EG
oo
fl
O E >i i-H O m - T3
U) £C OH] Ol ^1 tC rC
E n c u
ffi*> D 43
0) 0) U -H
O rH 4->
>i O O
i-H Z X
0) U
0)
c
• H
4->
o
o
w
I-H
G
0
•H
-p
3
r-l
0
U)
0)
a
i (d c
0) 4-1 O 01
(ti -H
O G
-H O
4->-rti m ft
O rH 0) 0
OO-
-H -H >i
ld3O'H
OOJO
I
U)
(U C W
OWE
0) -rH
G SH 4->
0 M
•HOG
4-1 O O
id -H
•H 0) 4-1
3 4-1 G 0
0 -H -H E
3 3
G 0) 0)
4-1 G 4-> 4-1
c o
O -H o HH
U W 4-1 O
o
•H
-p
o
Ul
0)
id
•H
4-1
ft
w
w
I -H
0)
o o
•H E
4-1
O)
•H 4-1
4J
ft
in SH
MH -H C
43 Cn o tn id
-HiH
W 4-1 4-i 43
0) U
0) -rH
.£ 3 C
0) •
H T3
U
Id O r-l
D1 ,G
o oi .
>1 4H
rH nj q_| U-l
w
o) id
•rH 4->
•o
w G SH
3 O id
•H
G SH
•H
>i -H
4-1 E
-P tT< tJ*
u in
SH a)
O SH
O
rH C
0) 0) -H
O)
n3+JOCU
13 O
•H -H G O Id Bl
C SH O 43 > id
•M o -H id Q)
E 4-) 4J r-l 73
O 3 0) 0) 0)
OT30)0)rH43
U 0)
0) 43
•rH (fl M O 0)
D SH 0) G IH SH
CO-HOICD
0 w 0) ft
O
in a)
4-> 0)
M
id ><
0) E
3 4J nj 4J n3
K 3
G >i 4H Ul
O O
IH rH
•H 0)
-0)0) MGQJ
W ID Q) rH O -H
W U rH Id 0)
S • • 3 0) 43
G CP
•H G
HI S-l CTi 4-> 4-) X
a) a) 'H >d O' -rH -H
W 2 0 3 C
•H -rH O O 0)
s
Cn
H
Q
H
ffl
30
-------
for detailed quantitative predictions. The 143-step mechanism in the
Bell Laboratories model is more detailed than necessary, even if meteoro-
logical variations can be represented as simply as a few cells in series.
Thus, the lumped mechanisms in the SAI and LIRAQ models represent a
reasonable compromise between chemical detail and computational simplicity.
Moreover, these mechanisms appear to represent properly the known chemical
features of the system.
The most important inorganic reactions are common to the mechanisms
in the SAI, LIRAQ, and Bell Laboratories models. There is some variability
in the choice of which secondary inorganic reactions are included.* As
noted in the previous section on the chemical kinetics of currently
available AQSMs, the representation of organic reactions in the mechanisms
is essentially similar (see Table 3).
With respect to sources of uncertainty in kinetic mechanisms, it is
necessary to consider the performance of mechanisms in the simulation of
smog chamber data. The best characterized data available appear to be
those obtained at the University of California, Riverside (UCR). Of the
mechanisms in the AQSM, the most detailed set of validation studies
available are those for the SAI mechanism and the UCR data (9). Because
of the essential similarity of the SAI and LIRAQ mechanisms, it is
reasonable to discuss uncertainties in the predictions of both mechanisms
based on the performance of the SAI mechanism.
The overall smog formation process as simulated by present mechanisms
can be described in terms of two radical pools. One of these pools is
*ln studies with the Bell Laboratories mechanism (see the following section),
it was found that the removal of O^D) chemistry had a noticeable effect on
the predictions of the mechanism. Also, not included in these lumped mecha-
nisms is radical scavenging by particles, an aspect included in the Bell
Laboratories mechanism. In sensitivity studies, it was found that the
removal of the radical scavenging term did not lead to substantially dif-
ferent predictions. Nevertheless, it would seem advisable that radical
removal (principally HO ) be included in the lumped mechanisms together
with J1 ( D) chemistry.
31
-------
the oxygen radical pool; it is associated with NO photolysis and the
production of ozone. The other radical pool can be referred to as the
peroxy-oxyl radical pool. In this pool, radical transfer reactions
convert peroxy radicals to oxyl radicals and vice versa, with the con-
comitant conversion of NO to NO and oxidation of hydrocarbons. Oxyl
radicals are formed when peroxy radicals convert NO to NO . Peroxy
radicals are formed when hydroxyl radicals react with hydrocarbons, and
hydroperoxy radicals are formed when alkoxyl radicals react with molecular
oxygen.
A major problem with kinetic mechanisms for photochemical smog is
that it is difficult to reproduce the initial rate of hydrocarbon dis-
appearance and the initial rate of conversion of NO to NO . This dif-
ficulty is often resolved by adding some initial source of peroxy or
oxyl radicals in addition to those formed by the reactions of oxygen
atoms with hydrocarbons, for example, by assuming an initial concentration
of nitrous acid, which photolyzes and supplies the initial radicals.
Whether nitrous acid is initially present in the smog chambers in the
amounts assumed is unknown.*
Once the pool of peroxy-oxyl radicals is established in a simulation,
the radical pool must be maintained, because radical sinks, such as the
reaction of hydroxyl radicals with NO or peroxy-peroxy combination
reactions, tend to consume more radicals than are produced by NO
photolysis and the subsequent reactions of oxygen atoms. The radical
concentration is maintained in the mechanism by the photolysis of carbonyl
compounds (and, in olefin systems by the ozone-olefin reactions). In
some cases it is obvious that too many radicals are present initially
and that the maintenance source of radicals in the mechanism is in-
*There is evidence that nitrous acid is found during the loading of smog
chambers (14). The amount required to simulate UCR experiments was found
by Whitten and Hogo (9) to be generally about one-third of the equilibrium
concentration of nitrous acid that could form from the initial concentra-
tions of NO, NO and HO. Whether the walls of the smog chamber are an
important source of initi.il i.-t-Ucals is unknown.
32
-------
adequate. It has been speculated that the walls of the chamber in some
way supply radicals to the peroxy-oxyl radical pool. The effect of such
a process would be greatest when the concentration of normal radicals
was the lowest — in a low activity and low hydrocarbon experiment. That
the walls may be supplying radicals is supported by the similar need for
a high initial HNO concentration (relative to equilibrium). A second
problem, noticeable in most of the SAI simulations of the UCR data, is
the poor fit between predicted and measured NO concentrations. Ihis
X
problem is usually reflected by an overprediction of the NO concentra-
tion, although it could also be caused by NO losses in the smog chamber.
In studies with the SAI mechanism and UCR butane photooxidation
data, it was found that for many cases one could achieve a reasonable
fit of predictions to measurements by using initial and maintenance
sources of radicals within the bounds of uncertainty of the various
reaction rates. However, it was not possible to obtain a good fit to
measurements of systems in which the hydrocarbon/NO ratio and the rate
X
of hydrocarbon oxidation are low. In simulating a series of experiments
on the propylene/butane system, it was found that it was necessary to
lower the carbonyl photolysis constant for radical production below that
which had been established in earlier simulations. In toluene simulations
in which the hydrocarbon/NO ratio was low, the inadequate maintenance
X
radical problem arose. On the other hand, high hydrocarbon/NO experiments
led to difficulties because of overly rapid radical production. In
general, the mechanism does not provide enough maintenance radicals to
sustain the rate of chemical reactivity observed at UCR. In general,
the compound most poorly simulated is PAN. PAN concentrations can be
simulated in runs with high concentrations of NO and propylene. For
the low NO runs, however, the predicted PAN concentrations vary from
X
too low to too high as the hydrocarbon concentration increases.
In a similar study, Dodge simulated "dirty" chamber effects by
assuming the walls to be a source of propylene at a rate sufficient to
explain the ozone yield obtained in a chamber experiment containing only
background levels of hydrocarbon and NO (12).
X
33
-------
The following aspects of kinetic mechanisms have been trea..^c as
"fitting" parameters wherein adjustments are made within accepted un-
certainty limits to obtain as close a match as possible between predicted
and measured concentrations (see the Appendix for a detailed discussion):
• Initial concentration of nitrous acid;
• Quantum yield of carbonyl photolysis;
• PAN chemistry;
• NO loss chemistry;
X
• Ozonide reactions;
• Chamber effects, such as desorption of hydrocarbons and free
radicals from chamber walls.*
As discussed above, variation of these parameters is usually required to
simulate the initiation and maintenance of the radical pool. An analysis
of several of these factors and of their role in predictions is presented
in the Appendix.
From the reported validation studies with kinetic mechanisms and
smog chamber data a degree of uncertainty in absolute levels of ozone
predictions of approximately ±50% is evident. Whether the source of
discrepancy between mechanism predictions and smog chamber data is our
lack of complete understanding of the prevailing chemistry or smog
chamber effects cannot be established with certainty. The production of
ozone in "clean" smog chambers is an indication that so-called chamber
effects do exist, and areas of uncertainty as to mechanism and rate
constants are clearly evident.
With respect to the sources of uncertainty in chemical kinetics we
offer the following conclusions:
*In atmospheric applications any reaction steps required to represent chamber
effects are removed from the mechanism.
34
-------
Although most mechanisms can represent the qualitative features
of the photochemical smog system, there are still important
sources of uncertainty in precise mechanisms, certain rate
constants, and appropriate initial conditions.
Absolute levels of ozone concentrations predicted by kinetic
mechanisms probably have an associated uncertainty of ±50%.
Relative changes in predicted ozone concentration levels
resulting from changes in initial conditions probably have a
somewhat lower uncertainty.
Meteorology
There are three basic meteorological variables of interest: the
wind field, mixing depth, and solar insolation. A problem, common to
all models, is that sparse and often unrepresentative measurements are
used to derive continuous fields over the region. The key question is:
how representative are the interpolated fields of the actual physical
processes in the atmosphere? Roth et al. (15) in their study of wind measuring
stations in the Los Angeles region found a substantial proportion of the
data, taken at identical or adjacent sites at the same time, differed
markedly. A 20% error in any of the measurements is not uncommon, the
uncertainty in the vertical velocity field being somewhat greater. The
basic effect of small perturbations in the wind field is to introduce an
artificial diffusion or smoothing process. Larger errors can affect the
time phasing and magnitudes of pollutant peaks at particular locations.
Mass-consistent wind fields, derived using objective analysis procedures
and appropriate weighting of station data, can substantially reduce the
effects of uncertainties. The artificial creation of convergence and
divergence zones can be minimized. Some problems still remain, however,
in creating three-dimensional wind fields from very limited amounts of
upper level data. To some extent errors in these measurements can often
mask physically meaningful calculations of vertical velocities.
All models considered in this study require specification of the
mixing depth. In most regions it is only measured, or calculated from
temperature profiles, at a very limited number of locations. A 20-30%
35
-------
error is typical; however, in regions of convergence or strong ' .iting
on surface slopes the accuracy can be much worse. (In view of the fact
that concentration predictions are very sensitive to mixing depth, it is
vital to use objective analysis procedures that simulataneously couple
the calculation of the wind field and inversion base location.)
One of the most influential parameters in determining the ground
level oxidant concentrations is the magnitude of the ultraviolet radia-
tion intensity. Photodissociation rates are typically derived from
measurements of total solar flux. This introduces an uncertainty in
photolysis rates of approximately 20% because the reaction rates depend
only on the UV portion of the solar spectrum. Broad band measurements
are difficult to correct for cloud cover and aerosol loading effects.
Extensive calculations were performed in the LIRAQ study in an attempt
to circumvent this problem.
Initial and Boundary Conditions
Most models depend on routine air monitoring data for their initial
and boundary conditions (I.e. and B.C.). These data are typically
interpolated to a fine mesh to provide the surface level I.C. and the
B.C. for the edges of the region. There are obvious problems with this
approach:
(a) The monitoring data are often not representative of the
concentration levels surrounding a monitoring station (see,
for example, Ott and Eliassen (16)). The nature of this
problem is site specific and must be evaluated for each monitoring
site within the modeling region.
(b) The monitoring data represent surface level measurements.
I.C. profile data are required for SAI and DIFKIN; all models
require the concentration levels above the inversion base as
an upper level B.C. This problem with B.C. can to a certain
extent be eliminated by extending the vertical coordinate of
the model domain above the inversion so background levels can
be used. If this is not feasible, and depending on the atmo-
spheric condition, a factor of three should be considered as
the minimal level of uncertainty associated with upper level
36
-------
boundary conditions. These uncertainty levels can be reduced
if upper level measurements are available. The problem with
upper level I.C.s can to a certain extent be eliminated by
starting the model well before the time period of interest.
(c) The most serious problem associated with I.C. and B.C. may be
uncertainties associated with the monitoring methods them-
selves. Trijonis (personal communication) has performed a
critical review and statistical analysis of the quality of
monitoring methods. Based on this work, the precision of the
data corrected for interference effects is recommended to be:
0 /O - excellent, NO - good (error 10%), total hydrocarbons
X j X
fair, and nonmethane hydrocarbons - poor. Of these data the
largest uncertainty is in the nonmethane hydrocarbon data,
which must be further split for validation according to the
requirements of the particular chemical mechanism. A minimum
of 50% uncertainty should be associated with these measure-
ments unless more refined results are available.
Summarizing, for most models the major uncertainties are associated
with the upper level data (factor of 3) and with the hydrocarbon measure-
ment (^50%). The problem with horizontal boundary conditions can largely
be removed by choosing the model boundaries away from strong gradients
and pollutant sources. Initial conditions uncertainties car. be minimized
by starting the calculation well before the time period of interest.
Emission Inventories
The assessment of the level of uncertainty in a particular emission
inventory is obviously a substantial undertaking and, most properly,
should be carried out when the inventory itself is compiled.*
Emissions from each class of source can be characterized according
to:
*Typical levels of uncertainties in mobile and fixed source activities
(e.g., number of vehicle miles traveled and number of units of fuel con-
sumed) should be identified. Then, the typical uncertainties in emission
factors (e.g., grams per mile of pollutant emitted per vehicle mile traveled
and gram of pollutant per unit of fuel consumed) should be combined with the
uncertainties in activities to produce net uncertainties in emissions.
37
-------
'."3 • .>;-•< ." c r spatial resolution,
© Lev,i of temporal resolution, and
a Source activity or emission factor,
The level of spatial resolution achievable is in principle as fine
as one desires since the locations of all sources can presumably be
specified (although traffic count data may not be available on a street-
by-street basis). Temporal emission rates will fluctuate some from day
to day. Emissions from some stationary sources may vary with ambient
temperature, but these variations are generally known as a function of
temperature. The major problem in properly specifying source emissions
is uncertainty in emission quantities arising from uncertainties in
source activities and emission factors.
Two basic factors are involved in emission specification, the
quantity emitted and its composition. Emission compositions are typically
estimated from handbooks such as AP-42. Each table in AP-42 includes a
qualitative estimate of the accuracy of the material on a scale that
varies from A (excellent) to E (poor). Recent studies aimed at estab-
lishing NO and SO emission inventories for stationary sources in the
X ^
South Coast Air Basin have presented estimates of the level of accuracy
of the overall inventories (17). These reports estimate that a ±20%
uncertainty in the total emissions is reasonable, whereas uncertainties
in individual source emissions can range as high as ±300%. A compensating
factor is that generally the large uncertainties are associated with,
small absolute emission levels. Mobile source emission estimates depend
to a large extent on the quality of the traffic data. The level of
uncertainty in the South Coast Air Basin mobile source emissions of NO ,
X
CO, and SO is probably of the order of ±15%.
Probably the most serious emission inventory problems are those
associated with hydrocarbon emissions. The level of uncertainty in the
stationary source hydrocarbon emissions in the South Coast Air Basin is
probably on the order of ±30%. within individual source classes the
38
-------
uncertainties can be as high as ±100%. Mobile source hydrocarbon emis-
sion uncertainties have been estimated in the range 15-50%. Two of the
AQSMs under consideration, namely the SAI model and LIRAQ, require re-
active hydrocarbon speciation in the emission inventory. (In the Bell
Laboratories model all reactive hydrocarbons are represented as propylene.)
Generally insufficient information is available concerning the hydrocarbon
composition of major hydrocarbon sources. It is necessary to estimate a
hydrocarbon breakdown into the four clashes. (Aldehydes constitute an
important class of reacting species, and virtually nothing is known
about aldehyde emissions.) it is difficult to estimate the uncertainty
associated with estimated hydrocarbon speciation. From the point of
view of the AQSM, errors in absolute hydrocarbon levels will be more
influential on oxidant predictions than will errors in class assignments
because reactivities do not vary enormously for the classes. Thus,
uncertainties in hydrocarbon emissions by class, while definitely leading
to uncertainties in oxidant predictions, are not detrimental to accurate
oxidant predictions as are uncertainties in hydrocarbon emissions by
total level.
If we attempt to place overall uncertainty limits on the emissions
for a region such as the South Coast Air Basin that has received consider-
able inventory attention, we might suggest the following ranges:
CO ±20%
+
NO 120%
x
HC ±30%
These ranges represent limits for accuracy-evaluation studies.
Any technique for evaluating oxidant control strategies (e.g.,
Appendix J, smog chamber data, or an AQSM) ultimately requires a total
emission inventory. The rollback-based methods generally require only
a total emission value for the entire region, whereas AQSMs require, in
addition, spatial and temporal resolution. Generally the major problem
associated with obtaining an accurate emission inventory lies in estimating
39
-------
the absolute amounts and not in estimating the spatial and tempoi
-------
tion. These processes have been studied for many years using smog
chambers, and several kinetic mechanisms have been proposed that provide
concentration predictions that agree at least qualitatively with the
experimental results. Yet, all the significant chemical reactions may
not have been identified, and furthermore, the rate constants for many
reactions believed to be important are not known very accurately.
Finally, it has not been established conclusively that a mechanism
validated with smog chamber data accurately represents actual atmo-
spheric chemical processes. Initial and boundary conditions and source
emissions lead to inaccuracies in predictions of AQSMs because of uncer-
tainties due to an imcomplete data base.
In this section we attempt to assess the accuracy of AQSMs for
photochemical oxidant. Recalling the discussion in the section on the
Fundamentals of AQSM, we endeavor to perform here an accuracy-evalua-
tion. Specifically, we seek to answer the question: Given the levels
of uncertainty arising either from incomplete understanding of the basic
physics and chemistry or from inaccurate input due to sparse data, what
are the corresponding levels of uncertainty in predictions of oxidant
levels? Once these levels can be ascertained (or, more precisely,
estimated), then we can attempt to answer the fundamental question
underlying this report, namely: At the present time are there any AQSMs
sufficiently validated/evaluated and appropriate for use in designing
urban oxidant control strategies?
Prior Accuracy-Evaluation Studies with AQSMs
Few of the studies present detailed sensitivity studies of model
predictions to changes in input parameters. The importance of sensitivity
results cannot be overemphasized. Practically they are of value in
assessing the level of detail and accuracy required in model input
parameters or the effects of uncertainties on predictions. The only
extensive, published sensivity study for AQSMs is that of Liu et al. for
the SAI model (18) . Most of their findings are summarized in the fol-
lowing table and detailed in Table 6.
41
-------
RANKING OF THE RELATIVE IMPORTANCE
OF THE INPUT PARAMETERS (SAI MODEL)
Variable
Wind Speed
K
H
\
Mixing Depth
Radiation Intensity
Emission Rate
CO
A
D
C
B
D
B
NO
A
D
C
B
A
A
°3
A
D
C
B
A
B
N°2
A
D
C
B
B
B
A — Most important
D — Least important
The effect of varying boundary and initial conditions in the 1973-
version of the SAI model was reported by Demerjian (19). Average ozone
concentration maps for the Los Angeles basin between the hours of 1:00
and 2:00 p.m. were presented for the base case and the case of boundary
conditions reduced by 50%. Only minor differences were found at the
eastern and northern edges of the basin where the maxima occur, but
significant differences were observed at the western and central por-
tions of the basin. The initial conditions in addition to the boundary
conditions were reduced by 50%. Reduction in the predicted ozone levels
at the northern and eastern edges of the basin was found to be of the
order of 20-30% corresponding to the reduced initial conditions.
The effect of varying initial conditions in DIFKIN was studied by
Demerjian. Comparative trajectory simulations were carried out for Los
Angeles on September 29, 1969, for the original initial concentrations
prescribed and ±20% of those values. It was found that for short time
intervals the predictions are quite sensitive to the initial conditions.
Table 7 summarizes accuracy-evaluation studies carried out with the
Bell Laboratories model.
42
-------
3
w
EH
H
1
EH
D
ft
2
H
a
H
CO
0
H
H
os
iO
o
t"*
i— 3
w
Q
O
S
H
f^-t
C/3
M
S3
EH
CM
0
£H
EH
1 — |
^
H
EH
CO
2
H
CO
M
EH
CM
0
>H
Si
1
D
CO
<1
,
v o
VV
W
,J
m
9
R*
En
e
2-i
~ ^
rH ^^
4-1 O
H
a
£}
e
H
X
r-H
O
o e
a
o
u
,-
o
H
-P
rH
n
>
CJ
-p
'
g
i-i
Tj
_.__)
2.
£
M
r-- LTI vD Lfi
"i ^r o^ ^r
^J" 'N rH ""O
-H- H- 1 1
i_n i — i^rnrn^^cN^ro
• i_T) CN CN ^J O
^ oj ,_) ,H + + | I O
— — - .
O r£
IT) ^
e
CM
CN rH
0 -C +1
LTI PH
CM O O
CM rH,
O + | O
0 A 0 £1
O 'O O TI3
0) 0)
>. X >i XI
XI M XI M
3 3
13 -P T) 4-J
0) 1-1 01 M
X! IP X! CL1 f>'3 o\° o'° 0-°
^ ft^ftOLOLDO
3 3 m CN CN m
ti rH't(^T>,>1>,>,
01 eajgxixixix)
a o o< o
T^i T3 T3 T3 T3 T3
>i C^CCDOJCDCD
r-1 ^rHTO^inuiW
E MSMfdrOfdfll
O O CD CD CD GJ
T3 4-)rC-t-JSHl-(JH!^
c ccnouuo
Oi rHr^HDCDCC U
M G^-lOl^t^rHrH (D
OH C^ W
_n w LO en to to \
_p r^4Jr^jJ4JjJ_p (-vj
c HCnc^cr; E
CJ VjCDVlCDCD(D'']J
e tnetreggg o
CD OJ (D CD CD CD O
^ tjpur?^^)^
•j} ajLQdi^cfiUlLOOO
r^ OJHSOrdrdrdfrJ4-J4->
0 0) O OJ OJ CD
£ ^g^S^SSTJT,
rd rd CD OJ
c c cc^cyiw
0 cnOtfiOOOO^fd
H OJH.OJH-'-'H-HCDCD
-4-1 ^4J^4->4-l4-l4-)MM
ffl rHnjrHrjtdfdfdun
LO >C/J>tnt/lU)U}QH
0
OJ
>1 t/3
-p \
-H r\)
> 6
H
tn o
2 LO
^ MH CD
0 -H 3
Q -H
-M ,15
i. i nj
!M a; 4-1 O
d t-0 O aJ
N XJ
T3 T3 H —
£ ^ VH
-i H 0
2 S X
roo^LDOOr^v"
^D k^) LT) CO 0~i ^ ^
CO -H rH t rH rH | -r
+ 1 + 1 +
COO^OLDOO^J'r'-i
LTl CO CN rH — 1 — 1
+ 14-1 1 +
D
0) U
U3 CD
~\ tn
3 =- ~' c5
• o LO LD o o LP i_n
OLncNCNmmr-Jr— i
O O >,>>->->>-
4-14-i.QXiXlXl^Xl
^^^^^^^^
OJOCDCDOJaOO
cnwcnc/]^^^':^
(QdjcOn^fdr^ u3td
CDCDCDOCJC oo
S-l^-(i_)V-ii-iiH ^V-f
OOOUUDOD
QHQKiQM,5t~
^_
U
CD
U)
4-1 CM >
H £ 4-J
> H
H L1"! 'Jt
W C
3D O 1
M~) ^3 CO 4^ 4-
y~j rH jz; c 03
H rtj 4J M JX
Q > r^
dj ^
rH CD Q C -
ro ui -H
O nj tri -^
-H XI C rd ,r
4J -^ rH -1 ^
!H X T!
CD H nj g ;
> T, X. '~:
^
-H
i_l
tJ^
C1
X
4-1
H
IT,
C
o
•H
4->
aJ
O
0
rH
2
O
4-J
O
c"
fd
-u
rd
CJ
?
CD
U)
fl3
J2
O
X
e
0
M
^
0
H
r3
H
^>
0
+-
•0
*ji
rj
i-^
•K
43
-------
_
y
y
^
rn
w
n
ry
0
y
™
,§*
>ij
S
H-t
, -i
3
M
m
M-i
S
G
T1
S
UJ
JzJ
>s
r
v/J
Q
&
^T
•*
^
P-»
tr^
H
Fj
m
w
CO
*
.n
i — I
CQ
CO
-P
rH
in
cu
T)
cu
•H
JfcJ
id
(/}
rl
cu
cu
G
fo
.H
m
c
•H
OP
VO
CN
G
•H
id
cu
rH
g
10
•H
G
id
O
cu
g
cu
45
4-1
g
Q
IH
•0
CU
T)
O
X
cu
cu
cu
to
G
0
•H
4->
O
id
cu
^-^.
p
H
O
^i
M
4J
in
•H
g
cu
45
0
^-^
Q
rH
O
1
0)
C
cu
iH
E*i
ft
O
4H
0 •
to
cu g
4-i in
id -H
SH G
id
cu o
S cu
0 g
rH
c
TI -H
G
Id 4-1
C
a id
0 -P
H 0
CU ft
s e
0 -H
rH
in
MH -rl
O
K*1
CU SH
in 4->
3 in
td -H
0 g
cu cu
tQ rC
O
^
rH ^
CU Q
CT.rH
rl —
td o
rH
G G
O 0
•H -rl
4-1 10
td 3
}H rH
-P 0
C G
cu o
CJ O
C
o
o •
0 O
•H
M -P
td o
cu id
ft cu
cu
45 X
4J 0
1
1 -H
G 1 g
CU 3 -H
U rH 10
G 0
0 C <
0 O ~
O
ro
0
*— * T3
X • cu
id g -H
cu ft IH
ft ft -H
-P
cu co to
45 0 3
4-> • -n
0
C w
•rl CU -rl
Q
>i
c? Id
T( rH
IH cu
§ C
CU CO -H
to to x
US Id -rl
cu g
SH G
o o cu
CU -H 45
T! 4J 4-1
Id
id SH cu
4-1 >
0 G 0
4-1 CU 43
o id
en G
COG
•rl O O
rd 'H
id PO4-J
cu o id
rH iH
rH O
Ti cu ft
CU > >H
T3 CU O
3 H D
rH C
O rl -H
X CU
0 ft f
ft O
to 3
id CU
U O >1
G 43 id
H td rH
^-^
•
to
G
0
•H
4-1
Id
^
O
rH
id
U
cu
4-1
c
0
4J
U
cu
IH
O
•H
^
CU
45
ft
to
a
0
4-1
*H
O
c
0
•H
4J
td
N
•H
r-l
•H
43
4J
to
*~^
G
0
CU
o
c
0)
H
*H
G
•H
4-1
C
id
G
•rH
g
O
id
4J
SH
cu
X
cu
•H CNO
cu
cu
H
4-1
•p
•H
rH
in
45
^
a
CJ
M-4
o
c
O
•H
4J
fl
rl
O
0
D
G
•rl
T3
cu
^
o
45
tn
>i
Tl
4J
to
^
id
rH
O
a
i^i
H
•H
!H
id
•H
M
^~*
rH
03
U
•H
id
SH
i
H
0
tn
0
cu
id
to
id
cu
cu
i5
1
£
O
•H
in
rH
CJ
G
O
U
rH
id
u
•H
Ti
id
M
1
H
0
to
O
CU
4J
[5
o
iH
0
o
4-1
in
•H
cu
4-1
id
^
tn
4J
•H
^
U
•H
4-1
tn .
•ri in
H CU
id to
cu to
H CU
u
in o
SH rl
id ft
cu
ft cu
ft tn
45
>i ft
4-J in
to td
•H tJ1
0
CU CU
-55
to
c
0
•H
4-1
CJ
id
rl
cu
4-1
C
•rl
^
CU O
45 C
4-1 -H
g
43 >i
rH
Ti C
CU O
rH
I lit
O 0
-P G
G 0
0 -H
U 4-1
U
cu id
43 CU
^
O
4-J CO
•H
'c £
,3
0
c
in o
in -H
O 4-1
H id
G
CNH
O g
a in
cu
4-1
•
rH G
1 -H
U Id
CU 45
tn u
m
1 rH
6 id
0 O
•H
o -d
cu id
rH rl
i cu
0
rH 4-1
1 O
0 G
rH
X -0
CN G
id
II
lyl r^
•H
cu in
tn
td rH
U 0
to
cu o
in SH
td cu
CQ id
SH
0
4-1
G
id
•P
to
4-1
O
G
in
-H
CN
o
a
-f"
|T*
o
^J
0
4-1
G
fd
4-1
to
C
0
CJ
CU
4-1
td
«
1
c
o
•H
tn
3
rH
CJ
c
0
u
,
w
C
O
•H
4-1
id
%
G
(U
D
G
0
O •
a cu
O -P
cu
O g
4-1 3
cu id
O ft
G
tO rH
-P td
W 4-1
O -H
ft >
g
•H id
G CM
0 0
u a
cu +
4J
3§
•
cu
u
G
cu
tr1
cu
in
G
0
u
cu
H
4J
4-1
•H
rH
'H
0
tn
c
o
•H
4-1
O
td
cu
M
0*
13
1
CU
•a
45
CU
r^C
1
G
0
•H
in
3
rH
O
C
0
CJ
1
o
id
cu
X
o
1
cu
T!
45 in
cu G
t! O
H -H
-------
Prior Validation Studies with AQSMs
Validation (or verification) of the AQSMs with historical data is a
necessary step in establishing the level of confidence that can be
placed on predictions. Extensive studies have been performed with the
SAI, LIRAQ, DIFKIN, and, to a lesser extent, Bell models (Table 8).
There are problems in comparing the results of validation studies with
different AQSMs, even for the same region, because of differences in
model formulation, spatial resolution, solution procedures, and different
methods of representing meteorology and chemistry. As can be seen from
Table 8, the variety of regions and conditions for which the AQSMs have
been validated is limited. In particular, there is a lack of validation
results for widely varying emission conditions. Clearly there is a need
to apply the AQSMs to different regions and under different emission
conditions (e.g., weekday vs. weekend). At this time we have to rely on
the results of past validation studies to draw qualitative conclusions
concerning the accuracy of the AQSMs. Each of the four AQSMs has been
applied to at least one urban area. Because of deficiencies in the data
bases used, it is difficult to discern whether discrepancies between
oxidant predictions and observations are the result of errors in the
model, inaccuracies in input parameters, or deviations arising from the
comparison of point data and volume-average predictions. At present
none of the four AQSMs are verified to the extent desired ultimately.
Liu et al. (18) compared the SAI (1973 version) and DIFKIN model
predictions for the 6 days in 1969 (20). In general, the models reproduced
concentration changes of the major pollutants as a function of time
reasonably well. The LIRAQ mudel was validated for three periods in
1973. The spatial distribution of predicted oxidant concentration
compared favorably with the data for the July and August periods, al-
though the predictions were generally lower than the station values.
Predicted NO values did not compare well with the data, possibly as a
result of the assumed vertical profile. The LIRAQ model produced gen-
erally poor predictions for the September days; a number of possible
reasons were cited, but none were quantitatively evaluated. The Bell
45
-------
!H 1 1
O E (D
i - to -ri o i x; w
3 o G > u c 4-> rH i c: • •
i
j
i
to
S
CO
Oi
&<
K
EH
H
[S
CO
W
H
Q
D
EH
CO
2
0
H
f£
^
i — 1
>=!;
>
03
H
I-H
H
i
i
j
13
a)
ITi
a
E
o
0
co
4J
C
18
3
rH
rH
O
Oj
in
o
• rl
^
a)
a
ai
E
•H
EH
CO
C
O
-rl
cn
H CM CM a) • 0) SH i 0 O CU £> 13 £>
34-) OO s£ ^> E2t> 4-1 rH -^4-)i (U 32 lT)rH4-lrH
T3CC-M --P -4-1 O 13 O ?• a> 13 E 1)
a) -H o f-TO oo oo -H in >* » -H CNE n
E ^ -H O >, Z IS Z rt) 13 MH CO 4-> JH X: C44-) O -H O
in w z x! aj o a)OrHn)Q)a)oco 014-1 2
o a M (Din a) ^H 4-i QJ4-)r}i32!a)£nto32
UH cu a> - a) SH d SH c 3 q > -ri rHK a)
^4-)>o> a>3 o)3 amaJrHSio cot) 13
cum 2-ri 3 3 E-'HrHit)a)4J-rHajcoa)
aix:-p "O t! OT)3sa) CNIS -P rH 4->
rH4-ia)o TIC TIC; ucuccr1 ooooido
coroo^it) a>i8 cuo) EO Cu) r^iH -H o -H
OJXiaJlD M M .-HCU-rlKTjra.^^-l
•H 13 3 M nj a) n) a) r\ia) 4J a) - n) a) 13 a)
T3cc c a> a> ox;i8viroci->H>H(T)M
3ir)HT33 E~ri E-^l W4-)>Hc7ia)-HS a*na
4-i a) o-P- O4J- 4->n)Via)<:cM
tni ^113 ooin oon -rnc 3rHdiO4Hani4H
in • (8 a n) a a) C CNM Dcocoo tCOOO
CrHtoaiB raa)o tocuo oaJoaJirliH-
-HOitooa) c >H C SH ooEn)rOO-HO-ri
4-)xi-HO> (T)-<8 18-18 -Moa) — ' >i to CO
ITJ+JCI -ri 4-)oo -POO oa) aw xicoiacora
13 nJW4-l 3OO 3UO SXjT^OlrHdJQirHXlr-HXi
•rlXlXl-PO rH SH rH S-l OC1) (USHTSO) 0)
i — I4-1OCII8 i — I-T3 i — I-H3 -4-l4-)^(I>a)i — l^d)£>a)
n) O a) 18 a) O f^^i 0 rn^ con) 3d) a)XIi8 a)XiCDXi
J>r£jg4_)&H ftOXI f^OXI OSaT^rH^Ei— 1-PJ4-1
1 1
13 in
3 O SH 4J
O C 0) -H 1 01
g rHcni4->a)£raocu
0 4-lCfT)O(Uf84-li I-HIOXI
•rl r) -H IT) t3 •• M >H -H • 3 4-1
4j ' . ^ kOtT'aJoooxiH
rd °j •3fO>OrHOOcuO
OC~ r^warHconjrHOcorac
^ -r4 p; I >i -H C -o 18 rOHlTi O-i O 3 H) -P
i — li OO IT) c^ia)OOn")XIGOO4-)
r\iO ^ O 13 a3Ci — ^XltOOi — 1 (T) -HO
m^^ CN4-)c 0] (DtOO^-14^ CsJI-rH 4-JC^
>i ,,, ii OH cooi-^ >i PJ co o (U o o E - u a) -P
tt ^, „. >i34-100 <8 OH 10) 4JCOCOrHCU C
r^J >-< UJ . rH CT1 a v I^K ^OlrHCOCLJ''^''^! V4M3
r^ P^ fcrl 33CU <^ i^CDiSaJXIf^XI^'^H'HOO
iD'O t^i -MO U n) 13
to id sC to a) a 3
is -H cues x; CEK
OS SH O 4-i O O
U . ItS'-H,— 3 MUSH
OtO^-> GCnE O 13~-^O
XlmqrH >iOO>,3 w 3MH
4-)XOOJ (8-HM 32 >i
3O-H^- patnlm C 0) >i
Omtn OJXil -ri -corn
coa)- o^3rH x M c
X;SHO ow— co oiaio
144 4-> rH 01 0) 0) t-3
OO'O -HglHCO -rl UJ C
CQ-HO OXOaJ- iHC H33
WE C NT3O-H (DO
c - (8 o >, -rH a) 4Jw -gx;--
•HCOa) hfNO) 0 0)03 -rl » 4J
4-)-rlrH> X-HT3rH r-i MMCOq
SHWXC< COSn-rHa IT)SH S-iaJCP
O(8oa) n)r-tr)SHE tn -H O-HOO
fljp^coC) COrH^>tTiaj EH<^ S4-)tOO
2
CK I-H
^ •** H
H K ti i_5
ft, H H H
CO | "1 Q CQ
46
-------
Laboratories model was validated against 0 concentrations for all the
days having characteristics of the hypothetical day of the simulation.
Interpretation of the results of the Bell Laboratories model is difficult
because the predictions apply to a region much larger than that repre-
sentative of the monitoring data.
The basic problem in interpreting validation results is associated
with the "tuning" of model parameters or inputs. ("Tuning" implies the
adjustment of those influential parameters that are imprecisely known
within their estimated upper and lower bounds to obtain as close a fit
of predicted and observed concentrations as possible.) In most serious
validation studies, all model parameters and inputs are determined a
priori based on the conditions of the simulation and are not adjusted to
modify the predictions. Nevertheless, there are usually certain inputs,
for example initial concentrations aloft and boundary concentrations
aloft, that are simply not available and must be estimated. These
concentrations, particularly the concentrations aloft during the day,
are often adjusted to obtain as good a fit as possible. Kinetic rate
constants and turbulent diffusivities are generally not adjusted nor are
the calculated wind and inversion fields. The key issue is the sensi-
tivity of the predictions to those inputs that are adjusted. If the
predictions are highly sensitive, for example, to the assumed ozone
concentration at the inversion base, and that concentration is unknown,
it would be difficult to place considerable confidence in the predictions
of the model. If, on the other hand, the assumed ozone concentration
aloft has only a mild influence on the predicted ozone and ozone aloft
is the only adjusted parameter, then one is safe in placing confidence
in the predictions.
We have chosen not to present here extensive examples of the com-
parison of predicted and observed oxidant concentrations from prior
validation studies for the four AQSMs. Such comparisons are, of course,
available in the primary references. A particularly noteworthy valida-
tion study is that recently carried out by SAI for the Denver airshed
with the current version of the SAI model (21). The work is noteworthy
47
-------
in that it is the first significant verification study carried out with
the SAI model for a region other than the South Coast Air Basin and the
first major verification study of the current version of the SAI model.
Figures 1-3 show comparisons of predicted and observed ozone con-
centrations at each of the monitoring stations on July 29, 1975, July
28, 1976, and August 3, 1976, respectively. Figure 4 shows the varia-
tion of averages over all stations of observations and predictions.
Finally, Figure 5 presents the correlation of observed and predicted
ozone levels for the three validation days at nine stations.
To obtain the results of Figures 1-5, very little parameter tuning
was reported. Aldehyde photolysis rates were adjusted slightly based on
prior trajectory simulations; likewise, initial and boundary conditions
were changed somewhat, although the adjustments in I.e. and B.C. had
less than a 1 pphm effect on the 0 levels predicted. The upper-level
concentration of 0 was assumed to be background (0.02 ppm) and was not
varied during the simulations. Thus, these results appear to reflect
the capability of one AQSM in the absence of parameter adjustment. From
these results a level of accuracy of ±50%, and in several cases a much
lower value, is evident.
Accuracy of AQSMs
In this section we address the question: How accurate are the
current AQSMs? Ideally one would attempt to answer this question through
detailed accuracy-evaluation studies. In these studies uncertainty
limits on the important input parameters would be prescribed. Through
the exercising of the model these input uncertainties would be trans-
lated into uncertainties in predicted oxidant. Time did not permit such
a study for the purposes of this report. Thus, based on an analysis of
the uncertainties associated with key input parameters, we hope to
indicate qualitatively whether any of the AQSMs can predict oxidant
levels with an accuracy necessary for control strategy evaluation. Also
48
-------
J Observed
Computed
15
COMPARISON
STATION
8 _9_ IP_ II II 1
9 10 11 12 1 2 34 5
Time of Day, By Hourly Interval!Start h°Ur|
I stop hour I
Figure 1. Denver air quality model validation comparisons for 29 July 1975.
49
-------
—v — Qb;jrved
0 Computed
C
o
c
d)
o
OJ
c:
o
M
O
-~0 0 0 PARKER RD
Time of Day, By Hourly Interval
Figure 2. Denver air quality model validation computed/observed comparison at
various stations for 28 July 1976.
50
-------
E
CX
CL
OJ
u
C_J
CJ
c
o
M
c
J 5
1ST
55 ? S 9 10 TT 12 1 2 3 4 ' 5
9 10 11 12 1 2 3 4 5 6 7
— v — Observed
Computed Time Qf D B Hour]v Interva1 [start hour I
hour J
Figure 3. Denver air quality model validation computed/observed comparison at
various stations 3 August 1976.
51
-------
DO
J_
3
10 ir
Q.
csilf)
co
-O
O
J_
OJ
Q.
CD
cn
(O
i-
OJ
>
na
O
o;
colcn
-103
Ill
0
-rH
-M
O
-rH
Tl
0)
)M
0,
U)
C
0
•H
Q)
W
W
O
•r-l
-P
(0
4-1
W
0)
0
01
0)
0)
M-l
O
O
•H
4-1
m
•H
VH
rd
(l)
.C
H
cu
in
&
-H
l
u"> O m
(iut|dd) BUOZQ
52
-------
NAAQS
LIMIT
Predicted 0^ Concentration (pphm)
Figure b. Denver air quality model validation computed/observed correlations
(data for 3 days, 9 stations, daylight hours).
53
-------
we will compare the four AQSMs in an attempt to determine which, f any,
of the four are to be preferred.
An important aspect of an AQSM is a proper balance among the degrees
of detail of the treatments of the main chemical and physical processes.
The Bell Laboratories model is the only one of the three that does not
involve any vertical resolution in the concentration distributions; and,
of course, the Bell Laboratories model involves the most detailed chemical
treatment of the four AQSMs. The important issue in this regard that
must be faced is, therefore: What degree of vertical resolution is
consistent with the representation of prevailing chemical processes? If
instantaneous vertical mixing up to an inversion base height is assumed,
then for this assumption to be valid, the characteristic time scale of
the chemical reaction processes must be long when compared to the time
scale for vertical mixing so that the chemical reaction rates based on
vertically integrated concentrations are essentially equal to the ver-
tically integrated chemical reaction rates based on the local concentra-
tions. An analysis of this question is presented in the Appendix. The
conclusion drawn from the analysis in the Appendix is that assumed
instantaneous mixing in the layer between the ground and an inversion
base is inconsistent with a treatment of chemistry as detailed even as
that in the SAI and LIRAQ models. We, therefore, conclude that:
Because of the inconsistency between the levels of detail of
the representations of chemistry and transport, the Bell
Laboratories model is not viewed as a likely candidate for
oxidant prediction when compared to the other three models
that contain vertical resolution.
Because of the specialized nature of DIFKIN as a trajectory model,
this model is not viewed as strong a candidate for general oxidant
prediction as the two grid-based models. With the inclusion of a better
chemical mechanism in DIFKIN, it will serve as a useful tool for special
studies for which a trajectory model is appropriate.
54
-------
As we noted, we would like to translate uncertainties in input
parameters into uncertainties in predicted oxidant levels. If we sum-
marize uncertainties estimated in this section we have:
Chemical kinetics (Ozone levels) ±50%
Initial concentrations (HNO , aldehydes)
Rate constants
Mechanisms of hydrocarbon oxidation
Meteorology
Wind speeds and directions ±20%
Mixing depth ±25%
Light intensity ±20%
Initial and Boundary Conditions ±50%
Initial concentrations aloft
Boundary concentrations aloft
Emission Inventories
NO ±20%
x
Hydrocarbons ±30%
In the absence of detailed accuracy evaluation studies, an estimate
of the uncertainty in predicted ozone levels as a result of the above
input uncertainties acting individually or in concert represents sheer
guesswork. In the Appendix we have singled out two of the most in-
fluential uncertainties, namely initial concentrations of radical-
producing species and the upper level boundary conditions, together with
the question of the degree of vertical resolution for more detailed
study. On the basis of the Appendix and of the prior validation studies
cited in the previous section, one is inclined to place an overall
uncertainty on oxidant level predictions from current AQSMs of ±50%. We
therefore conclude that:
Oxidant level predictions of current grid-based AQSMs (i.e.,
the SAI and LIRAQ models) have an estimated uncertainty of
±50%.
55
-------
CONCLUSIONS AND RECOMMENDATIONS
In this report we have attempted to estimate the Level of accuracy
associated with the oxidant predictions of current AQSMs. The question
we have intended to address is:
At the present time are there any AQSMs sufficiently validated/
evaluated and appropriate for use in designing urban oxidant
control strategies?
It is clear that it is difficult to provide an unequivocal answer
to this question, since one must eventually compare AQSMs with other
methods available for oxidant prediction and select the method most
compatible with the intended use of the method and the available data.
Since this report has not analyzed other oxidant prediction methods, we
cannot provide a recommendation on which method should be used under
which circumstances. The aspect we must consider is: In light of the
accuracy of AQSM oxidant predictions estimated in the preceding section,
do we consider these AQSMs "sufficiently validated/evaluated?"
It seems that AQSMs for photochemical oxidant have reached a level
of detail wherein major new chemical or physical changes are unlikely.
There are inherent uncertainty limits that will be difficult to reduce.
These are the result of the need to represent chemistry in a relatively
concise manner and of the lack of availability of all the required
inputs. We have estimated the current level of uncertainty in absolute
levels of oxidant predictions as approximately ±50%. It is anticipated
that the uncertainties associated with representing relative changes in
oxidant levels is somewhat smaller than ±50%.
In cerms of the four AQSMs surveyed we arrived at the following
conclusion:
• Of the four AQSMs surveyed, the SAI and LIRAQ models are deemed
most appropriate for use in designing urban oxidant control
strategies.
56
-------
The final question of interest is: Should AQSMs be recommended over
existing methods for oxidant prediction? Obviously this question cannot
be answered in the absence of a comparative study of the available
methods. It seems clear, however, that AQSMs represent ultimately the
preferred procedure for oxidant prediction. We offer the following
recommendation:
A comparative study of the SAI and/or LIRAQ models with the
chamber data of Dimitriades (22, 23) be carried out relative
to their utility and accuracy as techniques for designing
urban oxidant control strategies.
Even in the absence of detailed comparative studies, it seems safe
to conclude that either chamber data or AQSMs are more representative of
atmospheric phenomena than is the Appendix J curve. With such a study
as recommended above it would be hoped to establish which of the two is
the most useful at this time.
57
-------
EVALUATION OF KINETIC MECHANISMS
In principle, every reaction appearing in a photochemical smog
mechanism is subject to some degree of uncertainty, whether in the rate
constant or the nature and quantity of the products. In validating a
mechanism, the accepted procedure is to compare the results of smog
chamber experiments, usually in the form of concentration-time profiles,
with simulations of the same experiments using the proposed mechanism.
A sufficient number of experimental unknowns exist in all such mechanisms
so that the models can be "tuned" to fit most experimental data. The
inherent validity or accuracy of any mechanism should be judged on the
basis of this tuning procedure by evaluating how realistic the proposed
parameter variations are.
Uncertainties in the kinetic mechanism are related to inaccurately
known rate constants or products for reactions in the mechanism. Uncer-
tainties associated with comparison of the predictions of the mechanism
to experimental smog chamber data arise, in addition, because the proper-
ties of the photochemical reactor, associated equipment, and experimental
procedures are not completely known.
In this Appendix we discuss kinetic mechanisms for photochemical
smog. We focus particularly on those important aspects of mechanisms
that have some degree of uncertainty at present. Our object is to
assess the expected level of accuracy of prediction of current kinetic
mechanisms.
Inorganic Chemistry
The inorganic reactions important in photochemical smog are, by and
large, well established. Rate constant values for a number of the
inorganic reactions have been revised from earlier values as new deter-
minations have been carried out. The primary photochemical cycle of
NO -NO-0 is well established and need not be discussed. Aside from the
primary photochemical cycle, the most important inorganic reactions are
58
-------
those involving hydroxyl and hydroperoxyl radicals and nitrous and
nitric acids. Rate constant values for reactions involving OH and HO
are difficult to determine, and there is still considerable uncertainty
associated with several of them. Table A-l summarizes recent rate
constant determinations for the important inorganic reactions involving
OH, HO HN02' and HNO .
TABLE A-l. SELECTED INORGANIC REACTIONS
Reaction Rate Constant
@25°C (ppm, min Reference
Units)
1.
2.
3.
4.
5.
6.
NO + OH -> HONO
N0n + OH -»- HONO.
2 2
NO + HO0 -> NO0 + OH
2 2
N°2 + H°2 "* H°2N°2
NCL + H0n -> HONO + 00
22 2
HO^NO_ ->- HO« + NO,.
22 2 2
1.6xl04
4
1.6x10
1.5x10^
3
1.8x10^
1.2x10
a
1.2x10
< 4.5
-1
2-14 min
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(30)
(31)
Estimated based on ratio of rate constants, (k + k )/k = 0.1, as found by
Simonaitis and Heicklen (27). Howard (30) has reported a low pressure (third
order) value of k but extrapolation to 1 atm. is not possible.
The very recent determination of k by Howard and Evenson (29)
necessitates reevaluation of the influence of Reaction 3 on predictions
of mechanisms. Reaction 3 has always been very influential in the conversion
of NO, N02/ and, with a new value of k almost an order of magnitude higher
than the previously used value, its influence may be even more substantial.
59
-------
Olefin Chemistry
Olefin-OH' Reactions—
The olefin-OH' reaction is well established as a key reaction in
photochemical smog. The mechanism of the olefin-OH' reaction for the
common olefins is largely agreed upon, and rate constant values are
reasonably certain. The ethylene-OH* reaction is thought to proceed as
follows:
0
9 I .
CH2 = CH2 + OH -- *-CH2CH2OH 1.33 x 10 ppnT1 min"1
Subsequent steps in the presence of NO are (these reactions will be
discussed subsequently) :
00- 0-
i I
CH CH OH + NO - *^NO2 + CH CH OH
0'
i
CH CH OH - >-HCHO + -CH OH
HO
The main product of the ethylene-OH' reaction is, therefore, formaldehyde.
The propylene-OH* reaction is currently thought to proceed by two
paths, addition and abstraction:
°2 9°'
CH CH = CH + OH' >-CH CHCH OH
4-1-1
O 00- 3.8 x 10 ppm min (33)
00- 0.
I I
CH CHCH OH + NO -> NO + CH CHCH OH
0-
I
CH CHCH OH -> CH CHO +-CH OH
•CH OH >^HCHO + HO •
£ £,
60
-------
After the initial reaction, either an hydroxy-peroxyalkyl or a peroxyalkyl
radical is formed, leading, after reaction with NO, to a hydroxy-alkoxyl
or an alkoxyl radical. The likely reaction paths of these radicals will
be discussed.
Olefin-0 Reactions—
The possible mechanisms of olefin-C reactions were elucidated by
O'Neal and Blumstein (34). The propylene-0 reaction is currently
thought to proceed as follows:
CH CH = CH +
HCO • + CH CHO + OH-
HCHO + CH C(O)O • + OH'
Rate constants for olefin-ozone initiation steps are reasonably well estab-
lished. The principal areas of uncertainty in olefin-ozone chemistry are:
(a) the extent of stable ozonide formation, and (b) the split between
each of the two paths, as shown above. In propylene photooxidation
experiments, measurement of the acetaldehyde/formaldehyde ratio aids in
establishing the importance of each of the two reaction paths.
Paraffin Chemistry
The main oxidation reaction of paraffins is with hydroxyl radicals.
The products of reaction between paraffins and OH' in air are a peroxy-
alkyl radical and water. For butane, for example, hydroxyl radicals can
abstract a hydrogen from either the primary or secondary carbon as
follows:
+ OH'
CH CH CH CH 00' + HO
00-
I
CH CH CHCH + HO
61
-------
The peroxyalkyl radicals react with NO to produce NO and alkoxyl radicals.
Because alkoxyl radicals can thus be viewed as the effective product of
the paraffin-OH' reaction, alkoxyl radicals play an essential role in
the chemistry of paraffin systems.
Aldehyde Chemistry
It is now well established that aldehydes play an extremely important
role in smog chemistry. Aldehydes participate in the reaction process
in two ways: (a) photolysis to give stable products and radicals, and
(b) reaction with hydroxyl radicals to give acyl radicals that are
rapidly converted to peroxyacyl radicals. The detailed chemistry of
aldehydes is still somewhat uncertain because of uncertainties in aldehyde
photolysis rates and in rate constants for aldehyde-OH- reactions.
Formaldehyde Chemistry—
Photolysis of formaldehyde leads to hydrogen atoms and formyl
radicals:
HCHO + hv —H* + HCO •
The hydrogen atoms immediately form hydroperoxyl radicals. The possible
paths for reaction of the formyl radical are:
HCO- + 0 a+HCO •
b 3
-> HO • + CO
c
-> OH- + CO
Quantum yields as a function of wavelength for formaldehyde photolysis
are highly uncertain. At longer wavelengths formaldehyde photolysis
proceeds as:
HCHO + hv -> H + CO
62
-------
Osif and Heicklen (35) estimated the ratio of the rates of paths a and b
to be 5 and the ratio of the rates of paths c and b to be less than 0.2.
The peroxyformyl radical will react with NO as follows:
HCO • + NO >^NO + HO • + CO
-J £•*£,£
Note that path a results in the net conversion of two molecules of NO to
NO , whereas path b leads to only a single NO to NO conversion. Thus,
the split between paths a and b influences ozone formation.
The reaction of formaldehyde with hydroxyl radicals proceeds as
follows:
HCHO + OH- ->• HCO- + H20 2.2 x 10 pprn'^in"1 (36)
Formaldehyde photolysis and reaction with OH* lead to the formyl radical.
Further elucidation of the fate of the formyl radical is an important
need in determining formaldehyde chemistry.
Acetaldehyde Chemistry—
The principal acetaldehyde reactions are:
°2
CH CHO + hv *-CH302« + HCO3'
°2 4 -1 -1
CH CHO + OH- >-CH C(0)0 • + HO 2.2 x 10 ppm min (37)
Peroxy Radical Chemistry
Peroxy radical chemistry forms the basis for the production of
ozone and organic products. Because of the difficulties associated with
measurement of rate constants of reactions involving peroxy radicals,
there is still a considerable degree of uncertainty concerning peroxy
radical reactions.
63
-------
Reactions of Alkoxyl Radicals with Oxygen and Thermal Decomposition--
Two possible reaction paths for the alkoxyl radical are decomposition
and reaction with molecular oxygen. It is now established that the
decomposition of the shorter chain alkoxyl radicals is unimportant
compared to the reaction of the shorter chain alkoxyl radicals with
molecular oxygen (38, 39). Barker et al. have shown that the rate
constants for alkoxyl radical-0 reactions are almost independent of the
size of the radical (39). Unimolecular decomposition rates, on the
other hand, increase rapidly with increasing size of the radical. Thus,
as the alkyl group increases in size, ac some point unimolecular decom-
position becomes competitive with reaction with 0 . The point at which
the two rates become equal has not been firmly established.
The reaction of methoxyl radicals with 0 is:
CH 0- + 0 + HCHO + HO • 0.9 ppm min~ (38,39)
Decomposition of the methoxyl radical is negligible when compared to the above
reaction. The sec-butoxyl radical formed in the photooxidation of
butane undergoes decomposition or oxygen addition as follows:
CH CH CHCH - »-CH CH 0 • + CH CHO
~> £ O 3 £ £ ~J
0- 0
CH CH CHCH + O - *=~CH CH CCH + HO •
The relative importance of these two paths has not been established but can
be inferred from the relative amounts of methyl ethyl ketone, butyraldehyde,
and acetaldehyde formed in a butane photooxidation. The primary butoxyl
radical probably does not decompose but reacts with oxygen.
Hydroxy-alkoxyl radicals are formed in the olefin-OH* reaction.
For example, in propylene photooxidation we obtain the hydroxy-alkoxyl
radical through the reaction
00- O
l I
CH CHCH OH + NO -> NO + CH CHCH OH
64
-------
Although the fate of this radical is not known precisely, it is generally
felt at this time that unimolecular decomposition occurs exclusively:
0-
I
CH CHCH OH -> CH3CHO + 'O^OH
Because of the importance of the olefin-OH- reaction, uncertainties in
the fate of the hydroxy-alkoxyl radical will introduce uncertainties in
overall kinetic mechanisms.
Reactions of Peroxy Radicals with NO and NO—
Conversion of NO to NO occurs primarily by reactions of the form
RO • + NO ->• NO + RO-
Aside from the HO --NO reaction, rate constant values have not been
measured for the peroxyalkyl-NO reactions. Darnall et al. (40) have
postulated that longer chain peroxyalkyl radicals (C. ) will add to
NO to form an excited complex:
RO • + NO -> RO NO
RON02
RONO -^RO- + NO
+RONO
M
Darnall et al. estimated the ratio k /k, to be 0.09 for butyl, 0.16
for pentyl, and 0.6 for hexyl systems (40).
Peroxyalkyi-NO reactions have been studied by Simonaitis and
Heicklen (27). They reported that the ratio of the rate constants for
RO -NO and RO -NO reactions is 2.2. The peroxyalkyl-NO reaction may
proceed by:
65
-------
R02N°2
R'CHO + HNO
Simonaitis and Heicklen reported that production of the alkylperoxynitrate
occurs 75 percent of the time (27). For methylperoxy, for example,
HCHO
The peroxynitrate may react with NO to produce methyl nitrate or with
NO to give methyl nitrite (41). As in the case of the HO -NO reaction,
the RO -NO reaction diminishes the formation of ozone. At this time
the mechanism (and rate constants) of the RO -NO reaction is largely
uncertain. Because of its importance in ozone formation, this reaction
should be studied further.
The primary pathway of alkoxyl-NO reactions is (42):
4 -1 -1
RO* + NO -> RONO 4.9 x 10 ppm min
The alkyl nitrite will photolyze to give back the alkoxyl radical and
NO. The photolysis rates of the nitrites are, however, uncertain.
Since alkyl nitrites represent a reservoir for NO , uncertainty in the
X
photolysis rates of the alkyl nitrites leads to uncertainties in the
overall photooxidation chemistry.
Two reaction paths for the alkoxyl-NO reaction have been reported
(38,39):
RO- + NO
•> RCHO + HNO
66
-------
Barker et al. (39) estimated the ratio k /k to be 0.3, whereas Baker
b a
and Shaw (43) and Wiebe et al. (44) estimated k,/k = 0.11. Mendenhall
b a 3-1
et al. (38) determined the rate constant for RO- + NO as 7.8 x 10 ppm
-I 4-1
min , and Barker et al. found the rate constant to be 1.55 x 10 ppm
min (39). Thus, the rate constant is uncertain by about a factor of
two. In addition, there is some uncertainty as to the split between the
two reaction paths shown above. The split is important since path a.
removes both a free radical and an NO , whereas path b returns a hydroxyl
radical from the subsequent photolysis of nitrous acid.
The competition between NO and NO for peroxyacyl radicals is an
important factor in both PAN and ozone formation. For peroxyacetyl
radicals, for example:
O a 0 3
CH COO-+ NO -+NO + CH CO' 4 x 10 (45)
9 b ° 3
CH COO + NO -> CH COONO (PAN) 2.09 x 10 (45)
Radical-Radical Reactions—
The reactions of peroxyalkyl and peroxyacyl radicals with HO •
provide a sink for free radicals.
H02' + H02« -> H202 + 02 8.3 x 1Q3 pprn'^in"1 (46)
RO • + HO • -*• ROOH + O
RC(0)02- + H02' -> RC(0)OOH
67
-------
In addition, peroxyalkyl-peroxyalkyl reaction may be a source of alkoxyl
radicals:
RO • + RO • -> 2RO- + 0
A rate constant value of about 500 ppm min for the methylperoxy
recombination was reported by Parkes et al. (47), Whitbeck et al. (48),
and Hochanadel et al. (49).
Because of the experimental difficulties associated with measurement
of radical-radical reaction rate constants, at present the rate constants
of the above reactions are not available in general and must be estimated.
Since these reactions control the size of the overall radical pool, they
are important in kinetic simulation.
PAN Chemistry
The PAN formation reaction was given the section on Reactions of
Peroxy Radicals with NO and NO . Pate et al. (50) suggested that PAN
decomposes as follows:
0 O
ii ii _i
CH COONO -»• CH COO- + NO 0.0372 min (45)
PAN concentrations are controlled by the competition between NO and NO
for peroxyacyl radicals and by PAN decomposition. The peroxyacyl-NO
reaction is given in a preceding section. The acetyl radical resulting
from that reaction decomposes as follows:
so that the result of the competition strongly influences ozone production.
68
-------
Aromatic Chemistry
The lack of understanding of aromatic chemistry represents one of
the most serious weaknesses in our overall knowledge of the photochemical
smog system. Hansen et al. (51) and Hendry* have recently measured rate
constants of reactions involving aromatic hydrocarbons and hydroxyl
radicals. (Reaction of aromatics with atomic oxygen is slow and can
4 -1 -1
presumably be neglected.) A value of 1 x 10 ppm min for the toluene-
OH' reaction is consistent with the results of both groups. Toluene is
the aromatic species most studied.
Hendry has suggested that the main pathway is addition of OH- to the
rather than abstraction of a hydrogen from the methyl group because of
the similarity of the toluene-OH* rate constant to those of other alkyl
benzene-OH* reactions.
Schwartz (52) reported nitrotoluene and aerosol formation in toluene
photooxidation experiments. Akimoto et al. (53) found o-cresol,
a-nitro-toluene, and m-nitrotoluene resulting from the toluene-OH-
reaction.
The mechanism of aromatic-OH- reactions is largely speculation at
this time. As noted, this mechanism represents an important area of
uncertainty in kinetic simulations.
Generalized Mechanisms
The key issue in the use of a generalized mechanism is the selection
of the organic species to be included within each lumped class. In the
original Hecht et al. (7) mechanism (see Table 3) each lumped species
represents a different hydrocarbon class. The rate constants corres-
ponding to each class are to be determined as average rate constants
*Private communication from D.G. Hendry to G.Z. Whitten, 1976.
69
-------
calculated on the basis of the composition of the class. A problem with
this approach is that it is difficult to represent high concentrations
of less reactive species and low concentrations of highly reactive
species in the same class with one set of rate constants. As the re-
action proceeds, the more reactive species are preferentially depleted
first, and the effective lumped rate constants change with time. Thus,
when employing a lumped mechanism it is desirable to collect species
that not only behave mechanistically similarly but also have roughly the
same reactivity.
As noted earlier, Whitten and Hogo (9) have suggested a somewhat
different interpretation for the lumped reactant classes in the Hecht et
al. mechanism. In that mechanism, termed the Carbon Bond Mechanism,
each carbon atom is treated according to its bond type. Organics are
divided into four groups: (a) single-bonded carbon atoms (paraffins and
the single-bonded carbon atoms of olefins, aromatics, and aldehydes),
(b) fast double bonds (olefins excluding ethylene), (c) slow double
bonds (ethylene and aromatics), and (d) carbonyl bonds. Aside from the
inclusion of new information on several reactions, the principal change
from the Hecht et al. mechanism is that ethylene, the least reactive
olefin, is included with aromatics in the class of slow double bonds.
Summary of Uncertainties in Kinetic Mechanisms
The concentrations predicted by the kinetic mechanism are extremely
sensitive to the values of several reaction rate constants. Reactions
which are particularly sensitive are those governing the conversion of
NO to N02,
NO + HO -> NO + OH
^- j£
NO + RO -»• NO + RO
and the reactions which initiate the oxidation of olefins,
70
-------
O + olefins -»• radicals + stable products
OH + olefins ->- radicals + stab.le products
Hecht and Dodge performed a systematic sensitivity analysis on all the
reactions in the original Hecht-Seinfeld-Dodge lumped kinetic mechanism
(13). The conclusions of the study were compiled in the form of a
ranking of the reactions by their "sensitivity-uncertainty" index. This
index is an indicator of the combined sensitivity of the mechanism to
variations in the reaction rate constant and the experimental uncertainty
of the rate constant. Since this study was performed, only a few rate
constant determinations have been significantly improved. However, a
number of previously accepted kinetic data are currently in dispute.
For example, the reaction of NO with HO is now suspected of being an
addition reaction to form peroxynitric acid, HO NO , rather than a
pathway for forming nitrous acid, HONO, and 0 .
Since urban smog is initiated photochemically, kinetic mechanisms
must be able to predict the photolysis rates of pollutants that absorb
ultraviolet light. From Beer's Law, in an optically thin medium, the
"rate constant" governing the photolysis rate of a compound is given by:
X2
k = a -x ) f
-------
In smog chamber simulations, the photolysis rate is usually expressed
in terms of k , the rate constant for NO photolysis. From this, with
information on I (A), e(A), and <(>(A), photolysis rates of other species
can be predicted. Considerable uncertainty exists in the measurement of
c|>(A) for certain species. For instance, the photolysis of ozone can be
important in the formation of OH radicals. In the wavelength region of
interest, the primary quantum yields for the processes
O + hv -> Of1!) + 0(3P)
are still uncertain. While extinction coefficients are relatively easy
to measure in the laboratory for most species, quantum yield measurements
can be exceedingly difficult.
Another important photochemical process is the formation and sub-
sequent reaction of excited states. The rates of thermal reactions can
be enhanced by several orders of magnitude if one or more of the reactants
are vibrationally or electronically excited. For instance while ground
state 0(3P) atoms are unreactive toward such species as H , H^O, and
NO, singlet oxygen, O^D), reacts rapidly. Similarly, the oxidation of
SO in clean air probably takes place by the reaction of triplet SO_(3SO )
formed by the absorption of UV light by ground state SO followed by
internal energy transfer processes. SO may also be considerably more
reactive toward hydrocarbons than the ground state SO . Unfortunately,
both the formation and reaction mechanisms of most electronically excited
species are highly uncertain.
In summary, of all the unknowns in the homogeneous chemistry of
photochemical smog, the processes most often parameterized are:
72
-------
• the rates of reaction of RO with NO;
• rates of reaction and mechanisms of oxidation of the olefins,
aromatics and alkanes, i.e., numbers and types of radicals
formed;
• relative rate constants for the photolysis of absorbing species
such as 0 and HONO.
The second major set of unknowns in photochemical smog mechanisms
concerns the effects of smog chambers on the observed chemical processes.
Since mechanisms must be validated using smog chamber data, the char-
acteristics of the chamber must be built into the model explicitly.
Some of the specific effects or characteristics that must be considered
are: the spectral distribution and absolute intensity of the photolyzing
lamps; the absorption, desorption, and chemical reaction of species on
the walls; the initial loading of impurity species in the chamber on the
walls or in the gas; and the effects of leakage, sampling, and possible
temperature variations during the run. Of these effects, probably the
most important are the properties of the photolysis lamps. Photolysis
rates of absorbing species cannot be predicted with accuracy if I (A) ,
the incident light intensity distribution, is not known with accuracy.
This information must be coupled with the absolute rate of photolysis of
at least one species such as NO to compute the appropriate photolysis
rate constants.
Also important is the characterization of the initial contaminant
loading in the chamber. When mechanisms overpredict the length of the
induction period in which radical concentrations are initially building
up, it may be due to the presence of an absorbing species either in the
gas phase or on the walls that photolyzes. Actual measurement of the
species accounting for these effects is complicated by their low con-
centrations.
When these effects are not adequately characterized one usually
begins by parameterizing the N0_ photolysis rate constant k , the initj
concentration of trace photolyzable species such as HONO, and the wall
73
-------
absorption rate of ozone. If still less is known about the experimental
situation, the value of simulating the data becomes questionable.
The third major set of unknowns in simulating laboratory systems
concerns the reactions which take place heterogeneously, either on the
walls or on aerosols. Many reactions are thought to take place hetero-
geneously.
NO + NO + HO + 2HONO
NO + NO + HO + 2HNO
which produce nitrous acid and nitric acid. Evidence for the heterogeneous
nature of these processes comes from the strong dependence of measured
rate constants on the reactor surface to volume ratio. (The disappearance
of SO in smog chamber experiments also seems to have a strong hetero-
geneous component either as a result of reactions in droplets or the
wall-catalyzed formation of polymeric sulfur-oxygen species that remain
on the walls as films.) Recent work has shown that certain long-lived
free radicals such as HO can be lost to particles at appreciable rates.
Diffusion and subsequent loss of radicals to reactor walls occurs con-
stantly but these processes do not affect the homogeneous chemistry
appreciably. Heterogeneous processes, in general, are difficult to
account for in kinetic mechanisms and are usually ignored.
In summary, the three classes of phenomena that often require para-
meterization in kinetic mechanisms are the homogeneous chemistry, the
behavior of the chemical reactor, and heterogeneous chemistry. Tuning
the model to account for unknown or uncertain chemical and physical
effects is a legitimate procedure provided that the exact steps taken
are spelled out in detail and lie within physically realistic bounds. A
properly tuned mechanism is capable ot predicting the concentration-time
profiles of the stable species such as NO, NO , 0 , and hydrocarbons
74
-------
within 10-20% over a wide range of initial conditions. In theory, such
a mechanism, minus the steps included to account exclusively for chamber
effects, should be capable of predicting atmospheric concentrations with
the same accuracy.
75
-------
REVIEW AND ANALYSIS
Kent R. Wilson
ABSTRACT
The question of the utility for urban oxidant control strategies of
present Physical-Chemical Models (PCMs) for oxidant air quality is addressed.
Alternative models for calculating oxidant levels from precursor levels
are discussed: linear rollback, modified rollback, statistical models,
smog chamber analogies, and PCMs. Sources of error in PCMs are described,
including chemical reaction errors (initial conditions, sources and
sinks, kinetic schemes, and computational limitations), errors in physical
motion (windfields, omission of small scale motions, and grid size
effects), and finally errors due to the approximate treatment of the
interaction between physical motion and chemical reactions (turbulent
inhomogeneity and its effect with nonlinear kinetics). The utility of
present PCMs is categorized by particular usage: essential for sci-
entific understanding, probably useful for comparison among different
land-use, transportation, and both short and long term emissions control
alternatives (but not for prediction of absolute levels), marginally
useful for monitoring siting decisions, and unwise for use as legal
standards. Several areas for improvement of PCMs are suggested, in-
cluding (a) more general testing than just against monitoring data, (b)
the theoretical and experimental investigation of the effects of tur-
bulent inhomogeneity, (c) statistical extensions of PCMs to oxidant
distributions, reduced kinetic schemes and model output statistics, (d)
the use of specialized processors for greatly speeding up program execu-
tion and (e) the use of dynamic computer graphics to extract more under-
standing from the model calculations. It is concluded that while no
model is as yet very accurate, an iterative approach is likely to lead
to improvement of both models and of oxidant air quality.
77
-------
INTRODUCTION
The question that has been posed by EPA for this study is, "At the
present time, are there any air quality simulation models sufficiently
validated/evaluated and appropriate in designing urban oxidant control
strategies?"
In answering this question we will first consider briefly the
various alternatives for relating emission levels to oxidant levels,
including rollback, smog chamber analogies, statistical/empirical re-
lationships, and what we will call Physical-Chemical Models (PCMs); in
other words, largely deterministic models that attempt to simulate those
processes of physical motion and chemical reaction in the atmosphere
involved in photochemical air pollution. Second, we will discuss some
of the sources of error in PCMs, including input data, modeling of physical
motion, chemical reaction, interaction between physical motion and
chemical reaction, as well as errors involving comparison of theoretical
model output with experimental measurements, both smog chamber and
atmospheric measurements. Third, we will define utility and differ-
entiate among a number of different ways of using PCMs in urban oxidant
control strategies, concluding that the utility of PCMs ranges from
essential to pernicious depending upon the specific mode and purpose of
usage. Lastly, we will suggest a number of directions for improvement
of PCMs.
ALTERNATIVE EMISSIONS TO OXIDANT MODELS
Several alternatives (54) exist for predicting oxidant levels as a
function of hydrocarbon and (often) NO emissions levels, including simple and
X
modified rollback, the drawing of analogies between reactions in smog chambers
and in the atmosphere, various relationships derived from statistical/empirical
analyses of atmospheric data, as well as the use of physical-chemical models.
We will begin by admitting that there are no means at present to accurately
predict atmospheric ozone levels as a function of emissions levels. All of
78
-------
the above alternative approaches have substantial limitations and will con-
tinue for the foreseeable future to be less accurate than is desirable.
Nonetheless, each of the above approaches can be useful in the proper
context, and we will survey these approaches briefly in order to set the
context for a deeper discussion of the utility of PCMs.
Linear Rollback
Linear rollback results from the rather arbitrary assumption that there
is a direct proportionality between reactive hydrocarbon (RHC) emissions in a
region and maximum 1-hour average oxidant levels, and that this oxidant level
will be reduced proportionately to reactive hydrocarbon reduction (55,56).
Such an approach suffers from many defects. First, linear rollback only
relates oxidant levels to reactive hydrocarbon emissions, yet such levels
clearly also depend on NO emissions. Second, the chemical kinetics that link
X
reactive hydrocarbon concentrations to ozone concentrations are known both
experimentally and theoretically to be nonlinear, making suspect the linear
rollback assumption. Third, the use of maximal 1-hour oxidant is statisti-
cally unfortunate, a nonrobust measure of the oxidant distribution that is an
unstable indicator of the real atmospheric situation.
Modified Rollback
Modified rollback replaces the assumed linear relationship between RHC
emissions and maximal oxidant with a statistical relationship based on atmo-
spheric data. The most common version is the EPA Appendix J formulation (57-59),
which relates Lhe upper limit of maximal 1-hour oxidant levels to 6-9
a.m. nonmethane hydrocarbon concentrations. The Appendix J model has several
serious problems, among them being neglect of NO emissions altogether, of
post 9 a.m. RHC emissions, of transport (6-9 a.m. emissions and afternoon
maximal oxidant are liable to relate to two different air masses), the non-
robustness of an upper limit curve that depends only on a few extreme measure-
ments, the agglomeration of data from several cities with differing emissions
and meteorological patterns and the possibility that the pattern of the oxi-
79
-------
dant ana rK.itraethaiie hydrocarbon data may be due more to mutual ouireiation to
meteorological variation than to what would happen to oxidant levels if RHC
emissions themselves were varied, given that the system is not linear.
Aerometric Statistical Models
Several attempts have been made to model the relationship between emis-
sions levels and oxidant levels by statistical analysis of aerometric data
(54). Merz, Painter, and Ryason regressed oxidant onto early morning NO and
X
total hydrocarbon levels (60). Kinosian and Paskin (61) regressed oxidant
levels onto NO concentrations for various classes of total hydrocarbon levels.
Trijonis examined the joint distribution of morning hydrocarbon and NO
X
levels and related this to downwind oxidant levels (62). Trijonis et al. (63)
and Martinez et al. (64) examined recent historical trends in monitored oxi-
dant and precursors levels, the relation among these trends providing a po^-
sible basis for predicting future effects of emission control measures.
Caporaletti et al. (65) regressed hours over the Federal oxidant standard onto
RHC picked up by simplified windstreams over an air basin, thus producing a
statistical model that transforms a spatially resolved RHC emissions inventory
into a predicted, spatially resolved, statistical oxidant distribution.
Such aerometric statistical models have two major advantages. First,
they are based on observed atmospheric data, so that one might hope to make
successful predictions even when an understanding of the underlying mechanism
is not complete. The statistical relationships thus discovered might then
lead to later deterministic understanding of cause and effect. Second,
compared to PCMs, they are relatively easy and inexpensive to develop and to
use, and thus can be applied both widely and repeatedly, for example, to the
evaluation of many different land-use alternatives.
Statistical models, on the other hand, have limitations that should be
carefully considered. A key problem is that while statistical models are
derived from data from the recent past, they are often called upon to predict
future oxidant levels from RHC and NO levels falling far outside the range of
80
-------
the data used in the derivation. Such extrapolation leads to uncertainties
that are difficult to estimate (54).
Smog Chamber Analogies
Smog chambers have provided much of our understanding of photochemical
air pollution. From the time evolution of ozone concentration in an irra-
diated chamber as a function of initial precursor concentrations/ one can draw
analogies to the expected time evolution of oxidant in the real atmosphere as
a function of emissions levels. By modeling the chemical kinetics of the
reacting system and adjusting the rates of reaction and initial concentrations
to match the chamber observations, one can hope to correct for chamber effects
such as wall reactions and to extend the validity of the analogy beyond the
range of the chamber experimental data base (12, 22, 23). Strengths of this
approach include: (a) that it is calibrated to actual measured chamber experi-
mental data over a range of RHC and NO concentrations, and (b) that it could
JC
be extended, at least in principle, to cover a variety of precursors, for
example, different RHC mixtures. Weaknesses include: (a) that the effects of
physical motion of the atmosphere, including transport, diffusion, mixing of
air parcels of different histories and compositions, and the interaction
between turbulence and chemical reaction are neglected and may be quite
important, and (b) that chamber effects, such as wall interactions, may not in
reality be removed by kinetic modeling corrections, as the tuning of the
kinetic scheme may not correspond to the real chemical mechanism.
Physical-Chemical Models
The final alternative, and the one to which this study is most directly
addressed, is the use of physical-chemical models in an attempt to simulate
the atmospheric processes of physical motion and chemical reaction.
The need for physical-chemical models seems obvious. The phenomena of
photochemical air pollution are coo complex to be intuitively fully compre-
hended. We are dealing with a system that is:
81
-------
o time dependent, both in emissions and in meteorology,
• spatially varying, both in emissions and in meteorology, and
• nonlinear with respect to variation of input variables because of the
nonlineari ty of the kinetic equations.
One could write down ab initio expressions that represent in theory the
processes of interest as accurately as one desires, for example, microscopi-
cally as an exercise in time-dependent quantum mechanics, or on a more macro-
scopic level, in terms of hydrodynamics and chemical kinetics, we could des-
cribe the processes as:
i = ^({c^}, T) + Si(ir, t) , i = 1,...,N (Eq. 1)
in which at time t, ir is the velocity at position ir, {c } is the set of
concentrations of the N chemical species involved, R. is the rate of produc-
tion of species i by chemical reaction at temperature T, and S, represents
the sources and sinks of species i as a function of position and time. The
difficulty is that any ab initio approach is calculationally quite unfeasible.
Thus a series of approximations are made, aimed at achieving a balance among
what can be calculated, what can be measured, and what can be estimated.
Instead of calculating rate constants from quantum mechanics (a presently
impossible task for all but the simplest reactions), they are measured in the
laboratory. Instead of calculating the large scale time evolution of the
atmosphere, meteorological measurements are used for information on winds,
temperature, humidity, and solar radiation. Instead of trying to calculate the
small scale motions of the atmosphere, approximations are usually made such as
K theory, neglect of effect of turbulent motion on averaged reaction rates,
etc. (3,66,67). Several of the sources of error introduced by these ap-
proximations will be discussed in the following section.
General Observations
Now that the various alternative approaches for emissions to oxidant
models are listed, we can make two general observations. First, all of the
82
-------
above approaches have serious flaws, but all of them also have utility in the
proper situation. There is, as yet, no way to transform even a perfectly
accurate emissions distribution as a function of space and time into a satis-
factorily accurate spatial-temporal ozone distribution. The real question,
thus, is: How useful is a particular type of model under a specific set of
circumstances? Second, the distinction between statistical and deterministric
models is only one of degree. Statistical models are guided by deterministic
understanding of cause and effect. Primary consideration is given to RHC and
NO emissions in statistical models because of deterministic knowledge. Statis-
tical aspects are included, at least implicitly, in all deterministic PCMs.
Rate constants carry a statistical uncertainty, wind fields are constructed
from scattered measurements, a whole ensemble of detailed atmospheric motions
are consistent with all the available meteorological data, and the model itself
is usually tuned in one way or another to match smog chamber and atmospheric
observations.
SOURCES OF ERROR
Comparisons among PCMs and sources of error in PCM have been reviewed
several times recently (18,19,68). Basically we can divide error sources
into three categories: (a) chemical, having to do with concentrations and
rates of reaction, (b) physical, having to do with atmospheric motion, and (c)
chemical-physical interaction, having to do with the influence of motion on
reaction.
Chemical Reaction Errors
Errors in the chemical aspects of PCMs are of four types: initial condi-
tions, sources (emissions) and sinks, kinetics, and computational.
Initial Conditions—
Incorrect concentrations at t = 0 lead to errors in initial conditions for
integrating approximate forms of Equation 1. For example, questions that can
83
-------
be posed as initial condition uncertainty and involve long range transport (69-
72) and multiday carryover of reactants, intermediates and products, form much
of the basis of the serious disagreements as to the mechanism of elevated rural
ozone (73-76).
Sources and Sinks—
Lack of knowledge of correct species resolved spatial-temporal distri-
cts of sources
remains a problem.
butions of sources of emissions and sinks, i.e., S.(ir, t) in Equation 1,
Kinetics—
No existing kinetic scheme (2,7,10,11,74,77) takes into account more
than a fraction of the species and reactions that must be involved in photo-
chemical air pollution due to the great variety of reactive hydrocarbons.
Thus, one of two approximations is ordinarily applied, (a) picking a single
hydrocarbon, for example propylene (2), as a surrogate for all reactive hy-
drocarbons, or (b) lumping hydrocarbons together into one or more classes (7).
The first choice slights the chemical diversity of actual atmospheric reactive
hydrocarbons but remains more faithful to measurable rate constants. The
second choice allows an approximate match to the hydrocarbon distribution, but
at the cost of the use of rate constants that are removed from direct measure-
ment and represent average values or values arrived at by tuning a kinetic
calculation to match an average result.
As Demerjian, Kerr, and Calvert (11) have warned, there is a danger in
fitting rate constants to match gross smog chamber or atmospheric data in that
the tuning may not represent the real mechanism and so may not give correct
results when applied to a different situation with a different set of data.
They remind us that "computer fits of experimental data based on such inac-
curate choices of kinetic data obviously provide no validation of the mechanism
choice, but represent a sophisticated exercise in curve fitting."
84
-------
Other areas of concern are (a) how to include heterogeneous reactions,
both involving aerosols and surfaces that may act as sinks, (b) a general lack
of certainty in our knowledge of rate constants as well as possibly more subtle
effects, such as (c) the production of unrecognized electronically metastable
or vibrationally excited product states that might react rapidly before being
quenched (78).
Computational Limitations—
Other sources of uncertainty arise not from errors in the data input or
the differential equations used but rather from the limits to affordable
computation time that determine the fineness of the spatial and temporal mesh
size that can be integrated. For example, a reactive point source may be
simulated incorrectly if it is spread out over a sizable grid cell, par-
ticularly given the nonlinear nature of the kinetics (79). Comparisons of
model output representing averages over a grid cell may be incommensurate with
monitoring station readings that represent a microscale environment (66,67).
Physical Motion Errors
Errors in modeling the physical motion of the atmosphere can be divided
into three categories: external factors, approximations in the formulation,
and computational limitations.
External Factors—
Given that the usual approach is to define a wind field from rather
sparse and largely ground-level observations, the possible errors are con-
siderable, particularly in periods of stagnation.
Approximations in Formation—
For computational practicality, small scale atmospheric motions are not
treated explicitly, but instead usually handled by K theory, defining a
85
-------
horizontal K and vertical K eddy diffusivity, which attempt to represent the
averaged effect of small scale motions. This introduces the practical un-
certainty as to what values to assign to K and K_ as a function of position
and time. K is perhaps not important compared to horizontal advec-
H
tion, but K can certainly have a large effect. As an illustration, in La-
grangian modeling by Environmental Research and Technology (80), the values of
K used vary by one and one-half orders of magnitude as a function of height and
time.
Computational Limitations—
Because, given affordable computer time, only relatively large spatial-
temporal grids can be used, the concentrations {c,} actually dealt with in the
computations are spatially and temporally averaged, eliminating the direct
calculation of the effects of any but larger scale motion. Again, this leads
to incommensurability of prediction with the micro-scale measurements made at
actual monitoring sites (66).
Chemical-Physical Interaction Errors
A possibly important source of error is the effect of calculating as
homogeneously mixed an atmosphere that may in reality, due to the details of
small scale motion, be chemically quite heterogeneous (81,82). Because of
the nonlinearity of the differential equations representing reaction rates as
a function of concentration, the approximation of using spatially and tem-
porally averaged concentrations may lead to quite erroneous results. There is
experimental evidence, for example, from the LARPP project, that the value of
the ratio of
[03] [NO]/[N02]
varies by up to an order of magnitude from that expected from the photostation-
ary state one would expect from a well-mixed system (80). A deeper analysis
of these data by Calvert (83) indicates that the problem is worse in the early
morning,- when pockets of NO-rich air from surface emissions may be alternating
86
-------
with pockets of aged 0 -rich air fron higher elevations. That the problem is
not just a local one is indicated by power plant plume measurements indicating
persistence of NO/NO inhomogenieties for up to 90 km downstream (84) and air-
borne 0 measurements indicating inhomogeneities persisting even out to sea (71)
The ad hoc biasing of one or more homogeneous concentrations (80) to try
to deal with this problem is clearly an unsatisfactory solution, and more
sophisticated theoretical treatments (66 and private communication from J'.H.
Seinfeld, 1977), as well as additional experimental measurements, seem impera-
tive.
UTILITY AND AREAS OF APPLICATION
As (a) there is no way to predict ozone levels as a function of emissions
levels with great accuracy in any atmospheric situation, (b) the various uses
of such predictions imply quite varied needs and constraints, and (c) there is
the considerable variety described above of alternatives to PCMs for such
prediction, the question as to the utility of PCMs in designing urban control
strategies must be answered with discrimination as to specific use. As will
be seen below, it is the author's belief that the utility of PCMs ranges from
essential for some uses to disastrous in other contexts.
First, let us define useful as being significantly better than what is
being done at present. In this way we can hopefully keep moving in the di-
rection of improvement without being caught in the trap of inaction because of
the inevitable imperfection of our ability to predict. We should candidly
admit that oxidant prediction of all types is subject to large uncertainty
(85), that the field is yet very young, and that no part of it is free from
difficulties. Statistical/empirical modeling is often called upon to venture
forth onto statistically unfirm ground to predict what would happen as a
result of control strategies that would change the emissions pattern upon
which the statistical analysis is based. Smog chamber experiments are con-
ducted on too small a physical scale, with (a) the absence of both the large
and small scale physical motions that characterize the atmosphere and affect
-------
final ozone levels through transport and mixing and turbulent nonmixing ef-
fects on reactions, and (b) the presence of different boundary conditions
(wall effects instead of natural sources and sinks). PCMs are up against a
highly multidimensional problem that is really too complex, with too many
unknowns for the state of our chemical and meteorological understanding and of
our computational capacity.
Yet acknowledging all these problems, we can still gain more by using
models where appropriate than by not using them. Even rollback tells us the
direction to proceed with basin-wide hydrocarbon control and gives at least a
rough figure for the magnitude.
We will now consider different possible applications for PCMs (19) and
evaluate their utility in each case. The conclusions are summarized in Figure 1.
Short Term Prediction
If one could predict in advance the occurrence of oxidant episodes (54),
one could then try to reduce their levels by short term emissions control
strategies, for example, control of traffic and shutting down of particular
industries. In addition health warnings could be issued.
Episode Prediction—
In the short term time scale, oxidant level changes are mainly determined
by meteorological changes and not by emissions changes. These meteorological
changes are controlled mainly by physical processes or. a scale large compared
to an air basin, and thus air basin deterministic models of PCM scale are
inappropriate vehicles for such prediction. Statistical models (54), perhaps
linked to large scale deterministric weather predictior. -odels, are a more
appropriate tool.
Conclusion: PCMs are unlikely to be useful for e'issie prediction.
-------
en
a
o
0)
ft
-p
•9
,Q
O
cu
en
3
XI
•H
W
w
Q
O S
^ a:
O LU
O
cc
Q_
CE S
O CE
nr uj
to t—
CO
UJ
o
o
CO
o_
LU
p
UH
0)
W
P
• >i
W rH
S X5
u to
ft-§
IH M
O ft
w II
a>
•rH CN
-U
•H -
rH rH
•H (0
-M -H
3 '^
T! 0)
a in
fti W
0)
w
0) II
w
D r-H
OJ
J-l
89
-------
On the other hand, PCMs could be useful tools for the evaluation of urban
short terra emissions control strategies, to test the oxidant effect of varying
traffic patterns or industrial emissions in response to an episode predicted
by other means. A start in such a direction has already been made (86) in
terms of evaluating air pollution emergency plans, although the pessimistic
conclusions are somewhat in disagreement with a statistical analysis of the
response of the real atmosphere to weekend-weekday variations in emissions
(74). A kinetic study by the Bell Labs group (87), with a simplified physical
motion component, of the weekend-weekday effect again points the way toward
the use of PCMs in evaluating urban short term control strategies.
Conclusion: PCMs are probably already useful for episode control
strategy evaluation.
Long Term Prediction
Several classes of uses exist for PCMs in long term control strategies,
whose utilities need to be separately evaluated.
Scientific Understanding—
Control strategies should be evaluated from the basis of a thorough
scientific understanding of photochemical air pollution, in order that the
effects of alternative control strategies may be understood before a decision
among them is made. PCMs are an essential part of the process of moving toward
such a thorough scientific understanding. First, a PCM should ideally be an
embodiment of present physical and chemical understanding, cast in a mathe-
matical and algorithmic form approximating this understanding as closely as
one can, given the strictures of data availability and computational cost.
Given the multidimensional nature of the problem, with a spatially and tem-
porally varying set of inputs being nonlinearly transformed to a spatially and
temporally varying ozone output, there seems to be no reasonable alternative
to PCMs for embodying and dealing with scientific understanding of photochemial
air pollution.
90
-------
In addition, because PCMs, in contrast with all other means of predicting
ozone from emissions levels, are a simulation, an algorithmically parallel
structure to scientific understanding, when the output of PCMs diverges from
experimental evidence we must conclude that either: (a) the scientific
understanding is incorrect or incomplete, (b) the approximations used in the
solution are inappropiiaue, (c) an, error ir translating the scientific under-
standing into algorii hmi.c form has been made, (d) there Is insufficient data,
or (e) the experiaientaI evidence Js incorrect. If the other error sources can
be ruled out, then we can conclude that scientific understanding is indeed
incorrect or incomplete. Thus PCMs can and must be used as a means to test
scientific understand"! Tig against experimental data, and to prompt the improve-
ment of this understanding when it is found wanting.
Conclusion: PCMs are essential Lo the improvement of scientific under-
standing of photochemical air pollution and thus to the
rational choice among urban oxidant control strategies.
Land Use and Transportation planning Alternatives—
Many emissions to oxidant approaches consider the air basin as a lumped
unity. This simplifies the development and use of the prediction scheme, but
is an inadequate approach for land use and transportation planning, which
practically must consider geographical placement of alternative sites and the
spatial distribution of their oxidant effects. An implicit assumption of
lumped basin models is that the spatial distribution of emissions sources
remains constant. By contrast, one of the needs of land use and transporta-
tion planning alternatives is the evaluation of the effects of changes in this
distribution.
Thus spatial resolution in terms of emissions input is a key need for
land use and transportation planning alternatives, and spatial resolution in
oxidant output is useful in that it allows integration over population dis-
tributions and thus the calculation of relative dosage among planning alter-
natives. While statistical models with spatial resolution are possible and
have been implemented (65,88), the use of PCMs is certainly a reasonable goal.
91
-------
In this context, however, caution must be exercised. As is discussed
below, the use of PCMs in the foreseeable future for the definition of legal
standards for land use and transportation plans would probably be counter-
productive. Thus, PCMs at this time would be most valuably employed as re-
finement tools, to best distribute emissions over a region to minimize popula-
tion dosage (65), once total levels have been set by the application of other
standards.
It should be realized that to do this, PCMs need to be evaluated for
enough meteorological regimes that a weighted distribution of effects can be
calculated that reasonably represents the meteorological distribution for the
region.
In such a situation, where the aim is to choose among different spatial
distributions of emissions, it is the author's opinion that existing PCMs could
play a useful part.
Conclusion: Existing PCMs, if properly used, could play a useful
role in choosing among urban land use and transportation
alternatives providing different emissions pattern dis-
tributions, if the overall emissions totals were already
set.
Emissions Control Alternatives—
Emissions control alternatives that involve differing controls on various
classes of emitters entail not only differences in aggregated emissions totals
for RHC and NO , but also (a) changes in the relative levels between RHC and
NO emissions, (b) alternative distributions among various individual hydro-
X
carbons, and (c) alternative spatial and temporal distribution patterns, for
example, power plant NO emissions from elevated stacks versus ground-level rush
hour peaked NO emissions from mobile sources. While it would be unwise to
expect to predict absolute oxidant levels with accuracy, PCMs can be expected
to provide relative guidance among such control alternatives (89). Again,
weighting by meteorological regime distribution would seem important.
92
-------
The question has been raised as to whether the Bell Labs model (2), in
particular, is sufficiently validated to use in strategy evaluation. First,
it should be noted that while the chemical kinetic scheme of this model is
very extensive, the physical portion ot the model is as yet rudimentary,
involving only a few well-mixed boxes, into which emissions are immediately
distributed with spatial resolution only oa the scale of a county. Second,
only one hydrocarbon, propylene, is used as a surrogate for all RHCs. In
addition, trapped 0 is treated in a rather arbitrary fashion. Thus, at this
stage, the absolute agreement with monitoring station measurements must be
regarded as likely to be somewhat fortuitous. One would hope to see this
model evolve toward a better balance of attention to chemical reaction and
atmospheric motion, and then to see an error analysis in which the sensitivity
of the model to parameter change and the uncertainty in parameter values are
jointly evaluated to establish error estimates. All of this is to say that
one should not expect great absolute accuracy oc any PCM, and in particular,
not from an effort which, although very commendable and extensive for such an
early stage, still has a maturing process of balancing and testing to go
through.
Conclusion: PCMs could probably be used to provide relative guidance
among emissions control alternatives, but should not be
relied upon to provide absolute levels.
Monitoring Station Siting—
Monitoring stations should be sited so as to pick up the most information
on pollution distributions for the least number of stations. PCMs could be
used (again a distribution of meteorological regimes would be an improvement)
to evaluate approximate oxidant distributions for use in decisions on moni-
toring station siting. In actual practice, it is likely to be more cost
effective to use a combination of mobile stations for probing and then un-
certainty reduction techniques for sampling that have been extensively de-
veloped for petroleum and mineral prospection.
Conclusion: PCMs are possibly useful for monitoring station siting,
but unlikely to be cost effective unless they are already
implemented for another purpose.
93
-------
Legal Standards—
In the context of legal standards for oxidant control strategies, a major
emphasis must be placed upon having a simple set of rules that can be inter-
preted by many users with minimal variability as to conclusions drawn. PCMs do
not fulfill this need. Drawing legal standards in terms of PCMs would likely
lead to confusion, extensive litigation, and a loss of momentum in the implemen-
tation of control strategies. Our legal system is not suited to probabilistic
decision making and cannot be expected to deal easily or well with rules and
regulations with multidimensional uncertainty.
The difficulty with using PCMs as legal standards is that there are too
many parameters (reaction rates, initial conditions, eddy diffusion values,
etc.) whose uncertainty is sufficiently great that by judicious (but not
unreasonable) choice one could arrive at any point within a wide range of
final absolute oxidant predictions. To attempt to settle such a situation in
the context of litigation is an unpleasant prospect and should be avoided at
the outset.
Conclusion: PCMs are inappropriate as legal standards because the
multidimensional parameter uncertainties are too complex
to handle in the context of litigation.
DIRECTIONS FOR IMPROVEMENT
Several areas are obvious ones for needed improvements in PCMs: better
emissions inventories differentiated as to RHC type, measurement or improve-
ment of particular rate constants, more three-dimensional windfield and tempera-
ture information, better determination of vertical eddy diffusivities, improved
measurements of initial conditions, and better micromodeling to match in-
dividual monitoring test sites. In what is to follow, five aspects out of
many possibilities for improvement are selected for discussion, (a) model
testing, (b) the interaction between physical motion and chemical reaction
(c) statistical improvement, (d) computer improvements and (e) improved
interaction with the user.
-------
Model Testing
Given the error sources cited above, in the simulation of chemical reac-
tions (initial concentration conditions, sources and sinks of species, imper-
fections in the kinetic scheme, and compu-ational limits to grid size), in the
modeling of physical motion (lack of knowledge of gross wind field, approxi-
mations to avoid calculation of small scalt. motions, practical computational
limits to resolution), and in the simulation of the interaction between small
scale motion and chemical reaction due to tuxbulent inhomogeneity and non-
linear kinetics, one irr .,t certainly rest such PCMs against experimental reality.
It is important to realize that it is r.^i. -^nough just to get the right
final answer from the simulation; that it is not sufficient just to predict
ozone levels reasonably for a given situation. If one is going to have
confidence in using the model to predict the outcome in an altered set of
circumstances (for example, the effect of a control strategy that would
considerably alter emissions), the underlying cl'tmical and physical mechanisms
in the model irr.st be correct and not just tuned in a multidimensional curve -
fitting pro.rdure to give the right numbers in a particular case.
Thu:.> it. is important to test the individual components of the model
against experimental reality to insure their faithfulness to the actual chemical
and physical mechanisms and not just to test the model as a whole.
For example, as is common, the kinetic schemes used in PCMs should be
carefully checked against both measurements of individual rate constants and
against smog chamber measurements. We should not be satisfied in the long run
with the practice of sweeping the unknowns under the rug by tuning concentra-
tions and rates to make the kinetic schemes agree with particular smog chamber
results. The goal should be to understand both the experimental and the
modeling situation thoroughly enough that we can quantify the true mechanism.
Then, on the basis of such understanding, one can find appropriate approxima-
tions to speed up the computation, approximations whose validity can. be checked
against the true mechanism.
95
-------
Similarly, the physical motion aspects of the model should be checked
independently of the chemistry. Wind fields constructed with partial data can
be checked against the rest of the wind measurements. Vertical wind and
temperature calculations can be checked against further elevated measurements.
Transport calculations can be checked against tracer studies, both specific
release and CO distributions. Eddy diffusivity calculations can be checked by
tracer studies, perhaps by smoke release.
The interaction of physical motion and chemical reaction is of particular
concern, and is treated next.
Interaction Between Physical Motion and Chemical Reaction
As is discussed above, there is both theoretical and experimental evi-
dence that the atmosphere is not well mixed and that this could lead to
significant errors in PCMs. It would seem that a high priority should be
placed on both theoretical (66) and experimental studies of this problem, as
there are grounds for believing that both may be fruitful. For example, two
rapid response NO, NO , and 0 monitoring systems could be set up and their
cross-correlations measured as a function of distance between inlets. In this
way the spatial-temporal structure of concentration variations could be measured,
and these measurements could be repeated under various conditions, such as
near a freeway, far downwind from sources, at elevated locations, etc.
The question of "non-mixedness" needs to be resolved, as it gives rise to
uncertainty both as to the accuracy of present PCMs and as to the applicability
of smog chamber measurements for atmospheric analogy.
Statistical Improvements
It would seem worthwhile to investigate the incorporation of three
statistical aspects into the now largely deterministic PCMs, (a) an ensemble of
meteorological regimes in an attempt to model not just worst days but the
yearly distribution of ozone, (b) a statistically determined reduced kinetics
96
-------
package for greater speed, and (c) a version of model output statistics for
predicting ozone at specific monitoring stations.
Meteorological Regimes—
The idea has been suggested that one should not just model worst days,
but enough different categories of days that one can transform an emissions
pattern into a predicted yearly ozone distribution. Again, the transformation
is a nonlinear one, so that the shape of the ozone distribution is not pre-
cisely preserved under scale changes of the emissions pattern. (Perhaps this
project is already underway in the San Francisco Bay Area where Smalley
patterns have already been defined for meteorological regimes.) Certainly a
system that predicted an ozone distribution, or at least the number of hours
over the Federal standard, would provide easier comparison with mandated air
quality goals (65,81).
Reduced Kinetics--
If a PCM is to be applied repetitively, for example, to evaluating land
use and transportation alternatives, a large fraction of the computer expense
might be saved by statistically developing a reduced chemical kinetics module
(Private communication from W.S. Meisel, Technology Service Corp., 1974) (81).
For such a use, one needs a black-box with only a few input-outputs (perhaps a
few categories of RHC concentrations, NO, NO , 0 , and sunlight intensity).
^ -3
One does need, however, to have confidence that over the range of variables of
interest, the reduced scheme does give the same answer as a more complete
scheme. This is a statistical task, both to sample the range of the variables
of interest and then to fit a simpler model to the calculated points. In
other words, the suggestion is to work downward, systematically reducing a
larger kinetic scheme in which one has confidence, rather than starting with a
simplified scheme and trying to tune it to match particular measurements.
97
-------
Model Output Statistics—
An alternative to the buildup of microscale models (66) to bridge the gap
between PCM grid-cell scale output and the specific environment of a particular
monitoring station would be the adoption of the Model Output Statistics (MOS)
viewpoint from weather forecasting (76). One would then regress the oxidant
readings at a particular station back onto several variables, involved in and
predicted by the PCM, to arrive at a statistical scheme relating PCM output
and specific monitoring station oxidant level. Such a scheme would be useful
in evaluating situations in which emissions do not grossly change, but should
be used with caution in cases in which emissions are greatly changed from
those used to develop the MOS.
Faster Computers
Some guidance can perhaps be found in considering the development of
other modeling fields in which computation involving the solution of large
sets of coupled differential equations is also involved, fields such as stel-
lar, plasma, and molecular dynamics. Two key developments are the move toward
specialized processors to speed up the computation and dynamic computer
graphics to allow the user to better extract understanding from the calcula-
tion.
PCMs fit the criteria for advantageous use of specialized processors.
First, they involve in their exercise the repetitive use of the same or
similar code, so that run costs, if they were really used widely, would be
large compared to coding costs, which are higher for specialized processers.
Second, they are amenable to being split into parallel streams, which need
communicate only when a step is completed, and thus parallel hardware can be
employed.
The present limits of computer expense by such techniques could be
pushed back by two or three orders of magnitude. For example, a specialized
system that runs approximately one-fourth as fast as a CDC 7600 for molecular dyia
98
-------
mics is now running (90), and the processor cost for this machine, which is
available 24 hours per day, is approximately $65,000. On a larger scale, a
machine for solving aerodynamic-coupled differential equations is being
proposed by NASA (91), which will run 100 times faster than a CDC 7600 and
whose price is estimated at $30 million. Perhaps somewhere in between there is
a cost effective solution for PCMs.
Computer Visualization—
In other related modeling areas (90), dynamic computer graphics has
proved very useful in allowing the user to comprehend the multidimensional
time evolution of complex systems. Thus one could watch the time evolution of
various species' concentrations as well as the wind field, all perceived
three-dimensionally, and thus visualize what is rather incomprehensible when
presented as tables of numbers.
CONCLUSIONS
None of the several alternative routes to predict oxidant levels from
precursor emissions, linear rollback, modified rollback, aerometric statis-
tical models, smog chamber analogies, or Physical-Chemical Models (PCMs) are
capable of the accuracy desirable for evaluating urban oxidant control strat-
egies. Yet each is useful in the proper context. Rollback is arbitrary;
Appendix J is statistically unsound; statistical models are on shaky ground
outside the range of the data used to develop them; smog chamber analogies
leave out the effects of transport, mixing, and turbulent inhomogeneity and,
in addition, involve surface effects different from the real atmosphere; and
PCMs require large data inputs, large amounts of computer time, and involve
approximations in chemical kinetics and small scale motions that lead to
error.
99
-------
Thus the utility of PCMs, like other models, needs to be evaluated in the
context of particular usage. The view presented in this paper is that existing
PCMs are essential for the scientific understanding of photochemical air pol-
lution, and that they are probably useful for the choice among urban land use
and transportation alternatives and for choosing among emissions control
alternatives both for the short and long term as long as they are not used to
set absolute goals. PCMs appear only marginally useful for monitoring station
siting, and their use is judged to be most unwise as legal standards.
In conclusion, a feedback approach (63,92) to urban oxidant control is
endorsed, with the realization that oxidant control probably will still be a,n
issue 25 years from now. Control should be looked at as an iterative process
in which we make the best judgment we can at each stage, realizing our falli-
bility, but as progress is made we continue to re-evaluate the system. We
cannot expect perfection from PCMs now or in the future, but we can look for
improvements both in the atmosphere and in our ability to model it.
100
-------
REFERENCES
1. Dimitriades, B. , and A.P. Altshuller. International Conference on
Oxidant Problems: Analysis of the Evidence/Viewpoints Presented.
Part I: Definition of Key Issues. JAPCA, 27(4):299-307, 1977.
2. Gradel, T.E., L.A. Farrow, and T.A. Weber. Kinetic Studies of the
Photochemistry of the Urban Troposphere. Atmos. Environ. 10(12):
1095-1116, 1976.
3. Seinfeld, J.H. Air Pollution, Physical and Chemical Fundamentals.
McGraw-Hill, New York, 1974.
4. Liu, M.K., and J.H. Seinfeld. On the Validity of the Grid and Trajec-
tory Models of Urban Air Pollution. Atmos. Environ. 9(6/7):555-574,
1975.
5. Hecht, T.A., and J.H. Seinfeld. Development and Validation of a Genera-
lized Mechanism for Photochemical Smog. Environ. Sci. Technol. 6(1):
47-57, 1972.
6. Eschenroeder, A.Q., and J.R. Martinez. Concepts and Applications of
Photochemical Smog Models. In: Photochemical Smog and Ozone Reactions.
Adv. Chem. Series, No. 113, pp. 101-168. American Chemical Society, Washing-
ton, D.C., 1972.
7. Hecht, T.A., J.H. Seinfeld, and M.C. Dodge. Further Development of
Generalized Kinetic Mechanism for Photochemical Smog. Environ. Sci.
Technol., 8 (4):327-339, 1974.
101
-------
8. MacCracken, M.C., and G.D. Sauter, (ed.). Development of an Air Pollu-
tion Model for the San Francisco Bay Area. Final Report to the National
Science Foundation. University of California, Livermore, Calif., 1975.
9. Whitten, G.Z., and H.H. Hogo. (Systems Applications, Inc.) Mathematical
Modeling of Simulated Photochemical Smog. EPA-600/3-77-011, U.S. Environmental
Protection Agency, Research Triangle Park, N.C., 1977. 306 pgs.
10. Niki, H., E.E. Daby, and B. Weinstock. Mechanisms of Smog Reactions.
In: Photochemical Smog and Ozone Reactions. Advances in Chemistry,
Series 113, American Chemical Society, Washington, D.C., 1972. pp. 16-57.
11. Demerjian, K.L., J.A. Kerr, and J.G. Calvert. The Mechanism of Photo-
chemical Smog Formation. Adv. Environ. Sci. Technol., 4:1-262. John
Wiley-Interscience, New York, 1974.
12. Dodge, M.C. Combined Use of Modeling Techniques and Smog Chamber Data
to Derive Ozone-Precursor Relationships. Int. Conf. Ox. Poll., Proc.
2:881-889. EPA-600/3-77-001b, U.S. Environmental Protection Agency,
Research Triangle Park, N.C., 1977.
13. Dodge, M.C., and T.A. Hecht. Rate Constant Measurements Needed to Improve
a General Kinetic Mechanism for Photochemical Smog. Inter. J. Chem.
Kinetics, 7(Symp.l):155-163, 1975.
14. Chan, W.H., R.J. Nordstrom, J.G. Calvert, and J.H. Shaw. Kinetic Study
of HONO Formation and Decay Reactions in Gaseous Mixtures of HONO, NO,
NO , HO, and N . Environ. Sci. Technol., 10 (7):674-682, 1976.
15. Roth, P.M., S.D. Reynolds, P.J.W. Roberts, and J.H. Seinfeld. Develop-
ment of a Simulation Model for Estimating Ground Level Concentrations of
Photochemical Pollutants. Report 71-SAI-21, Systems Applications, Inc.,
San Raefael, Calif., 1971.
102
-------
16. Ott, W., and R. Eliassen. A Survey Technique for Determining the Represen-
tativeness of Urban Air Monitoring Stations with Respect to Carbon Monoxide.
J. Air Poll. Contr. Assoc., 23(8):685-690, 1973.
17. Bartz, D.R. Control of Oxides of Nitrogen from Stationary Sources in the
South Coast Air Basin. California Air Resources Board, Report 2-1471,
1974.
18. Liu, M.K., D.C. Whitney, J.H. Seinfeld, and P.M. Roth. Continued Research
in Mesoscale Air Pollution Simulation Monitoring: Volume I. Assessment
of Prior Model Evaluation Studies and Analysis of Model Validity and
Sensitivity. EPA-600/4-76-016a, U.S. Environmental Protection Agency, Research
Triangle Park, N.C., 1976.
19. Demerjian, K.L. Photochemical Air Quality Simulation Modeling: Current
Status and Future Prospects. Int. Conf. Ox. Poll., Proc. 2:777-794.
EPA-600/3-77-001b, U.S. Environmental Protection Agency, Research Triangle
Park, N.C., 1977.
20. Reynolds, S.D., M.K. Liu, T.A. Hecht, P.M. Roth, and J.H. Seinfeld.
Mathematical Modeling of Photochemical Air Pollution: III. Evaluation
of the Model. Atmos. Environ., 8(6):563-596, 1974.
21. Anderson, G.E., S.R. Hayes, M.J. Hillyer, J.P. Kellus, and P.V. Mundkur.
Air Quality in Denver Metropolitan Region. EPA-908/1-77-002. Systems
Applications, Inc., 1977. 460 pp.
22. Dimitriades, B. An Alternative to the Appendix-J Method for Calculating
Oxidant and NO Related Control Requirements. Int. Conf. Ox. Poll., Proc.
2:871-879, EPA-600/3-77-001b, U.S. Environmental Protection Agency, Research
Triangle Park, N.C., 1977.
23. Dimitriades, B. Oxidant Control Strategies, Part I. Urban Oxidant
Control Strategy Derived from Existing Smog Chamber Data. Envir. Sci.
Technol., 11(1) .-80-88, 1977.
103
-------
24. Cox, R.A., R.G. Derwent, and P.M. Holt. Relative Rate Constants for
the Reactions of Hydroxyl Radicals with Hydrogen, Methane, Carbon
Monoxide, Nitric Oxide, and Nitrous Acid at Atmospheric Pressure and
296°K. J. Chem. Soc. Faraday Trans. I., 72 (9):2031-2043, 1976.
25. Sie, B.K.T., R. Simonaitis, and J. Heicklen. The Reaction of OH with
NO. Int. J. Chem. Kinetics, 8(1)-.99-106, 1976.
26. Hudson, R.D. (ed.) Chlorofluoromethanes and the Stratosphere. NASA
Reference Publication 1011, 1977.
27. Simonaitis, R., and J. Heicklen. Reaction of HO with NO and NO .
J. Phys. Chem. 78(7):653-657, 1974.
28. Cox, R.A., and R.G. Derwent. Kinetics of the Reaction of Hydroperoxy
with Nitric Oxide and Nitrogen Dioxide. J. Photochem., 4(1-2):139-153,
1975.
29. Howard, C.J., and K.M. Evenson. Kinetics of the Reaction of HO with NO.
Geophysical Research Letters, 4:437, 1977.
30. Howard, C.J. Kinetics of the Reaction of HO with NO . J. Chem. Phys.,
1977 (in press).
31. Graham, R.A., Winer, A.M., and J.N. Pitts, Jr. Temperature Dependence
of the Unimolecular Decomposition of Pernitric Acid and Its Atmospheric
Implications. Chem. Phys. Letters, 1977 (in press).
32. Cox, R.A. The Photolysis of Gaseous Nitrous Acid - A Technique for Obtain-
ing Kinetic Data on Atmospheric Photooxidation Reactions. Int. J. Chom.
Kinetics Symposium, Vol. 1:379, 1975.
33. Atkinson, R.A., and J.N. Pitts, Jr. Rate Constants for the Reaction of?
HO Radicals with Propylene and the Butenes Over the Temperature Range
297-495°K. J. Chem. Phys., 63:3591-3595, 1975.
104
-------
34. O'Neal, H.E., and C. Blumenstein. New Mechanisms for Gas-Phase Ozone-
Olefin Reactions. Int. J. Chem. Kinetics, 5 (3):397-413, 1973.
35. Osif, T.L., and J. Heicklen. Oxidation of HCO Radicals. J. Phys. Chem.,
80(14):1526-1531, 1976.
36. Morris, E.D., Jr., and H. Niki. Mass Spectrometric Study of the Reaction
of Hydroxyl Radical with Formaldehyde. J. Chem. Phys., 55:1991-1992,
1971.
37. Morris, E.D., Jr., D.H. Stedman, and H. Niki. Mass Spectrometic Study
of the Reactions of the Hydroxyl Radical with Ethylene, Propylene, and
Acetaldehyde in a Discharge-Flow System. J. Amer. Chem. Soc., 93:
3570-3572, 1971.
38. Mendenhall, G.D., D.M. Golden, and S.W. Benson. The Very-Low-Pressure
Pyrolosis (VLPP) of n-Propyl Nitrate, tert-Butyl Nitrate, and Methyl
Nitrate. Int. J. Chem. Kinetics, 7 (5):725-737, 1975.
39. Barker, J.R., S.W. Benson, and D.M. Golden. The Decomposition of Di-
methyl Peroxide and the Rate Const
J. Chem. Kinet., 9(1):31-53, 1977.
methyl Peroxide and the Rate Constant for CH 0 + 0 -> CH 0 + HO . Int.
J £, £, Z.
40. Darnall, K.R., W.P. Carter, A.M. Winer, A.C. Lloyd, and J.N. Pitts, Jr.
Importance of RO + NO in Alkyl Nitrate Formation from C -C Alkane
Photooxidations under Simulated Atmospheric Conditions. J. Phys. Chem.,
80(17):1948-1950, 1976.
41. Spicer, C.W., A. Villa, H.A. Wiebe, and J. Heicklen. Reactions of
Methylperoxy Radicals with Nitric Oxide and Nitrogen Dioxide. J. Amer.
Chem. Soc., 95(1):13-20, 1973.
105
-------
42. Batt, L. , R.D. McCullock, and R.T. Milne. Thermochemical and Kinetic
Studies of Alkyl Nitrites (RONO)-D (RO-NO), the Reactions Between RO
and NO, and Decomposition of RO. Int. J. Chem. Kinetics Symposium,
7 (Symp. 1):441-461, 1975.
43. Baker, J., and R. Shaw. Reactions of Methye, Ethoxye, and t-Butoxye
with Nitric Oxide and Nitrogen Dioxide. J. Chem. Soc., 1965. p. 6965.
44. Wiebe, H.A., A. Villa, T.M. Hellman, and J. Heicklen. Photolysis of
Methyl Nitrite in the Presence of Nitric Oxide, Nitrogen Dioxide, and
Oxygen. J. Amer. Chem. Soc,, 95(1):7-13, 1973.
45. Cox, R.A., and M.J. Roffey. Thermal Decomposition of peroxyacetyl-
nitrate in the Presence of Nitric Oxide. Environ. Sci. Technol.,
11:900, 1977.
46. Hampson, R.F., and D. Garvin. Chemical Kinetic and Photochemical Data
for Modeling Atmospheric Chemistry. NBS Technical Note 866. National
Bureau of Standards, Washington, D.C., 1975. 112 pp.
47. Parkes, D.A., D.M. Paul, C.P. Quinn, and R.C. Robson. The Ultraviolet
Absorption by Alkylperoxyl Radicals and Their Reactions. Chem. Phys.
Letters, 23(3):425-429, 1973.
48. Whitbeck, M.R., et al. A Kinetic Study of CH 0 and the (CH ) CO
•J £ J ~j
-------
50. Pate, C.T., R. Atkinson, and J.N. Pitts, Jr. Rate Constants for the Gas
Phase Reaction of Peroxyacetyl Nitrate with Selected Atmospheric Con-
stituents. J. Environ. Sci. Health, All(1):19-31, 3976.
51. Hansen, D.A., R. Atkinson, and J.N. Pitts, Jr. Rate Constants for the
Reaction of OH Radicals with a Series of Aromatic Hydrocarbons. J. Phys.
Chem., 79(17) -.1763-1766, 1975.
52. Schwartz, W. Chemical Characterization of Model Aerosols. EPA-650/3-
74-011. U.S. Environmental Protection Agency, Research Triangle Park,
N.C., 1974.
53. Akimoto, H. Photo-Oxidation of Toluene-NO -O -N System in Gas Phase.
Z, £ £
Presented at the 12th Informal Conference on Photochemistry, National
Bureau of Standards, Washington, D.C., 197(5. (Proceedings to be pub-
lished) .
54. Myrabo, L.N., K.R. Wilson, and J.C. Trijonis. Survey of Statistical
Models for Oxidant Air Quality Prediction. Assessing Transportation-
Related Air Quality Impacts, pp. 46-62. Special Report 167. National
Academy of Sciences, Washington, D.C., 1976.
55. Daniel, W.A., and J.M. Heuss. Ambient Air Quality and Automotive Emis-
sion Control. JAPCA, 24 (9):849-854, 1974.
56. deNevers, N., and J.R. Morris. Rollback Modeling: Basic and Modified.
JAPCA, 25(9):943-947, 1975.
57. Environmental Protection Agency. Air Quality Criteria for Nitrogen
Oxides. Publication Number AP-84. U.S. Environmental Protection Agency,
Washington, D.C., 1971.
58. Federal Register. 36(158):15489, August 14, 1971.
1.07
-------
59. Schuck, E.A. , A.P. Altshuller, D.S. Earth, and G.B. Morgan. Relationship
of Hydrocarbons to Oxidants in Ambient Atmospheres. JAPCA, 20(5):297-
302, 1970.
60. Merz, P.H., L.J. Painter, and P.R. Ryason. Aerometric Data Analysis —
Time Series Analysis and Forecast and an Atmospheric Smog Diagram. Atmos.
Environ., 6 (5):319-342, 1972.
61. Kinosian, J.R., and J.J. Paskind. Hydrocarbons, Oxides of Nitrogen, and
Oxidant Trends in the South Coast Air Basin, 1963-1972. Internal Working
Paper. California Air Resources Board, Division of Technical Services,
1973.
62. Trijonis, J.C. Economic Air Pollution Control Model for Los Angeles
County in 1975. Environ. Sci. Technol., 8 (9):811-826, 1974.
63. Trijonis, J.C., T. Perg, G. McRae, and L. Lees. Oxidant and Precursor
Trends in the Metropolitan Los Angeles Region. Int. Conf. Ox. Poll.,
Proc. 2:1077-1093. EPA-600/3-77-001b, U.S. Environmental Protection
Agency, Research Triangle Park, N.C., 1977.
64. Martinez, E.L., N.C. Possiel, E.L. Meyer, L.G. Wayne, K.W. Wilson,
and C.L. Boyd. Trends in Ambient Levels of Oxidant and Their Possible
Underlying Explanations. Int. Conf. Ox. Poll., Proc. 2:1103-1112. EPA-
600/3-77-OOlb, U.S. Environmental Protection AGency, Research Triangle
Park, N.C., 1977.
65. Caporaletti, J.M., L.B. Myrabo, P. Schleifer, A. Stanonik, and K.R.
Wilson. Statistical Oxidant Air Quality Prediction Model for Land
Use and Transportation Planning. Atmos. Environ., 11 (5):449-458, 1977.
66. Lamb, R.G. Continued Research in Mesoscale Air Pollution Simula
-------
67. Reynolds, S.D. The Systems Applications, Incorporated Urban Airshed
Model: An Overview of Recent Developmental Work. Int. Conf. Ox. Poll.,
Proc. 2:795-802, EPA-600/3-77-001b, U.S. Environmental Protection Agency,
Research Triangle Park, N.C., 1977.
68. Seinfeld, J.H. Accuracy of Prediction of Urban Air Pollution Concen-
trations by Diffusion Models. Assessing Transportation-Related Air
Quality Impacts, pp. 34-45. Special Report 167. National Academy of
Sciences, Washington, D.C., 1976.
69. Kauper, E.K., and B.L. Niemann. Transport of Ozone by Upper-Level Land
Breeze — An Example of a City's Polluted Wake Upwind from its Center.
Int. Conf. Ox. Poll., Proc. 1:227-236. EPA-600/3-77-001a, U.S. Environ-
mental Protection Agency, Research Triangle Park, N.C., 1977.
70. White, W.H., D.L. Blumenthal, J.A. Anderson, R.B. Husar, and W.E. Wilson,
Jr. Ozone Formation in the St. Louis Urban Plume. Int. Conf. Ox. Poll.,
Proc. 1:237-247. EPA-600/3-77-001a, U.S. Environmental Protection Agency,
Research Triangle Park, N.C., 1977.
71. Siple, G.W., C.K. Fitzsimmons, K.F. Zeller, and R.B. Evans. Long Range
Airborne Measurements of Ozone off the Coast of the Northeastern United
States. Int. Conf. Ox. Poll., Proc. 1:249-258. EPA-600/3-77-001a, U.S.
Environmental Protection Agency, Research Triangle Park, N.C., 1977.
72. Hester, N.E., R.B. Evans, F.G. Johnson, and E.L. Martinez. Airborne
Measurements of Primary and Secondary Pollutant Concentrations in the
St. Louis Urban Plume. Int. Conf. Ox. Poll., Proc. 1:259-274, EPA-600/
3-77-OOla, U.S. Environmental Protection Agency, Research Triangle i/drk,
N.C., 1977.
L09
-------
73. Antell, M. A Mechanism Accounting for the Production of Ozone in Rural
Polluted Atmospheres. Int. Conf. Ox. Poll., Proc. 1:439-450. EPA-600/3-
77-OOla, U.S. Environmental Protection Agency, Research Triangle Park,
N.C., 1977.
74. Chang, T.Y., and B. Weinstock. Net Ozone Formation in Rural Atmospheres.
Int. Conf. Ox. Poll., Proc. 1:451-466. EPA-600/3-77-001a. U.S. Environ-
mental Protection Agency, Research Triangle Park, N.C., 1977.
75. Graedel, T.E., and D.L. Allara. The Kinetic Ozone Photochemistry of
Natural and Perturbed Nonurban Tropospheres. Int. Conf. Ox. Poll.,
Proc. 1:467-473. EPA-600/3-77-001a, U.S. Environmental Protection
Agency, Research Triangle Park, N.C., 1977.
76. Glasson, W.A., and P.H. Wendschuh. Multiday Irradiation of NO -Organic
X
Mixtures. Int. Conf. Ox. Poll., Proc. 2:677-686. EPA-600/3-77-001b.
U.S. Environmental Protection Agency, Research Triangle Park, N.C., 1977.
77. Pitts, J.N., Jr., and B.J. Finlayson. Mechanisms of Photochemical Air
Pollution. Angew and te Chemie Intl., 14(1):1-15, 1975.
78. Busch, G.E., and K.R. Wilson. Triatomic Photofragment Spectra: I.
Energy Partition
3626-3638, 1972.
Energy Partitioning in NO Photodissociation. J. Chem. Phys., 56(7):
79. Sklarew, R.C. The Use of Airshed Models in Developing and Testing
Control Strategies. Report on 1974 UC-ARB Conference on Technical Bases
for Control Strategies of Photochemical Oxidant: Current Status and
Priorities in Research. Statewide Air Pollution Research Center, U. of
Calif., Riverside, Calif., 1976.
80. Taylor, G.H., and A.Q. Eschenroeder. Tests of the DIFKIN Photochemical/
Diffusion Model Using Los Angeles Reactive Pollutant Program Data. Int.
Conf. Ox. Poll., Proc. 2:817-825. EPA-600/3~77~001b, U.S. Environmout-al
Protection Agency, Research Triangle Park, N,C., L977.
-------
81. Eschenroeder, A. Discussion. In: Assessing Transportation-Related Air
Quality Impacts, pp. 89-92. Special Report 167. National Academy of
Sciences, Washington, D.C., 1976.
82. Eschenroeder, A. The Use of Airshed Models in Developing and Testing
Control Strategies. Report of the 1974 UC-ARB Conference on Technical
Bases for Control Strategies of Photochemical Oxidant: Current Status
and Priorities for Research. Statewide Air Pollution Research Center,
U. of Calif., Riverside, Calif., 1976.
83. Calvert, J.G. Test of the Theory of Ozone Generation in Los Angeles
Atmosphere. Environ. Sci. Technol., 10(3):248-256, 1976.
84. Hegg, D., P.V. Hobbs, L. Radke, and H. Harrison. Ozone and Nitrogen
Oxides in Power Plant Plumes. Int. Conf. Ox. Poll., Proc. 1:173-184.
EPA-600/3-77-001a. U.S. Environmental Protection Agency, Research
Triangle Park, N.C., 1977.
85. Conn, W.D. The Difficulty of Forecasting Ambient Air Quality — A
Weak Link in Pollution Control. J. Amer. Inst. Planners, 41(5):334-346,
1975.
86. Dabbert, W.F., and H.B. Singh. A Preliminary Investigation of the
Effectiveness of Air Pollution Emergency Plans. Int. Conf. Ox. Poll.,
Proc. 2:837-848. EPA-600/3-77-001b. U.S. Environmental Protection Agency,
Research Triangle Park, N.C., 1977.
87. Farrow, L.A., T.E. Graedel, and T.A. Weber. Urban Kinetic Chemistry
under Altered Source Conditions. Int. Conf. Ox. Poll., Proc. 1:1.37-144.
EPA-600/3-77-001a. U.S. Environmental Protection Agency, Research
Triangle Park, N.C., 1977.
88. Myrabo, L.N., P. Schleifer, and K.R. Wilson. Oxidant Prediction Model
for Land Use and Transportation Planning. California Air Environment,
4(3):3-5, 1974.
Ill
-------
89. Reynolds, S.D., and J.H. Seinfeld. Interim Evaluation of Strategies for
Meeting Ambient Air Quality Standard for Photochemical Oxidant. Environ.
Sci. Technol., 9(5):433-447, 1975.
90. Wilson, K.R. Multiprocessor Molecular Mechanics. In: Computer Net-
working and Chemistry, P. Lykos, ed. American Chemical Society, Washing-
ton, D.C., 1975.
91. Yasaki, E.K. Super-Super Computer. Datamation, 23(3):150-152, 1977.
92. Lees, L., M. Braly, M. Easterling, R. Fisher, K. Hestner, J. Henry, P.J.
Home, B. Klein, J. Krier, W.D. Montgomery, G. Pauker, A. Rubenstein,
and J. Trijonis. Smog, A Report to the People. California Institute of
Technology, Environmental Quality Laboratory, Pasadena, Calif., 1972.
93. Boris, J.P., and D.L. Book. Flux Corrected Transport. I. Shasta, A
Fluid Transport Algorithm That Works. J. Computational Physics, 11(1):
38-69, 1973.
94. Reynolds, S.D., M.K. Liu, T.A. Hecht, P.M. Roth, and J.H. Seinfeld. Urban
Airshed Photochemical Simulation Model Study: Volume I. Development and
Evaluation. Systems Applications, Inc., (EPA Contract No. 68-02-0339).
EPA-R4-73-030a. U.S. Environmental Protection Agency, Researcti Triangle
Park, N.C., 1973.
95. Reynolds, S.D., P.M. Roth, and J.H. Seinfeld. Mathematical Modeling of
Photochemical Air Pollution: I. Formulation of the Model. Atrnos,
Environ., 7 (11):1033-1061, 1973.
96. Roth, P.M., P.J.W. Roberts, M.K. Liu, S.D. Reynolds, and J.H. SeinfeJd.
Mathematical Modeling of Photochemical Air Pollution: II. A Model
and Inventory of Pollutant Emissions. Atmos. Environ., 8(2):97-130,
1974.
112
-------
97. Reynolds, S.D., J. Ames, T.A. Heoht, J.P. Meyer, ad D.C. Whitney.
Continued Research in Mesoscale Air Pollution Simulation Modeling:
Volume II. Refinements in the Treatment of Chemistry, Meteorology,
and Numerical Integration Procedures. EPA-600/4-76-016b, Systems Ap-
plications, Inc., San Raefael, Calif., 1976.
98. Reynolds, S.D., et al. Continued Development and Validation of a Second
Generation Photochemical Air Qualit^ Simulation Model: Volume I.
Refinements in the Treatment of Meteorology, Chemistry, Pollutant
Removal Processes, and Numerical Analysis. Final report for EPA Contract
No. 68-02-2216. Systems Applications, Inc., San Raefael, Calif. (In
preparation.)
99. Hindmarsh, A.C. GEARB: Solution of Ordinary Differential Equations
Having Banded Jacobian. Lawrence Livermore Laboratory Report UCID-
30059, Rev. 1. 1975. 53 pp.
100. Edelson, D. A Simulation Language and Compiler to Aid Computer Solution
of Chemical Kinetic Problems. Computer and Chem., l(l):29-33, 1976.
101. Martinez, J.R., R.A. Nordsieck, and M.A. Hirschberg. User's Guide to
Diffusion/Kinetics (DIFKIN) Code. Final Report on Contract 68-02-
0036, EPA-R4-73-012B. General Research Corporation, 1973.
102. Deardorff, J.W. A Three-Dimensional Numerical Investigation of the
Idealized Planetary Boundary Layer. Geophys. Fluid Dyn., 1(4):377-410,
1970.
103. Lamb, R.G., W.H. Chen, and J.H. Seinfeld. Numerico-Empirical Analysis
of Atmospheric Diffusion Theories. J. Atmos. Sci., 32 (9):1794-1807,
1974.
104. Lamb, R.G. Continued Development and Validation of a Second Generation
Photochemical Air Quality Simulation Model: Volume II. MJUM osc'.-'l <•>
Modeling Studies. Final report for EPA Contract No. 68-02-?2l6. Systems
Applications, Inc., San Raefael, Calif. (In preparation.)
113
-------
BIBLIOGRAPHY
Chaum, D., B. Elkus, J. Rice, and K.R. Wilson. (To be published in Atmos.
Environ.)
Cox, R.A., Photolysis of Nitrous Acid in the Presence of Carbon Monoxide and
Sulfur Dioxide. J. Photochem. , 3(4):291-304, 1974.
Glahn, H.R., and D.A. Lowry. The Use of Model Output Statistics (MOS) in
Objective Weather Forecasting. J. Appl. Meteor., 11(8):1203-1211, 1972.
Morris, E.D., Jr., and H. Niki. Reaction of Dinitrogen Pentoxide with Water.
J. Phys. Chem., 77(16):1929-1932, 1973.
Simonaitis, R., and J. Heicklen. Reactions of HO with NO and NO and of
OH with NO. J. Phys. Chem., 80(1):l-7, 1976.
115
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO.
EPA-600/3--77-118
4. TITLE AND SUBTITLE
INTERNATIONAL CONFERENCE ON OXIDANTS, 1976 -
ANALYSIS OF EVIDENCE AND VIEWPOINTS
Part VI. The Issue of Air Quality Simulation
Model Utility
6. PERFORMING ORGANIZATION CODE
3. RECIPIENT'S ACCESSION-NO.
5. REPORT DATE
November 1977
7. AUTHOR(S)
1. John H. Seinfeld
2. Kent R. Wilson
8. PERFORMING ORGANIZATION REPORT NO.
9 PERFORMING ORGANIZATION NAME AND ADDRESS
1. California Inst. of Technology, Pasadena, CA.
2. Univ. of California-San Diego, La Jolla, CA.
10. PROGRAM ELEMENT NO.
1AA603 AJ-13 (FY-76)
11. CONTRACT/GRANT NO.
1. DA-7-2143A
2. DA-7-2191.T
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTF, NC
Office of Research and Development
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Fdnal
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
Partially funded by the Office of Air Quality Planning and Standards.
16. ABSTRACT
In recognition of the important and somewhat controversial nature of the
oxidant control problem, the U.S. Environmental Protection Agency (EPA) organized
and conducted a 5-day International Conference in September 1976. The more than
one hundred presentations and discussions at the Conference revealed the existence
of several issues and prompted the EPA to sponsor a followup review/analysis effort.
The followup effort was designed to review carefully and impartially, to analyze
relevant evidence and viewpoints reported at the International Conference (and
elsewhere), and to attempt to resolve some of the oxidant-related scientific issues.
The review/analysis was conducted by experts (who did not work for the EPA or for
industry) of widely recognized competence and experience in the area of photo-
chemical pollution occurrence and control.
John H. Seinfeld, California Institute of Technology, and Kent R. Wilson,
University of California at San Diego, review the issue of Air Quality Simulation
Model (AQSM) utility. The strengths and weaknesses of the various modeling
techniques are discussed, and the authors offer their recommendations on future
studies.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
* Air pollution
* Computerized simulation
* Utilization
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
EPA Form 2220-1 (9-73)
b.IDENTIFIERS/OPEN ENDED TERMS
19 SECURITY CLASS (This Report)
_UNCLASSIF:IED
20 SECURITY CLASS (This p-jye)
„JttJOEASSJLEIW?- .... „ _
116
c. COSATI Field/Group
13B
14B
21. NO. OF PAGES
124_._
22. PRICE
-------
Q
<
Q_
CO
UJ
UJ
Q_
Q
Z
<
UJ
0
|_
CO
O
0_
u
z
UJ
0
it m
n n
-1 <
< Q.
H UJ
Z
UJ
5
Z
O
(t
>
Z
UJ
CO
D
11)
+^
(0
DC'
in
in
«
U
1
X
•+-"
l_
3
O
U.
(0
o
^ S
O 0) c ™
UJ a o £
O ? ro 2
DC 03 E sz
°- ^ o O
-1 i£ "c •-
^ CO — "CO
^
------- |