&EPA
United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/4-82-009
June 1982
Air
An Evaluation of
the Empirical Kinetic
Modeling Approach
Using the St. Louis RAPS
Data Base
-------
EPA-450/4-82-009
An Evaluation of
the Empirical Kinetic
Modeling Approach Using
the St. Louis RAPS Data Base
By
Gerald L. Gipson
Air Management Technology Branch
Monitoring and Data Analysis Division
U.S ENVIRONMENTAL PROTECTION AGENCY
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
June 1982
-------
This report has been reviewed by the Office of Air Quality Planning and Standards, EPA, and approved for
publication. Mention of trade names or commercial products is not intended to constitute endorsement or
recommendation for use.
Publication No. EPA-450/4-82-009
-------
TABLE OF CONTENTS.
Page
Preface • • • v
List of Tables vi
List of Figures ..' vii
Executive Summary , vi i i
1.0 Introduction 1
1.1 Background 3
1.1.1 OZIPP Trajectory Model ... 5
1.1.2 The EKMA Technique 9
1.2 RAPS Data Bas.e 13
2.0 Level II Analysis. f 17
2.1 Methodology for Developing Input Data 17
2.1.1 Air Parcel Trajectory 17
2.1.2 Initial Concentrations 19
2.1.3 Emissions 19
2.1.4 Boundary Conditions 23
2.1.5 Dilution i 23
2.1.6 Chemical Mechanism 24
2.2 Peak Ozone Predictions 24
2.3 Sensitivity Tests. 30
3.0 Level III Analysis 41
3.1 Model Input Data 41
3.3 Sensitivity Analysis 45
4.0. Comparisons of EKMA with a PAQSM 49
4.1 Description of the Urban Airshed Model 52
4.2 Summary of Airshed Model Simulations 54
4.2.1 Base Case Simulations 54
4.2.2 Simulations of Changes in VOC Emissions 51
4.3 Comparison of EKMA wi th Airshed 65
4.3.1 Independent Model Tests 65
4.3.2 Common Basis, Tests 69
-------
TABLE OF CONTENTS (Continued)
Page
5.0 Conclusions and Recommendations 73
6.0 References 77
Appendi x A A-1
Appendix B B-l
Appendix C C-l
Appendix D D-l
iv
-------
PREFACE
The Urban Airshed Model Simulations described in tins ivport weiv
conducted by the Source Receptor Analysis Branch (SRAG) of the Office of Air
Quality Planning and Standards in EPA. The author gratefully acknowledges th?
invaluable assistance of Conrad Newberry and Gerald Moss of SRAB in providing
the Airshed Modeling information needed to conduct this project. In addition,
the author extends special thanks to Robert Kelly, Norman Possiel and Edwin L.
Meyer for their numerous contributions and valuable suggestions. Finally, the
author is especially grateful to Mrs. Carole Mask for typing and editing the
final report.
-------
LIST OF TABLES
Table No. Page
1-1
1-2
2-1
2-2
4-1
4-2
4-3
4-4
Mathematical Representation of the OZIPP Trajectory
Model
Example Emission Reduction Calculation Using a Single
Ozone Isopleth Diagram
Model Test Cases
Example Sensitivity to Mixing Height Profile
Summary of Miscellaneous Measurements and Airshed Model
Inputs
6-9 a.m. Urban Core Precursor Predictions Versus
Observati ons
Emission Reductions Needed to Reduce Peak 03 to 0.12 pptn
Emission Reductions Needed to Lower Peak Os to 120 ppb .
8
1?
18
39
%
60
68
72
-------
LIST OF FIGURES
Figure No. Page
1-1 Example Ozone Isopleth Diagram 4
1-2 Conceptual View of the OZIPP Trajectory Model 6
1-3 Illustration of EKMA Procedure 11
1-4 RAMS Station Locations 14
2-1 Air Parcel Trajectory for July 19 Test Case 20
2-2 Air Parcel Trajectory for July 19 Test Case Demonstrating
How Fresh Precursor Emissions are Considered 22
2-3 Location of Upper Air Network Stations 25
2-4 Level II Predictions Versus Observations of Peak Ozone . 26
2-5A Comparison of Trajectories Calculated by Alternative
Methods: July 19 (Day 201) 32
2-5B Comparison of Trajectories Calculated by Alternative
Methods: October 1 (Day 275) 33
2-6 Radiosonde Measurements of Wind Direction Aloft for
June 7 (Day 159) 35
2-7 Trajectory for June 7 Derived From Surface and Aloft
Wind Data 36
2-8 Graphical Depiction of Characteristic Curve 38
2-9 Sensitivity of Level II Model Predictions to Emissions
Inventory Spatial Resolution 40
3-1 Model Predictions Versus Observations of Peak Ozone
(Level III) 43
3-2 Sensitivity of Level III Model Predictions to Mixing
Heights 46
3-3 Sensitivity of Level III Model Predictions to Post-0800
Emissions and Initial Concentrations 48
4-1 St. Louis Modeling Region 53
4-2 Airshed Predictions Versus Observations of Renional
Peak Ozone 57
4-3 Airshed Predictions Versus Observations of Peak Ozone .. 59
4-4 Summary of Airshed Model Simulations 62
4-5 Airshed Model Sensitivity of Regional Peak Ozone to
Hydrocarbon Reductions 64
4-6 Comparison of Airshed and EKMA (Independent Tests) 66
4-7 Comparison of Airshed and EKMA (Common Basis Tests) .... 71
VII
-------
EXECUTIVE SUMMARY
The Empirical Kinetic Modeling Approach (EKMA) is a technique developed
for estimating the degree of emission reduction necessary to achieve the ozone
National Ambient Air Quality Standard (NAA'QS). The objective of this study has
been to investigate several approaches for assessing the validity of the
relationships underlying EKMA. The most direct means of validating the model
is to compare changes in ozone levels that result from changes in precursor
emis.ai.ons. to thos.e predi.cted by the model. Such an approach requires the
existence of well documented trends in both ozone and its precursors over a
period of time during which precursor emissions have been appreciably altered.
Such data, especially trends in precursors, are generally not available over a
time period of sufficient interest to make such comparisons. As a result,
three indirect methods for evaluating EKMA have been explored.
The first method entails using the sample, Lagrangian model underlying
EKMA to_raake predictions of peak ozone levels on a given day, and comparing
those predictions to observations. For this approach, detailed meteorological,
emissions, and air quality data compiled under the Regional Air Pollution Study
(RAPS) in St. Louis, Missouri, were used to develop the model inputs. This
approach does not answer the key question of how accurately EKMA predicts
changes in peak ozone accompanying reductions in precursors. However, success-
ful prediction of observed peaks gives some confidence that the model provides
a reasonable approximation of the physical and chemical phenomena leading to
ozone formation. This, in turn, should provide greater confidence in using the
model to make predictions of changes in ozone occurring because of changes in
precursors. The results of using this approach are not particularly encouraging.
Substantial underpredictions of peak ozone occurred in a number of cases.
However, a detailed assessment of the results suggested that performance could
be improved in some cases with revised model inputs. Thus, it is not clear
from these results whether the tendency towards underprediction is due to the
modeling concepts, or due to uncertainties in the model inputs.
-------
The second method evaluated is. yery similar to the one just described.
Again, the model underlying EKMA is used to make predictions of peak ozone, and
these are then compared to observed levels. In this approach, however, a less
complex data base is used, with certain simplifying assumptions made in the
formulation of the model inputs. Because of limited data, this level of
analysis is the more commonly used method of applying EKMA. As with the first
approach, the primary question of how well EKMA predicts changes in peak ozone
Is. not directly addressed, but this approach does give some indication of how
well the model simulates ozone formation. The results found with this approach
are much, better than those found using the first approach. In nearly every
case, predicted peak ozone concentrations are within ±30% of observations.
Furthermore, sensitivity tests indicate that this agreement is not substanti-
ally altered by variations in selected input variables. These results suggest
that the ozone formation processes are reasonably simulated by the model.
Therefore, the possibility exists that the simplified approach may "smooth out"
some anomalies, associated with the development of the model inputs using the
more rigorous approach. However, because of the simplifying assumptions used
in deriving the model inputs, the possibility that agreement between observations
and predictions may be somewhat fortuitous cannot be entirely ruled out.
The final approach attempts to address more directly the question of how
well EKMA predicts changes in peak ozone. In this approach, a number of
emission reduction scenarios are simulated with a complex Photochemical Air
Quality Simulation Model (PAQSM). These results form the basis for evaluating
corresponding EKMA predictions. An important limitation of such an evaluation
is that no absolute guarantee exists that the PAQSM predictions are correct.
Nevertheless, the PAQSM does represent the state-of-the-art in photochemical
modeling and, as such, serves as one means of testing simpler models. However,
making such comparisons is complicated by differences in the mode of applica-
tion of each type of model. EKMA did not agree precisely with the PAQSM in
every case. However, when estimates of the degree of control needed to reduce
peak ozone to 120 ppb (the level of the ozone NAAQS) were made for each of
three days with EKMA and the PAQSM, differences between the estimated emission
reductions were less than about 10%. Using the PAQSM as the standard, EKMA
was not found to systematically over or underestimate required emission reductions
-------
The findings of the above three approaches, for evaluating EKMA have led to
several recommendations for future evaluation. First, the trajectory model
underlying EKMA should be compared with more complex trajectory jnodels. Also,
methodologies for formulating the trajectory model inputs need more study.
Additional tests should be conducted to identify potential differences in model
predictions that could occur as the result of using various chemical mechanisms.
Finally, the relative Importance of boundary condition assumptions in the PAQSM
and EKMA simulations need additional study.
-------
1.0 INTRODUCTION
The Empirical Kinetic Modeling Approach (EKMA) is a procedure developed
by the U.S. Environmental Protection Agency (EPA) to determine emission
reductions necessary to achieve the National Ambient Air Quality Standard
(NAAQS) for ozone.1,2 EKMA mathematically relates peak ozone concentrations
to its precursors - nonmethane organic compounds (NMOC) and oxides of nitrogen
(NO ). These relationships form the basis for estimating reductions in
/\
precursor emissions needed to attain a prescribed reduction in peak ozone
concentrations. The objective of this study has been to investigate several
approaches for assessing the validity of the 03-precursor relationships
generated by the model underlying EKMA.
When any photochemical model is used to predict changes in ozone levels
resulting from changes in precursor emissions, the most direct means of
validating that model is to compare the predicted change to that which has
actually taken place. Such an approach requires the existence of well docu-
mented trends of both ozone and its precursors over a sufficient period of
time during which precursor emissions have been appreciably altered. For many
areas of the country, sufficient data are available to establish the trends in
ozone over a number of years, but changes in precursors that have taken place
during that same time are not as well known.3,4 This uncertainty in precursor
trends often introduces large uncertainties in comparing the ozone trends
predicted by a photochemical model with those that have been observed.
Furthermore, little, if any, ozone trend data are available for areas of the
country which have moved from a position of a significant ozone problem to one
of compliance with the ozone NAAQS, or vice versa. As a result, the data
necessary to validate a model over the range for which it is to be used are
not available.
The aforementioned problems associated with direct validation of
photochemical models have led to the development of some indirect approaches
for assessing the validity of EKMA. The first approach consists of using
detailed meteorological, emissions, and air quality data with EKMA to predict
observed ozone peaks, rather than changes in peak ozone levels as is usually
done. The use of detailed data is commensurate with a Level II analysis
-------
described in Reference 5, and the technique of comparing predicted and observed
ozone peaks is similar in concept to the approach normally used to validate
more complex photochemical models. Such an approach does not directly answer
the key question of how accurately EKMA predicts changes in peak ozone accom-
panying reductions in precursors. However, successful prediction of observed
peaks gives some confidence that the model provides a reasonable approximation
of the physical and chemical phenomena leading to ozone formation. This, in
turn, leads to greater confidence in using the cause effect relationships
contained with/in the model to make predictions of changes in ozone occurring
as a result of changes in precursors.
The second approach investigated for assessing EKMA's validity is nearly
identical to the first in that model predictions of peak ozone are compared to
observed levels. The difference results from using a less complex data base
to formulate model inputs. This approach was taken because the detailed data
of the type needed for a Level II analysis are not available for many areas of
the country. To test the potential effects of limited data on the predictions
of peak ozone, the model inputs used in the Level II analysis were modified
such that they are consistent with those required for a Level III analysis, as
described in Reference 5. Once again, the comparison of predicted peak ozone
levels, with those observed does not address the primary question of how well
th.e modeling approach predicts changes in peak ozone, but provides some indi-
cation of how well the model represents ozone formation phenomena when a much
more limited data base is used.
The final approach attempts to address more directly the problem of how
well EKMA predicts changes in peak ozone resulting from changes in precursors,
but with some important limitations. In this approach, the EKMA predictions
of changes in peak ozone are compared to the predictions of a detailed Photo-
chemical Air Quality Simulation Model (PAQSM). PAQSMs have been identified as
having the greatest potential for estimating the changes in ozone levels
resulting from changes in precursor levels because of their detailed mathe-
matical representation of chemical and physical processes leading to ozone
formation.1 For this third approach, a number of control strategies in which
precursors, are changed from existing, or base case, levels are simulated with
-------
the PAQSM and then replicated with the simpler EKMA model. The effects of the
control strategies predicted by each model are then compared for consistency.
Presumably, the more complex PAQSM should provide the best estimate of the
effects of changing precursor levels because of its more detailed nature, and
therefore serves as the basis for evaluating the predictions of the simpler
EKMA. One important limitation of such an analysis, however, is that no
absolute assurance exists that the PAQSM predictions are correct. Neverthe-
less, the PAQSM does provide state-of-the-art estimates of the effectiveness
of potential control programs and, as such, serves as one means of testing
simpler approaches.
The data base used to conduct the three analyses just described was
compiled during the Regional Air Pollution Study (RAPS) in the Metropolitan
St. Louis Area.6 This particular data base was selected because of the
detailed information collected, and because a complex PAQSM had been applied
using the data base. With this data base, sufficient information exists not
only to apply the different types of models, but it also affords the oppor-
tunity to examine several facets of model performance, especially those
related to the models' capabilities to simulate existing ozone levels.
The remainder of Section 1 provides background information on EKMA and a
more detailed description of the data base used in the study. Sections 2 and
3 describe the Level II and Level III analyses where predictions of peak ozone
are compared to measured levels. The comparisons between the PAQSM and EKMA
are discussed in Section 4. Finally, conclusions and recommendations arising
from the findings of all three analyses are presented in Section 5.
1.1 Background
The application of EKMA involves the use of an ozone isopleth diagram, an
example of which is shown in Figure 1-1. In such a diagram, peak hourly
average ozone concentrations are plotted as a function of early morning
precursor levels. To estimate the emission reductions needed to achieve the
ozone NAAQS, the diagrams are used in conjunction with ambient measurements of
ozone and precursor concentrations. The diagrams themselves are generated by
means of the Kinetics Model and Ozone Isopleth Plotting Package (OZIPP)/,3
-------
5 0.2 0.4 0..6 0.8 1,0 1,2 1,4 1.0 1,0
0.2 0.4 0.0 0.0 1.0 1.2 1-4 1.0 1.0 2.0
NMHC.PfnC
Figure 1-U Example Ozone Isopleth Diagram0
-------
This computerized program produces an isopleth diagram on the basis of a
simplified trajectory model which mathematically simulates ozone formation.
In so doing, the trajectory model takes into account the effects of emissions,
meteorology and pollutants which may have been transported into an area.
Thus, while the isopleths are plotted as a function of early morning precursor
levels, the positioning of the isopleths of the diagram are influenced by
these other variables as well.
Of the three approaches taken for evaluating EKMA that were previously
described, two involve making predictions of peak ozone, and in the other,
EKMA is used to make predictions of changes in ozone from some existing, or
base case condition. Thus, the first two involve testing the trajectory model
underlying EKMA for its ability to predict peak ozone levels that have been
observed. The third approach, predicting changes from base case conditions,
consists of actually using the isopleth diagram and the EKMA technique. While
both have been discussed extensively elsewhere,1,2,8,9 they are briefly
described below.
1.1.1 OZIPP Trajectory Model
The conceptual basis for the simple trajectory model in OZIPP is
similar to a Lagrangian photochemical dispersion model (see Figure 1-2). A
column of air is advected along a trajectory by the wind. The height of the
column is equal to the mixing height (i.e., the column extends from the
earth's surface throughout the mixed layer). The horizontal dimensions are
selected such that concentration gradients are small, and thus the effects of
horizontal exchange of air between the column and its surroundings can be
ignored. Within the column, the air is assumed to be uniformly mixed at all
times.
Initially, the column contains NMOC, NO and possibly ozone resulting
/\
from prior emissions and/or possible transport from upwind areas. As the
column moves along the trajectory, the height grows in accordance with the
temporal variation in mixing height. As the mixing height increases, air
above the column is mixed downwind into the column instantaneously, resulting
in two phenomena. First, pollutants within the column are diluted due to the
-------
INITIAL
CONDITIONS
KEY
= TIME PERIOD 1
= MIXING HEIGHT AFTER TIME i
= PRECURSORS INJECTED INTO
COLUMN DURING TIME i
= SUNLIGHT INTENSITY DURING
TIME i
Figure 1-20 Conceptual View of the OZIPP Trajectory Model
-------
increased volume of the column. Second, entrainment of pollutants from aloft
may occur. Several studies have documented that, under certain conditions,
significant pollutant concentrations (especially ozone) may exist above the
early morning mixed layer.10,11,12 The trajectory model simulates the effect
of entrainment by assuming constant pollutant concentrations aloft, and by
assuming their entrainment into the column at a rate proportional to the
growth of the mixed layer. In addition to these two effects, the trajectory
model also simulates the injection of fresh precursor emissions into the column,
which may occur as it moves along the trajectory. Thus, while pollutant
concentrations within the column are diluted by the growth of the mixed layer,
this effect may be offset somewhat by the entrainment of pollutants from aloft
and by the injection of fresh precursor emissions.
In addition to the physical processes described above, the OZIPP trajectory
model simulates the chemical interactions taking place among the pollutants
within the column. The latter is accomplished by means of a chemical kinetic
mechanism which describes the chemical reactions and their corresponding rates
of reaction. For those reactions affected by sunlight, the rate constants are
determined by a theoretical estimation of the diurnal variation in solar
radiation. Thus, the trajectory model calculates the instantaneous concentra-
tions of all reactive species included within the mechanism as a function of
time. From the time profile of ozone within the column, the model calculates
the maximum one-hour-average concentration occurring during the simulation.
The mathematical representation of all the aforementioned processes is
summarized in Table 1-1.
The application of the trajectory model in OZIPP requires the extraneous
development of a number of specific model inputs. First, the trajectory path
which depicts the column movement through time and space must be defined. Once
this is done, the pollutant concentrations initially within the column (i.e.,
at the trajectory starting point) can be approximated, and the emissions
occurring along the path can be estimated. Additionally, the growth in the
mixed layer and the concentrations of pollutants above the early morning mixed
layer must be estimated. The procedures used for developing these inputs for
the Level II and Level III analyses are described in Sections 2 and 3,
-------
Table 1-1. Mathematical Representation of the OZIPP Trajectory Modela.
C. = the concentrations of the ith chemical species in the mixed volume
1 at time = t
C.(@0) = the initial concentration of the ith species in the mixed volume
1 at time = 0
c aloft _ tne concentration of the ith species above the mixed volume
i
MH = f(t) mixing height at time = t
~- = f'(t) rate of rise at time = t
KD =
dilution rate at time = t
dt
|dC,l
l-rz—f - -Kn. C- decrease in species i caused by mixing height rise
lat J di1
\AC 1
aloft .
]—nrf = + 1C, C.a ° increase in species i caused by entrainment due to
L. Jent mixing height rise
E. = h(t) mass of species C. emitted per unit .time at time
= t (concentration - height units)
dC 1
U = E./MH increase in species C- caused by emissions into a
^•"emis mixing volume with unit area and height, MH, at time t
func (C.,j=l,n) production of species C. due to chemical formation at
J time t
L. = func (C-,j=l,n) loss of species C.. due to chemical reaction at time t
•r} = P^- L^ increase or decrease in species C, caused by chemical
-•chem reactions of all species at time t
P.
dC.
TOTAL Ldtldil ldtlent ' I dtUis ' i dtfchem
SOLUTION:
t fnr ~l
C, = C.(@0) + ; J i i numerical integration used to solve initial
o I dtj TOTAL value, differential equation problem
a adopted from Reference 9
-------
respectively. Finally, the OZIPP trajectory model internally generates the
information necessary for the chemical mechanism, with only the date of the
simulation input in order to derive the appropriate photolytic rate constants.
1.1.2 The EKMA Technique
Following the procedure described in the previous section will lead
to the prediction of one time/concentration profile of ozone, from which a
single, peak, one-hour average ozone concentration can be calculated. More
often than not, the predicted peak level will not agree precisely with the
observed ozone concentration. Such disagreement may be due in part to uncertain
ties in the model inputs, and to the simplistic manner in which the trajectory
model simulates ozone formation. (Recall that a number of simplifying assump-
tions were made in the formulation of the model, including the existence of a
well defined column or air parcel, no horizontal dispersion, instantaneous
mixing, etc.) The EKMA technique circumvents the potential problem of disagree-
ment between a model prediction and the observed peak when the model is applied
for regulatory purposes.
When the OZIPP computer program is used to generate an isopleth
diagram, the same basic information described in the preceding section is
required as input, except for the initial concentrations of NMOC and NO . To
A
generate an isopleth diagram, OZIPP performs repeated simulations with dif-
fering assumed initial concentrations of NMOC and NO . Thus, a peak one-hour
X
average ozone concentration is calculated for each set of initial concentra-
tions, and the results are plotted as ozone isopleths (i.e., all combinations
of NMOC and NO that yield a constant level of ozone are connected by a single
s\
curve, or isopleth). Note that a particular isopleth diagram inherently
incorporates the other information input to the model, i.e., the mixing height
growth, the photolytic reaction rates, the levels of pollutants aloft, and the
emission pattern. (That is, these factors affect the positioning of the
isopleths on the diagram.) The emissions themselves are actually input to the
model as factors relative to initial concentrations. Thus, for example, if the
initial concentrations are decreased by 50% from one simulation to the next,
then the emissions will also be decreased by that same percentage. In effect,
then, the isopleth diagram graphically depicts the model predictions of peak
-------
ozone under varying conditions of precursor initial concentrations and
precursor emissions, with all other variables held constant. As a consequence,
the isopleth diagram can be used to estimate the model predicted effect of
changing precursors from one level to another.
In order to use the isopleth diagram to estimate changes in ozone
resulting from changes in precursors, a starting point must be established on
the diagram to which all changes are related. In the trajectory model appli-
cation, the absolute initial concentrations of NMOC and NO are estimated from
A
available air quality data. If these two values are used to define a starting
point, the corresponding ozone level will more than likely not equal the
measured peak since the model will not always predict the peak ozone precisely.
The EKMA procedure circumvents this problem by using the measured ozone peak
and the measured NMOC/NO ratio to establish the starting point on the diagram.
P\
The starting point itself is the intersection between the NMOC/NO ratio line
X
and the ozone isopleth corresponding to the observed ozone concentration. This
procedure, in essence, calibrates the model to the observations in order to
evaluate changes from existing conditions. Figure 1-3 illustrates the procedure
graphically.
With the starting point defined, the isopleth diagram can be used to
estimate the effects of changing precursors from existing levels, or to esti-
mate to what degree precursor levels must be changed in order to achieve a
prescribed reduction in peak ozone. The latter is illustrated by the example
contained in Table 1-2. Note that the changes in precursors are expressed on
a relative basis, i.e., as a percent change from existing levels. As pre-
viously stated, the precursor changes imply that both initial precursor concen-
trations and subsequent emissions must be changed by the same degree, with all
other variables held constant. As such, the use of a single isopleth diagram
provides estimates of the precursor reductions needed to reduce the observed
ozone level to the desired goal for only those conditions that led to the
observed ozone peak (e.g., dilution, pollutants aloft, sunlight, etc.). For
the example problem shown in Table 1-2, the NMOC reduction of 72% is not
necessarily the reduction needed to achieve the ozone NAAQS, since the latter
is statistical in form and allows, on average, one daily maximum one-hour
10
-------
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1-8
0.2 0.4 0.6
0.8 i'.o r.2
KflHC.PPttC
1.4 1.6 1.8 2.0
Point 1 corresponds to model prediction of peak ozone
Point 2 corresponds to model calibration to observed peak ozone
Figure 1-30 Illustration of EKMA Procedure,
n
-------
Table 1-2, Example Emission Reduction Calculation Using a Single
Ozone Isopleth Diagram,
GIVEN: 03 Daily Design Value 3 .24
Design NMOC/NOX = 8:1
Anticipated Change in NOX = -20%
Base case diagram shown below
FIND: Percent reduction in VOC emissions needed to reduce ozone from .24
to 0.12 ppm
SOLUTION:
as
1.5
0 <
SSSi :«CE SIBGRSr. 31 V.Ct ' . .12
STEP 1: The base case point is found by the intersection of the 8:1 NMOC/NO
ratio line with the .24 ozone isopleth (Point 1) on the diagram. At
Point 1, (NMOC)i = 1.64 and (N0xh = .205
STEP 2: The post-control NOX coordinate is calcualted as follows:
(NOX)2 = (.205) x (1 -
.164
STEP 3: The post-control point is located at the intersection of the .164 NOX
coordinate and the 0.12 ppm ozone isopleth (Point 2). At Point 2,
(NMOC)2 = 0.46
STEP 4: The VOC emission reduction is calculated as
% reduction = (1 - ---) x 100 =
12
-------
average ozone concentration above 0.12 ppm each year.* In the comparisons of
the EKMA procedure with the PAQSM predictions, it must be emphasized that no
attempt has been made to establish the prescribed reduction in precursors
needed to achieve the ozone NAAQS with either model. Rather, attention is
focused on the precursor reductions needed to lower peak ozone on each of chret.
days to a level of 0.12 ppm. This is a subtle, but important point, and will
be addressed more completely in Section 4.
1.2 RAPS Data Base
As previously described, an extensive air quality management data base was
compiled as part of the Regional Air Pollution Study (RAPS). The data base
contains all of the elements needed to evaluate air quality simulation models:
air quality, emissions, and meteorological data sufficiently resolved in space
and time to develop input data and provide measures for evaluating model
performance. The data base as it relates to this study is briefly described
below.
The bulk of the air quality data used in the study was collected at 25
Regional Air Monitoring Stations (RAMS) spaced concentrically throughout the
study region (see Figure 1-4). These stations were located such that they
would not be unduly influenced by any one source or group of sources. At each
station, hourly average concentrations of the following pollutants are available:
ozone, nitrogen dioxide (N02), nitric oxide (NO), total oxides of nitrogen
(NO ), nonmethane organic compounds (NMOC), and carbon monoxide (CO). The size
X
of the network, the quality of the data and the duration of measurements
provided the best available temporal and spatial resolution of ambient
Reference 2 describes the procedure for establishing the precursor
reductions needed to achieve the ozone NAAQS. In essence, a number of
high ozone days are modeled, and the precursor reduction needed to achieve
the ozone NAAQS is selected such that it is consistent with the acceptable
number of ozone peaks above 0.12 ppm, and other possible changes in
existing conditions which may take place (e.g., changes in the levels of
pollutants aloft).
13
-------
Figure 1-40 RAMS Station Locations,
14
-------
pollutant levels necessary to evaluate a model such as the trajectory model
underlying EKMA.
The emissions data employed in this study include an hourly resolved
point and area source emissions inventory for NO , CO and Volatile Organic
X
Compounds 0/OC). The area source emissions were spatially resolved by means
of the RAPS grid system which consists of about 2000 variable-sized grids.13
Thus, estimates of emission rates from both area and point sources are avail-
ab.le by hour and by grid for any day in 1975 and 1976, the principal period
for ambient measurements.
The primary source of meteorological data consisted of continuous
measurements of wind speed, wind direction and temperature made at each of the
25 RAMS. Additional data were collected from the RAPS Upper Air Sounding
Network (UASN). In this program, radiosondes were conducted three times per
day, fiye times per week, at a minimum of two stations. These soundings
furnished vertical temperature profiles from which mixing heights could be
estimated. The latter were supplemented by data collected in the early morning
hours from the operation of sodar instruments.
The data base compiled under the RAPS program has been used by EPA in an
in-depth study to evaluate the performance of a number of PAQSMs.14 One phase
of that study consists of applying a complex photochemical dispersion model
developed by Systems Applications, Incorporated - the Urban Airshed Model.16
A number of days have been simulated to evaluate the Airshed Model's per-
formance in characterizing ozone formation under conditions encountered during
the data collection period. In addition, a series of detailed sensitivity
tests have been conducted for three of these days in order to:
(.1) identify those model inputs which most significantly affect
model predictions, and subsequently require most care in data collection
efforts;
(2) isolate possible sources of error in model inputs or formulation;
15
-------
(3) estimate the effects of changes in precursor emission levels on
predicted ambient ozone levels.
The results of this last set of sensitivity tests provide the basis for
evaluating the EKMA predictions of changes in ozone accompanying changes in
precursors, and are discussed in Section 4.
16
-------
2.0 LEVEL II ANALYSIS
As previously described, the trajectory model underlying EKMA has been
used with a comprehensive air quality management data base to make predictions
of peak ozone for comparison with observed levels. Several criteria were
employed to select ten trajectories on nine different days during 1976.
Because EKMA will necessarily be used for those days with the highest ozone
levels, primary interest focused on predicting the highest ozone levels
measured in the region. Further, enough days were selected to insure that the
model's performance was evaluated for a variety of atmospheric conditions. A
few days with lower ozone concentrations were also included to test for a
possible systematic bias in the model's predictions. These considerations led
to the selection of the ten test cases summarized in Table 2-1.
In the discussion that follows, the procedures for estimating the model
inputs are first described, followed by a comparison of the model predictions
of peak ozone with those observed. In addition, a detailed assessment of each
simulation is also made. Finally, Section 2.3 describes the results of a
series of tests designed to assess the sensitivity of the model predictions to
uncertainty in some of the model inputs.
2.1 Methodology for Developing Input Data
The first step in performing a simulation involved deriving an air parcel
trajectory corresponding to the time and location of the observed peak ozone
concentration. This trajectory represents the path an air parcel would have
traveled to reach the site of interest at the specified time (thereby repre-
senting the movement of the theoretical column described in Section 1.1.1).
Once the column movement had been defined, the remaining information necessary
to simulate a test case could then be developed: the initial concentrations,
emissions, boundary conditions (e.g., ozone aloft), and dilution data.
Methodologies used to develop each of these are described below.
2.1.1 Air Parcel Trajectory
Air parcel trajectories were calculated for each of the test cases
from the minute-by-minute measurements of wind speed and wind direction taken
17
-------
Table 2-1. Model Test Cases
Time of Peak 03, Peak 03>
Date Julian Day RAMS Site Local Daylight Time Concentration, ppm
10/1/76
7/13/76
6/8/76
6/7/76
6/8/76
8/25/76
10/2/76
9/17/76
7/19/76
8/8/76
275
195
160
159
160
238
276
261
201
221
102
114
115
122
103
115
115
118
122
125
1500-1600
1600-1700
1700-1800
1600-1700
1400-1500
1400-1500
1700-1800
1300-1400
1300-1400
1800-1900
.24
.22
.22
.20
.19
.19
.19
.15
.15
.12
-------
at each of the 25 RAMS. First, ten minute vector averages of wind speed and
direction were calculated for each of the 25 sites. The individual site
averages were then averaged to obtain an overall regional average wind speed
and direction for each ten minute period. The regional average wind speed and
direction were used to track a trajectory backwards from the site and time of
the observed peak ozone concentration until 0800 Central Daylight Time (CDT).
Figure 2-1 illustrates the back trajectory for the July 19th test case, with
the hourly segments of the trajectory shown along the path.
2.1.2 Initial Concentrations
The initial concentrations of nonmethane organic compounds (NMOC),
total oxides of nitrogen (NO ), and ozone represent the pollutant levels
A
initially within the theoretical model column at 0800 CDT. They were esti-
mated from the hourly-averaged concentrations at the three RAMS stations
nearest the trajectory starting point. The first step in the procedure was to
select the three RAMS sites closest to the trajectory starting point. At
these three sites, instantaneous concentrations corresponding to the simula-
tion starting time were computed by averaging the hourly pollutant levels for
the hour immediately preceding the starting time and the hour following the
starting time. For example, the 0800 CDT instantaneous concentration for a
pollutant at one site would be calculated by averaging the 0700-0800 and the
0800-0900 hourly levels. The initial column concentrations were then computed
V
as a weighted average of the three instantaneous levels, with the weighting
factors equal to the square of the reciprocal of the distance between each
RAMS site and the trajectory starting point (i.e., l/r?). In performing these
calculations, any concentrations below the minimum detectable limit of an
analyzer were set equal to the following lower limits: .005 ppm for NO ;
A
0.1 ppmC for NMOC; and .005 ppm for ozone.
2.1 .3 Emissions
The trajectory model also simulates the impact of precursor emissions
occurring after the simulation starting time. The RAPS emissions inventory
was used to estimate hourly emission rates for nonmethane organic compounds
(NMOC) and for NO . However, the emissions encountered by the column during
19
-------
-4330
•4240
700 710 720 730. 740 750
UTM EASTING
Figure 2-1„ Air Parcel Trajectory for July 19 Test Case,
770
780
20
-------
each hour are actually input to the model as fractions of initial concentration
of NMOC or NO . As described below, the fractions themselves were computed by
J\
comparing the emission densities encountered by the column of air during each
hour to the pollutant density initially in the column.
From the RAPS emission inventory, an average emission density for
each hour of the column trajectory path could be computed. This was done by
(1) establishing the 10 km x 10 km grid network shown in Figure 2-2; (2)
summing the hourly point and area source emissions occurring within each grid
square encountered by a trajectory segment; (3) dividing the total emissions
in each grid square by the area of that grid square (i.e., 100 km2); and (4)
weighting the resulting emission densities consistently with the proportion of
the trajectory segment in each grid square. For example, consider the tra-
jectory path shown in Figure 2-2. An emission density for the first hour
would be calculated from the total emissions occurring between 0800 and 0900 LOT
within grid squares (1) and (2). Since roughly 2/3 of the trajectory segment
between 0800 and 0900 occurs in grid square (2), the emission density in (2)
is weighted by a factor of "0.67," whereas the emission density in grid
square (1) is weighted by a factor of "0.33."
The actual fractions input to the OZIPP model were calculated for
both organic compounds and NO using the following expression:
/\
where
e. = fraction of initial concentration to be added during hour
1 i to represent emissions occurring during hour i
Qi = emission density* for hour i, moles/m2
Emissions derived from the RAPS inventory were expressed on a mass basis
(e.g., kilograms). To convert to a molar (or ppm) basis, the following
conversion factors were used: 46 gra/mole for NO and 14.5 for NMOC. For
NO , the inventory gives NO as equivalent N02, and for NMOC, one ppmC is
assumed equivalent to CH2.5-
21
-------
-4330
09 (J LOT 2/3 of hotfrly
•4240
700
710
730... 740 750
UTM EASTING
770
780
Figure 2-20 Air Parcel Trajectory for July 19 Test Case Demonstrating How
Fresh Precursor Emissions are Considered.,
22
-------
Ho - initial morning mixing height (described in Section 2.1.5), n
Co = initial pollutant concentration, ppm or ppmC
P = density of air, 41 moles/m3
Equation 2-1 represents the ratio of the emission density at a particular Lou
to a hypothetical column density based on the initial column conditions.
This procedure has the effect of assuming that the absolute mass of
emissions occurring during any one hour is instantaneously dispersed through-
out the theoretical model column. Note that, in equation 2-1, the area of the
column does not appear. However, this term is implicitly accounted for by the
emission density term, Qi. Hence, it is possible to vary the impact of fresh
emissions simply by changing the size of the grid squares used in the modeling
exercise. As subsequently described, model sensitivity tests to grid square
size were performed.
2.1.4 Boundary Conditions
Boundary conditions for the trajectory model include pollutant
concentrations found in the layer above the early-morning mixed layer. A
procedure used to estimate the level of ozone aloft has been described in
Reference 1 and was used in this study. Hourly ozone concentrations measured
between 1100 and 1300 CDT at upwind, rural-type monitors were averaged to
obtain an estimate of the levels aloft. Precursor pollutants aloft were
assumed to be negligible.
2.1.5 Dilution
Dilution in the trajectory model results from the change in mixing
height which occurs during the day. In the OZIPP model, the mixing height is
assumed to rise as a function of the time after sunrise in accordance with a
"characteristic curve" derived empirically from data taken during the RAPS
study.15 Because of the existence of detailed radiosonde and sodar data, the
OZIPP program was modified such that a day specific mixing height profile
could be used in place of the characteristic curve. (Use of the character-
istic curve is considered in the sensitivity analysis portion of the discussion.)
23
-------
Radiosonde measurements were taken at the four Upper Air Network
(UAN) stations shown in Figure 2-3. For every day except October 2, soundings
were taken at stations 41 and 42 at about 0500, 1100 and 1700 CDT. On some of
the days, additional soundings were taken at the other locations. The vertical
temperature profiles corresponding to each radiosonde were used to estimate
the height to which pollutants emitted near the ground would be likely to mix
radpidly (i.e., the mixed layer).* The mixing heights derived from radiosondes
taken at different sites but at approximately the same time were averaged to
obtain a regional average mixing height in early morning, mid-morning and
afternoon. These estimates were supplemented by measurements taken by SODAR
to better establish the mixing height as a function of time during the early
morning. In this fashion, mixing heights were estimated at hourly intervals
throughout the morning hours, and the profile for each day derived by fitting
a monotonically increasing smooth curve through the data points.
2.1.6 Chemical Mechanism
The chemical kinetics mechanism currently incorporated in OZIPP was
used in all ten tests. The N02/N0 fraction corresponding to observed initial
J\
concentrations of N02 and NO were used in all cases. The hydrocarbon reactivity
factors recommended in References 1 and 2 were also used.
2.2 Peak Ozone Predictions
The model inputs developed in accordance with the previously described
procedures are summarized for each test case in Appendix A. The peak ozone
concentrations predicted with these inputs are graphically, compared with the
observed ozone concentrations in Figure 2-4. The solid 45 degree line ema-
nating from the origin represents perfect agreement between observed peak
hourly ozone concentrations (abscissa) and predicted peaks (ordinate). The
In some cases, subjective interpretations of the radiosonde data were
invoked to arrive at the best estimate of the mixing height. For
October 2, only radiosondes conducted at the Salem, Illinois National
Weather Service Radiosonde Station were used since RAPS radiosonde data
were not available.
24
-------
90d45','/ \\ 90°'30'
90°'15
''
39°00'
38°30'
i
to
©
90olOO'
8SC|45'
0
0
i x/'
T°Y
_rv // m
f\ O
"OC
38^
3B°30-
38°15'
90°, 45'
90°,30'
90C,15'
O
O RAMS STATIONS
gg UAN SITES
A VORTAC LOCATIONS
90C|OC'
38°15'
Figure 2-3. Location of Upper Air Network Stations
25
-------
CD
a- i^h-
40 BO 120 160 200 240 280
OBSERVED OZONE. PP8
Figure 2-40 Level II Predictions Versus Observations of Peak Ozonec
26
-------
dashed lines on either side of this line represent agreement to within ±30%,
and the numbers on the graph indicate the Julian day corresponding to each of
the ten test cases. Of the ten simulations performed, only four are within
±30% of the observed peak. In practically every case, the observed peak ozone
concentration is underpredicted, sometimes by a substantial amount.
To examine potential causes for the tendency towards underprediction, a
detailed examination of each simulation was undertaken. The goal of the
analysis is to compare the temporal patterns of ozone, key precursors and
related variables (such as the NMOC/NO ) predicted by the model with measured
X
levels. In addition, the relatively inert pollutant carbon monoxide (CO) was
also modeled to provide supplemental information on the emissions/mixing
height relationships used in the analysis. The latter was possible only
because the RAMS monitors were sited such that they would not be unduly
influenced by any one source or group of sources. In most instances, such an
analysis would be inappropriate because CO monitors are normally located to
detect peak levels originating from a localized group of sources.
Because the model computes species concentrations in a moving reference
frame (i.e., in a moving column of air), and ambient measurements are recorded
in a fixed reference frame (i.e., at fixed monitoring sites), some means of
transformation is necessary for comparison. The procedure used is similar to
that for estimating initial concentrations. Pollutant levels are interpolated
on the hour along each trajectory from ambient measurements taken at the fixed
RAMS monitoring sites. First, trajectory nodes are established by locating
the trajectory position at the start of each hour. Next, the three RAMS sites
nearest each node are selected. For each chosen site, the concentrations at
the start of an hour are obtained by averaging the hourly average level for
the preceding hour with the hourly-average for the following hour (e.g., the
1100-1200 CDT average ozone level would be averaged with the 1200-1300 CDT
average ozone level to obtain an estimate of the ozone concentration occurring
precisely at 1200 CDT). Finally, the concentrations at the start of an hour
are combined into a weighted average using the square of the reciprical distance
between the node and the site. In a sense, the concentrations computed at
each node represent instantaneous levels, and are directly comparable to the
27
-------
model predicted instantaneous levels. It should be noted, however, that some
situations may arise in which the interpolated values might not be represen-
tative of the column concentrations. For example, some situations may occur in
which the interpolated value may be based on measurements taken 20-40 kilo-
meters from a trajectory node. In general, these occurrences are rare, and the
use of interpolated measurements provides one means of comparing concentration
calculated in a Lagrangian reference frame with those measured in a fixed
frame.
The model predicted temporal patterns of ozone, NO , NMOC, NO, N02,
A
NMOC/NO , N02/N0 , and CO versus the interpolated values for each simulation
A X
are contained in Appendix B. The interpolated averages are represented as the
circle, and the range of values used in calculating a particular interpolated
value is indicated by the vertical bars. The solid line is the model predic-
tion of a particular variable as a function of time. At a minimum, one might
expect the model prediction to fall within the range of values used in the
interpolation scheme. Each of the simulations are qualitatively described
below.
10/1/76 (Day 275). This day was one of general stagnation, with the
trajectory meandering within the urban area. The temporal pattern of CO
predicted by the model agreed fairly well with the interpolated pattern,
although the absolute levels were underpredicted in the morning. The results
for NMOC were similar, in that mid-morning levels were underpredicted. NO ,
NO, and N02 predictions agreed reasonably well with interpolated values. x
Predictions of ozone corresponded extremely well with interpolated levels up
to about 1400 CDT, after which ozone was underpredicted.
7/13/76 (J)ay 195). The trajectory for this day led from the rural
area southeast of the city, passed through the urban area about mid-day, and
ended slightly north of the city at 1600 CDT. Precursor concentrations, both
predicted and interpolated, were at or near minimum levels throughout the
morning. The interpolated levels of CO, NMOC and NO all increased in the
afternoon (corresponding to passage of the trajectory through the urban area),
although the absolute levels of each were still relatively low. Model predic-
tions of these three pollutants were all lower than interpolated levels through-
out the afternoon period. Ozone predictions agreed with observations until
early afternoon, after which the model underpredicted the interpolated ozone
levels.
6/8/76 (Day 160, Site 115). The peak ozone on this day was measured
at 1600-1700 CDT slightly northeast of the city. The trajectory begins north-
west of the city, passes through the city in later morning, and veers north-
28
-------
east to the site of maximum ozone. The interpolated NMOC, NO and CO patterns
are at the highest levels in early morning, decrease rapidly ?n mid-morning,
and level off to low levels during the afternoon. The same pattern is predicted
hy the model, but the interpolated concentrations of NMOC and CO from early-
to mid-morning are slightly underpredicted. Interpolated and model predicted
concentrations of NO , NO and N02 agree fairly well. The ozone predictions
are slightly higher £han the interpolated levels up to about noon, after '"hie!-
the model predictions are somewhat lower than the interpolated levels.
6/7/76 (Day 159). The trajectory on this day begins to the northwest
of St. Louis, moves slightly southward until mid-morning, and then veers
northward passing close to Alton, eventually leading to the site of peak ozone
about 40-50 km north of the central St. Louis urban core. Interpolated con-
centrations of CO, NMOC and NO are all on the low side, with the highest
levels occurring in the early horning. Model predictions of NMOC and NO are
somewhat lower than interpolated levels. However, the model predictions of CO
track the interpolated levels fairly well, although concentrations tend to be
slightly underpredicted in mid-afternoon. Ozone predictions agree with inter-
polations very well up to about noon, but are significantly lower in the
afternoon period.
6/8/76 (Day 160, Site 103). The trajectory for this simulation is
similar to that for the 6/8, Site 115 simulation discussed above. The model
predictions of CO agree very well with interpolated levels, as do the NMOC and
NO patterns. In the early morning, the predicted values of NO , NO and N02
are somewhat higher than the interpolated levels. Ozone predictions are
similar to the other days in that interpolated and model predicted levels
agree reasonably well, but the predictions fall off more rapidly in the
afternoon than the interpolated levels.
8/25/76 (Day 238). The measured peak ozone concentration on this
day occurred to the northwest of the city in early afternoon. The trajectory
started southeast of the city, and tracked from the southeast to the north-
west, passing quite well to the east of the urban area. The model predicted
CO levels were lower than the interpolated levels, even though the latter were
relatively low. The same pattern was found with NMOC, but NO predictions
agreed fairly well with interpolated levels. As might be expected, ozone was
significantly underpredicted, with the model prediction leveling out in the
afternoon.
10/2/76 (Day 276). October 2 was the day following the October 1
stagnation period. Because this was a Saturday, some of the meteorological
information normally collected were unavailable, and of course, and it is
likely that more uncertainty exists in the emissions estimates for this day
than for a weekday. Interpolated concentrations of CO, NMOC and NO remained
relatively high in the early morning hours and then decreased substantially in
late morning. The model predictions of these pollutants followed the same
general pattern, but the early morning peak NMOC was underpredicted. Ozone
was again substantially underpredicted.
29
-------
9/17/76 (Day 261). The trajectory on this day started in the
general urban area and tracked to the south. The peak occurred about 20 km
south of the city in early afternoon. The model predictions of CO, NMOC and
NO all agreed reasonably well with the interpolated levels, although CO and
NM$C were slightly overpredicted in late morning. Ozone predictions were
always greater than the interpolated levels, resulting in an overprediction of
the one-hour-average peak.
7/19/76 (Day 201). July 19 was marked by persistent and moderate
winds out of the south. The trajectory started well south of the city and
passed through the urban area in early morning. The measured peak occurred
about 50 km to the north in early afternoon. Although model predicted CO
concentrations agreed well with interpolated levels, both were quite low,
being just slightly above the minimum detectable limit of the analyzers. Late
morning to early afternoon concentrations of NMOC predicted by the model are
somewhat lower than the interpolated levels, although again both are rela-
tively low being below 0.5 ppmC. Except for a one-hour spike in the inter-
polated NO levels, model predictions and interpolated NO concentrations
exhibited good agreement. The afternoon ozone concentrations predicted by the
model were somewhat less than the interpolated levels.
8/8/76 (Day 221j. The trajectory for this day tracked from east to
west, passing through the city in mid-morning. The peak ozone, which the
lowest of the ten test cases, was measured about 40 km due west of the urban
area. Both model predicted and interpolated concentrations of NMOC, CO and
NO were all near the minimum detectable limits of the analyzers. Ozone
levels predicted by the model were somewhat lower than the interpolated
levels.
The findings discussed above suggest that one possible reason for the
underprediction of peak ozone is that its precursors are also underpredicted,
especially NMOC. Often, the inert pollutant CO was underpredicted as well.
The one case (Day 261) in which both precursor and CO predictions agreed
fairly well with interpolated values, ozfrne was slightly overpredicted.
Possible reasons for the underpredicted precursor levels could be erroneous
mixing height profiles, or trajectory paths, since these factors directly
affect model predicted concentrations of precursors. This is further addressed
in the sensitivity analysis of the following section.
2.3 Sensitivity Tests
In this section, attention is focused on sensitivity tests conducted to
examine the relative importance of some of the model input variables. It
should be noted, however, that because of the number of test cases and the
number of model input variables, a full, factorial design experiment was
30
-------
beyond the scope of this analysis. Rather, individual tests were chosen to
illustrate particular facets of model behavior.
The establishment of the sensitivity of model predictions to individual
model inputs is complicated by the interactions among variables. For example,
consider the model sensitivity to ozone aloft. If the mixing height does not
change during the day, then the levels of ozone aloft will not affect model
predictions at all. On the other hand, if the mixing height grows substan-
tially, model predictions will be significantly affected.* Thus, it is not
possible to generalize about the sensitivity of the model prediction to the
level of ozone aloft. Other such model interactions also occur, and some of
those are illustrated in the discussion that follows on the sensitivity of the
predictions to key model inputs.
Trajectory. Probably the most important part of the Level II
analysis is the derivation of the air parcel trajectory. The approach previ-
ously described was based on regional average wind vectors for ten minute
periods. Another possible approach would be to derive the trajectories using
a 1/r2 weighting factor for the stations within a preset scanning radius.
Such a technique was employed to generate alternative trajectories for each of
the ten test cases, and they are shown in Appendix C (along with the ones
previously described). In some cases, the two techniques agree fairly well.
In others, the disagreement is more pronounced. For the latter situation, it
is not possible to say which is correct.
To illustrate the potential effects of the differing trajectories on
the modeling results, consider the two test cases shown in Figure 2-5. For
the July 19 case, the two methods yield virtually identical trajectories, and
the model predictions of peak ozone are essentially the same (approximately
.11 ppm for both simulations). Conversely, the difference in trajectories for
October 1 leads to significant differences in model predictions. With the
Note that in the extreme, the model predictions of peak ozone approach
the levels of ozone aloft as the growth in mixing height becomes very
large.
31
-------
4330
Regional Average Winds
~H 1/R^ Weighted Average Winds
4240J—
730 740 750
UTM EASTING
Figure 2-5A0 Comparison of Trajectories Calculated by Alternative Methods
July 19 (Day 201)
32
-------
4330
4260
4250
4240-
I
H
i
1 F
1
Regional
/R2 ':/eig
4-i
//•£
/ /?
Average Winds
hted Average Mind
cliYv4£3j\\
COLUMBIA
WATER L
5 ~
^^""-™— •— •— .M^M,
-
i
-'
-*
•"
1
— ~— .
700
710 720
730 740 750
UTM EASTING
760 770
780
Figure 2-5B0 Comparison of Trajectories Calculated by Alternative Methods
October 1 (Day 275)
33
-------
trajectory based on regional average winds, the predicted peak ozone was
O.ldfc ppm, which was significantly lower than the observed .244 ppm. When the
trajectory based on the l/r2 weighting of the nearest stations is modeled, the
model prediction of peak ozone is substantially improved. The l/r2 trajectory
leads to higher initial concentrations, which in turn result in a higher
predicted ozone peak of .211 ppm. Thus, in one case, the differences in
trajectories are unimportant, but in another case they substantially affect
model performance.
The two techniques for estimating trajectories that have been
considered so far were based on surface level winds (i.e., measurements were
taken at a height of 10 meters). Information on the speed and direction of
winds aloft (within the mixed layer) are limited, but suggest that in some
cases substantial wind shear may exist, i.e, within the mixed layer, differing
wind directions are found aloft. Figure 2-6 illustrates the vertical profile
of wind direction measured by a midmorning radiosonde taken on June 7 (Day 159)
By combining the measurements of winds, aloft with the surface measurements, a
trajectory such as the one shown in Figure 2-7 could be hypothesized as a more
representative wind flow (the original trajectory is shown for reference). It
must be emphasized, however, that the measurements of winds aloft are of short
duration, taken at discrete points in time, with multi-hour intervals without
any measurement. The representativeness of the discrete measurements for
intermediate times is open to question. But, if the trajectory based on
combined surface/aloft winds is more representative of the wind flow on that
day, and were modeled as such, then the model's prediction of peak ozone would
increase because of the higher initial precursor concentrations near the new
starting point. Thus, model performance on this day would be improved.
The major point to be emphasized here is that using different
techniques to derive trajectories can lead to different estimates of air
parcel paths. In some cases, the differences in trajectories can lead to
substantially different modeling results. Determining which is the "correct"
trajectory is difficult because of no standard with which to compare the
different results, i.e., the trajectory most representative of the actual
parcel path is unknown.
34
-------
1052 LOT
QJ
>
O)
ISl
QJ
>
O
-Q
(O
1/1
S-
OJ
•(->
-------
^
'«
t
: k
i"\P
»
[SiVvt:
: f' . .
4 4- -
• f - t
< ::
•A t/» rt»T&,
!
4320
4310
4300
03
Surface Winds
Surface and Aloft Winds
t 428
700
710
720
730 740 750
UTM EASTING
Figure 2-7. Trajectory for June 7 Derived From Surface and Aloft Wind Data,
36
-------
Dilution. As. previously mentioned, uncertainties exist in estimating
the mixing height profiles.. In OZIPP, the profile is based on a number of
days, and the growth of the mixing layer is characterized by a "typical"
pattern. To compare the effects of using this characteristic profile rather
than the day-specific one, each test case was simulated using the characte^-
istic curve (shown in Figure 2-8), which is incorporated in OZIPP- Table 2-2
summarizes the differences in input and their effect on model predictions.
Note that the same afternoon maximum mixing height was used in both cases, but
for simulations with the characteristic curve, the mixing height was assumed
equal to 150 meters at sunrise in every case. This often leads to significant
differences in the mixing height at 0800 LOT, the start of the simulation.
Despite these differences, the effects on model predictions of peak ozone were
relatively small, the largest one being to increase the peak on one day by
only .022 ppm.
Post 0800 Emissions. As indicated in Section 2.1, the post-0800
emissions were derived on the basis of a 10 km x 10 km grid system. The
effect of choosing a different sized grid network on predictions of peak ozone
was examined by using 5x5 and 20x20 km grid networks, as well. Figure 2-9A
summarizes the results when the different grid networks were used with day-
specific mixing height profile. The results corresponding to the use of the
characteristic curve profile are shown in Figure 2-9B. In each figure, the
series of vertical bars depict the range within which model predictions varied
when the grid square size was varied. As can be seen, model predictions do
not vary substantially with grid square size, suggesting that the spatial
resolution of the emissions inventory is not an extremely critical factor.
37
-------
1.00
FRACTION OF DAYLIGHT HOURS
Figure 2-80 Graphical Depiction of Characteristic Curvec
38
-------
Table 2-2. Example Sensitivity to Mixing Height Profile
Mixing Heights, m
Test Case
10/1/76
07/13/76
06/8/76 (115)
06/7/76
06/8/76 (103)
08/25/76
10/02/76
09/17/76
07/19/76
08/08/76
Original Inputs
Sunrise
100
280
150
170
150
100
100
110
100
100
0800 LOT
100
280
150
170
150
100
100
no
100
100
Maximum
950
1670
2220
1920
2220
1800
1810
1670
1850
1440
Characteristic
Sunrise
150
150
150
150
150
150
150
150
150
150
Curve
0800 LOT Maximum
170
336
350
320
150
250
190
200
300
250
950
1670
2220
1920
2220
1800
1810
1670
1850
1440
Change in Ozone*
prediction, ppm
+ .015
-.003
+ .011
+ .002
+ .008
-.001
+ .009
+ .022
+ .003
+ .003
* A •+' means that the characteristic curve profile led to a higher predicted
peak ozone concentration than did the original inputs.
-------
'261
2tfl159I ,195
238^
^276
A) Day Specific Mixing
Height Profile
40 80 120 160 200 240 280
OBSERVED 02QNE, PP8
195
B) Characteristic Curve Mixing
Height Profile
I i i L i I I I I i I I i
40 80 120 160 200 240 2BO
085ERVEO OZONE. PP8
Figure 2-90 Sensitivity of Level II Model Predictions to Emission
Inventory Spatial Resolution,,
40
-------
3.0 LEVEL III ANALYSIS
The Level II analysis just described makes use of a comprehensive data
base. Such detailed information will not always be available, so alternative
procedures for developing model inputs have been formulated (Reference 2).
The purpose of this section is to assess the effect of using a less detail~J
data base on the predictions of peak ozone. The procedures for developing the
model inputs are first described, followed by a comparison of the model predic-
tions of peak ozone with those observed. Section 3.3 summarizes the results
of a sensitivity analysis designed to test the sensitivity of the model
predictions to uncertainty in the input data.
3.1 Model Input Data
The basic differences between the Level II and Level III data bases are
in the assumptions about available meteorological and emissions data. For the
Level III analysis, it is assumed that only enough wind data are available to
ascertain that the peak ozone level is, in fact, downwind of the urban area.
Second, only enough radiosonde data are available to estimate the 0800 LOT
mixing height and the maximum afternoon height. Finally, the spatial and
temporal resolution of the emissions inventory is limited to a seasonal,
countywide emissions inventory for VOC and NO . The procedures for estimating
X
the model inputs have been described in depth in Reference 2, and the
methodologies used here are briefly described below.
Trajectory. With a Level III data base, insufficient information
exists to establish an explicit air parcel trajectory. Thus, the column of
air is assumed to originate in the center city and begin moving at 0800 LOT
towards the site of peak ozone at a uniform speed.
Initial Concentrations. In the Level II analysis, initial
concentrations were estimated from the three monitors nearest the trajectory
starting point. Because the column of air originates in the urban area in the
Level III analysis, the initial concentrations were estimated from the early-
morning, urban average levels. The latter were calculated as the mean of the
individual 6-9 LOT averages at RAMS sites 101, 102, and 104-107. These six
sites were deemed representative of the urban area in general.
41
-------
Post-Q8QO Emissions.. The technique, for calculating post-0800
emissions is similar to that used in the Level 1,1 analysis, except that the
new assumed trajectory and countywide emissions, were used. The VOC and N0x
emissions for each hour were determined by the location of the trajectory
segment at each hour and the corresponding countywide emissions density. The
emissions fractions were computed with these data in the exact same manner as
described in Section 2.2.3.
Boundary Conditions. As in the Level II analysis, only ozone aloft
was considered (i.e., precursors aloft were neglected). The same estimates
made for the Level II tests were used in this analysis.
Dilution. For the Level III analysis, only the mixing height at
0800 LOT and the maximum afternoon mixing height are estimated, and the
characteristic curve defines the rate of the mixed layer's growth. Reference 2
recommends a minimum morning mixing height of 250 meters. The latter accounts
for the effects of mixing due to mechanical turbulence caused by increased
surface roughness in the urban area, and was used for all the Level III
0800 LOT mixing heights. The maximum afternoon mixing heights were the same
as those derived for the Level II analysis.
Chemical Mechanisms. As in the Level II analysis, all tests were
conducted with the chemical kinetics mechanism currently incorporated in
OZIPP. All default hydrocarbon reactivity factors recommended in Reference 2
were used. Default N02/N0 ratios of 0.25 and .10 were assumed for initial
X
concentrations and post-0800 emissions, respectively.
3.2 Predictions of Peak Ozone
The model inputs derived according to the procedures just described are
*,
summarized for each test case in Appendix D. The model results are graphically
depicted in Figure 3-1. Once again, model predictions (prdinate) are plotted
versus observations (abscissa), with the region between the dashed lines
corresponding to predictions within ±30% of observations. Of the ten test
cases, nine are within this range, but most are still slightly lower than
42
-------
40 80 120 160 200 240 280
OBSERVED OZONE, PPB
Figure 3-1. Model Predictions Versus Observations of Peak Ozone
(Level III).
43
-------
observations. Nevertheless., the Level III approach, shows, marked improvement
over the Level II analysis..
The rationale underlying the Level III approach rests on two bases.
First, it is clear that the Level II approach exhibits several problems. In
particular, a consistent wind field is difficult to define, even with a large
number of monitors and some measurements aloft. Any trajectory derived from
such data will likely have some uncertainty, which in certain cases can
critically affect the modeling results (see Section 2.4). Second, the data
typically available are of such a limited nature that they preclude a Level II
analysis in most instances (i.e., insufficient wind data exist to adequately
characterize a trajectory, and emissions data are not sufficiently resolved
spatially or temporally to use with a precise trajectory). In the Level III
analysis, therefore, the observed peak ozone is assumed to be largely produced
from precursors within the city in early morning and emitted subsequently.
A priori, one might expect that the Level III approach would lead to an
overestimation of observed ozone since:
(1) the early morning levels of precursors within the urban area
are often greater than those in surrounding areas;
(2) the higher levels of early morning precursors might lead to the
highest ozone levels in the afternoon;
(3) a sparse network of ozone monitqrs may,not detect the maximum
ozone level that actually occurs; and
(4) the model inputs are geared to generate the maximum ozone
level.
Because of the dense monitoring network in the RAPS, observations are not
as likely to underestimate the true maximum ozone level occurring within or
downwind of St. Louis. Thus, the close and only slightly biased predictions
summarized in Figure 3-1 are encouraging. However, because of the less
rigorous approach used in Level III compared to Level II, there is a greater
likelihood that agreement between observations and predictions is fortuitous.
44
-------
'}. 3 >ns Hi vj ty_Ana l^sjji
Just as for the Leyel II analysis, a full factorial design sensitivity
analysis was not possible. Rather, attention was focused on the variables
most likely to have the greatest impact on predictions of peak ozone, namely
dilution, post-0800 emissions, and initial concentrations. The basic approach
was to vary a key input variable by ±25%, and then establish the range over
which the predicted peak ozone levels might vary, given such an uncertainty in
the input. While the choice of a 25% variation is admittedly arbitrary,
careful compilation of a Level III data base should yield model inputs within
±25% of the true, or most representative, value.
Dilution. To examine the sensitivity of model predictions to
dilution, variations in the 0800 LOT mixing height and then the maximum
afternoon mixing height were considered. Simulations were repeated with the
morning mixing lowered by 25%, and then increased by 25%. Figure 3-2A sum-
marizes the results. The vertical bars indicate the range over which the
predictions varied in each test case, and the circle shows the original predic-
tion described in the previous section. For every test case, model predictions
increased when the morning mixing height was increased (i.e., dilution was
lowered). Furthermore, the model sensitivities differ according to the test
case. For example, predictions differ by 46 PPB for the Day 275 simulation
(10/1/76), but by only 5 PPB on Day 195 (7/13/76). Also, nonlinearities are
not readily apparent in the model response (i.e., the base case simulation
described in the previous section falls at or near the midpoint of the range
over which the morning mixing was varied). Finally, except for one test case,
model predictions remain within ±30% of the observed level over the entire
range of morning mixing heights evaluated.
Figure 3-2B shows the sensitivity of the model predictions to
changes in the maximum afternoon mixing height. It should be noted that the
±25% variation in afternoon mixing heights results in a much larger variation,
in absolute terms, than the variation in morning mixing height just discussed.
Similarly, the model response is also greater, but once again the sensitivity
varies according to the test case. In every test case, model predictions were
increased when the afternoon mixing height was lowered (i.e., dilution was
45
-------
m CM
a.
a.
a
• a
LU CM
z
o
tvi a
a in
o
UJ
cc
160/103
A) Morning Mixing Height
' I I I I I I i I I
I i i
40 80 120 160 200 240 280 320
OBSERVED OZONE, PPB
CD CM
a_
a.
o
• a
UJ CM
z
o
M a
a CD
a
UJ
>- a
e_> CM
B) Maximum Afternoon Mixing Height
'i i I t i I i I i 1 I I i I I I
40 80 120 160 200 240 280 320
OBSERVED OZONE . PPB
Figure 3-20 Sensitivity of Level III Model Predictions to Mixing Height.
46
-------
lowered). Note also, that model predictions are increased by a greater amount
when dilution is lowered than they are decreased by increasing the dilution
rate. However, in most instances, the predictions remain within ±30% of
observations over the entire range of variation.
Post-0800 Emissions. Figure 3-3A shows how model predictions vary
when both NO and VOC emissions are simultaneously changed by ±25%. Again,
X
day to day variations in sensitivity are found, with Days 195 and 221 exhi-
biting the greatest sensitivity. For most test cases, changing post-0800
emissions by ±25% does not result in model predictions outside the ±30% range
of agreement with observations. In every test case, increasing emissions
increased ozone. Finally, equally proportional changes in ozone are found
when emissions are decreased or increased.
Initial Concentrations. In Figure 3-3B, the sensitivity of model
predictions to variations in initial concentrations is shown. Again, both
NMOC and NO concentrations were simultaneously altered by ±25%. Increasing
/\
initial concentrations resulted in increased ozone in every case. The pat-
terns are similar to those found for the previously described sensitivity
tests: (1) variations in sensitivity are found from day to day; (2) the base
case simulation predictions fall at about the midpoint of the range of predic-
tions; and (3) most predictions remain within ±30% of the observed levels over
the range evaluated.
The sensitivity tests have focused on variations in a single variable
Simultaneous variations in two or more variables could lead to wider variations
in predictions. Furthermore, it is difficult to generalize about model sensi-
tivity to any particular variable. For example, the predictions for Day 195
were more sensitive to post-0800 emissions than to initial concentrations. On
the other hand, just the opposite was true for Day 275, suggesting post-0800
emissions are more important in the simulations for Day 195 than they are for
Day 275. Even though generalizations are difficult, model predictions are
apparently more sensitive to variations in afternoon mixing height than to
changes in the others.
47
-------
160/103
A) Post-0800 Emissions
' I I I I i i i i i
I i I i I
40 80 120 160 200 2<0 280 320
OBSERVED OZONE, PPB
o
to r-j
Q_
o.
o
' a
LD CM
I
a
M a
a (O
t- a
LJ CSI
B) Initial Concentrations
40 80 120 160 200 240 2BO 320
OBSERVED OZONE. PPB
Figure 3-3„ Sensitivity of Level III Model Predictions to Post-0800
Emissions and Initial Concentrations<>
48
-------
4.0 COMPARISONS OF EKMA WITH A PAQSM
The final approach for evaluating EKMA consists, of comparing the control
effectiveness predictions of EKMA with those of a more complex PAQSM. Because
PAQSM's are generally perceived as providing the best estimate of the rela-
tionship between ozone and its precursors, the results of the PAQSM can serve
as one basis for evaluating the performance of the simpler models. However,
two major shortcomings exist: (1) the PAQSM simulation of base conditions may
not always agree precisely with observations; and (2) no absolute guarantee
exists that the more complex PAQSM will accurately simulate changes from base
case conditions. A complete validation, as implied by this last limitation,
cannot be performed until changes from ba^e conditions are actaully imposed
and the observed effects are compared with the PAQSM predictions. Unfortu-
nately, the data base necessary for such complete model validation does not
exist. While the second shortcoming cannot be circumvented, the first can be
alleviated to some degree by evaluating only those PAQSM results corresponding
to cases in which model predictions and observations agree reasonably well in
the base case. Even though no absolute validation is possible, the possi-
bility of errors in the input data should be minimized because of the agreement
between PAQSM predictions and observations in the base case. As a result,
greater confidence is usually placed in the PAQSM predictions of control
program effectiveness for those days when the PAQSM simulates base case
observations accurately.
As described in Section 1.2, the Urban Airshed Model has been applied to
the St. Louis Metropolitan Area for a number of days to simulate ozone formation
under conditions encountered during the study period. In addition, a series
of detailed sensitivity tests were conducted for three of these days in which
VOC emissions were altered from base case conditions. Thus, tne principal
question addressed in this study is whether or not changes in peak ozone
levels accompanying cnanges in VuC emissions simulated by the Airshed Model
are similar to those predicted by EKMA.
49
-------
The preceding question would, at first, seem straightforward. However,
the manner in which these models are applied and the information provided by
each complicates the problem to some extent. For example, the Airshed Model
predicts pollutant concentrations, resolved both temporally and spatially
throughout the modeling region. The regional peak ozone concentration
predicted by the Airshed Model may not agree with the observed peak simply
because an ozone monitor was not positioned near the location of the predicted
peak. The implications of such a disagreement with regard to regulatory
applications have not been fully resolved. Also, the Airshed Model prediction
at the site of the observed peak may be higher or lower than the measured
level. Like the aforementioned problem, procedures for incorporating these
potential discrepancies in a regulatory application have not been determined
to date. On the other hand, EKMA is by definition an empirical approach,
depending on measured levels of ozone. The effects of proposed changes in
precursor emissions (i.e., NMOC and NO ) are evaluated relative to the initial
A
measured ozone level. Thus, the distinct possibility exists that EKMA and the
Airshed Model could have significantly different base case conditions in terms
of peak ozone.
To circumvent this problem, two separate procedures were followed in
comparing the two models. The first entails using Airshed Model results
despite any discrepancies between predicted and observed ozone peaks in the
base case, and then using EKMA exactly as it would be applied in practice.
The basis for comparison with these tests is the relative change in ozone
maxima from the base case level. For example, if an Airshed Model simulation
were conducted in which hydrocarbon emissions were reduced by 50%, the percent
change in the regional peak ozone from the base case is computed, and that
relative change is compared to the relative change in peak ozone predicted
using EKMA. To some degree, this puts the model predictions on a common
basis. The second approach consists of developing the inputs for EKMA on the
basis of the Airshed Model simulations (i.e., the Airshed Model peak ozone,
rather than the observed peak, would be used as input to the EKMA). For
example, if the Airshed Model predicted a peak ozone level of 0.30 ppm, this
value would be used to establish the starting point on the EKMA isopleth
50
-------
diagram, regardless of the observed peak ozone value. In effect, this
procedure involves -modeling a model," and also puts the model predictions on
a common basis.
Of the two procedures described in the preceding paragraph, the first is
the most rigorous test of EKMA's performance. For those situations in which
EKMA would be applied, the Airshed Model predictions would not normally be
available, and EKMA would have to be relied upon solely. Thus, these tests
provide some indication of how closely the models agree when they are applied
independently of one another. If the models disagree, the second set of tests
may help determine if the causes are due to the differences in model inputs.
For example, consider the situation in which the peak predicted by the airshed
Model is at a location with no nearby ozone monitor. The predicted sensi-
tivity of this peak to changes in precursor emissions may be different from
that predicted at the location of the measured peak. Thus, even though EKMA
may not agree with the Airshed Model in the rigorous test, agreement may exist
if the conditions corresponding to the Airshed Model predicted peak are
evaluated. Of course, a major limitation of any simpler model is that it does
not have the spatial and temporal predictive capabilities of the PAQSMs (i.e.,
it relies entirely on observations). This is a limitation that cannot be
overcome.
The discussion that follows is divided into three parts. In the tirst
portion, the urban Airshed Model is briefly described. This section is
followed by a summary of the Airshed Model simulations. Finally, Airshed
Model predictions are compared to those obtained with EKMA in the third section
In viewing these comparisons, it must be emphasized that this study is
not intended to determine the necessary control level for the St. Louis area,
but rather to compere two air quality models. It should be remembered that
the RAPS data base covers the 1975-1976 time period. In a regulatory analysis,
the use of a more current data base is desirable.
51
-------
4.1 Description of the Urban Airshed Model
The Airshed Model is a complex PAQSM of the Ejulerian .(or grid) type.16 A
network of grid cells is overlaid on the region to be modeled and the physical
and chemical phenomena leading to ozone formation are mathematically simulated.
These processes include emissions of primary pollutants (i.e., VOC and l\IOx)_
into each cell, the advection of pollutants from cell-to-cell, chemical
transformations of pollutants into intermediate and secondary species, trans-
port of pollutants into the modeling region from upwind areas, and entrainment
of pollutants from aloft due to growth in the mixed layer. The model computes
pollutant concentrations within each cell continuously, and thus attempts to
reproduce pollutant-concentration time profiles measured at each monitoring
site within the modeling region.
In the St. Louis application, the area modeled is 60 km wide in the East-
West direction and 80 km in the North-South. The horizontal dimensions of
each cell are 4 km x 4 km. Each 4x4 km area is divided into four individual
cells in the vertical dimension: the bottom two cells making up the mixed
layer, and the two upper cells corresponding to the region above tne mixed
layer. Pollutant concentrations initially within each cell are estimated from
available ambient measurements. Hourly emissions of primary pollutants
injected into each cell are input throughout the simulation period which
begins at 0500 Central Standard Time (CST) and ends at 1700 CST. The chemical
transformations are represented by a detailed chemical mechanism named Carbon
Bond II, which describes the NMOC-NO -03 photochemical interactions. The
A
concentrations of pollutants transported into the modeling region are deter-
mined from measurements taken at locations outside, and upwind, of the modeling
region. Finally, cell-to-cell advection is fixed by a wind field analysis
routine-which resolves measured wind data into u (east-west) and v (north-
south) components within each cell.
Figure 4-1 illustrates the horizontal grid structure relative to the
St. Louis area and the monitoring network. Note that the modeling region
encompasses 21 RAPS monitoring sites from which many of the modeling inputs
are derived. They also provide the air quality measurements necessary to
52
-------
n
km (Approx.)
Scale
Figure 4-1e St0 Louis Modeling Region,
53
-------
evaluate the model's performance. The outer four sites (i.e., Sites 122-125)
are used to determine the boundary concentrations which set pollutant transport
into the modeling region, for a raore detailed discussion of the Airshed Model
application, the reader is referred to Reference 14.
4.2 Summary of Airshed Model Simulations
This section describes the results of applying the Airshed Model to
simulate ozone formation on three days in the St. Louis region. Recall from
Chapter 1 that the Airshed Model should reproduce observed ozone concentrations
reasonably well to serve as the basis for evaluating the performance of
simpler models. While a detailed evaluation of the Airshed Model's performance
in the base case simulations is beyond the scope of this study, some rudi-
mentary analysis can provide insight into overall model performance. Following
this assessment, simulations in which precursors are changed from the base
case are described.
4.2.1 Base Case Simulations
The three selected days are June 7, 1976 (Julian Day 159), July 13,
1976 (Day 195) and October 1, 1976 (Day 275). Each of these days is charac-
terized by different meteorological regimes, thus serving to provide a broad
basis for model evaluation and for investigating the effectiveness of control
programs. On Day 159, the winds are light in early morning, developing into a
persistent southeasterly flow by early afternoon. The peak one-hour average
ozone concentration was 198 parts per billion (ppb), measured at Site 122,
located well to the north of the central urban area. Early morning concen-
trations of NMOC and N0x within the urban core were relatively high, and the
estimated ozone concentration above the early morning mixed layer was the
highest of the three days. On Day 195, winds were also out of the southeast but
somewhat stronger in early morning than on Day 159. The measured NMOC and NO
concentrations in the urban core on this day were much lower than those of the
other two days, while a moderate level of ozone aloft was estimated. Never-
theless, the measured one-hour average peak ozone was the second highest of
the three days - 223 ppb at Site 114, slightly north of the urban core.
Day 275 had the highest measured ozone concentration of the three days -
54
-------
244 ppb at Site 102 in the northern portion of the urban core. On this day
the winds were light and variable throughout the day, which is typical of a
stagnating air mass. Elevated early morning precursor levels were measured
within the urban core and the estimated level of ozone aloft was the lowest of
the three days. Some of the key measurements and Airshed Model inputs
characterizing the three days are summarized in Table 4-1.
Of principal interest (although not necessarily the best indicator
of model performance) is the Airshed Model's prediction of regional peak ozone
concentration. Figure 4-2 was constructed to compare the model's predictions
with observations in terms of magnitude, location, and timing of the ozone
peaks. In this figure, the large rectangle indicates the modeling region
while the smaller one shows the relative location of the urban core. The
location of the Airshed's prediction of peak ozone is denoted by a "O" while
the location of the observed peak is marked by an "0." The magnitude and time
of the corresponding peaks are shown to the right of each diagram, along with
the percent difference between observed and predicted peaks. The figure
clearly indicates significant differences between predicted and observed peaks
in terms of magnitude and location. However, these differences are not neces-
sarily indicative of model accuracy. For example, the Airshed regional peak
on Day 159 is at a location with no nearby monitor, while the actual observed
peak was outside the modeling region. On the other hand, the predicted
regional peak for Day 195 is somewhat lower than the observed peak, indicating
a model underprediction. For Day 275, the magnitude of the peaks is similar,
although the locations are somewhat different.
To provide a better indication of overall model performance, the Airshed
Model predictions of ozone at each of the 21 monitoring locations within the
modeling region were compared to the measured peaks.* The most rigorous test
Actually, the Airshed Model predicts cell-wide average concentrations,
not point concentrations necessary for direct comparison with monitored
pollutant concentrations. In the discussions that follow, the average
concentrations in the ground level cell corresponding to each monitor are
compared to the measured levels.
55
-------
Table 4-1. Summary of Miscellaneous Measurements and Airshed Model Inputs.
Julian Day
159 195 275
Measured 03 peak, ppb 198 223 244
Site of measured 03 peak 122 114 102
Time of observed peak, COT* 1600-1700 1600-1700 1500-1600
Measured 6-9 a.m. CDT urban 1.8 0.2 1.9
average** NMOC, ppmC
Measured 6-9 a.m. CDT urban .205 .048 .236
average** NO , ppm
A
Measured 6-9 a.m. CDT urban 7.6:1 7.0:1 8.1:1
average NMOC/NO ratio
/\
03 level aloft input to Airshed Model, ppb 114 78 63
0800-1500 CDT network average 1.0 2.3 0.6
wind speed, m/sec***
Maximum afternoon mixing height 1972 1853 529
inpjut to Airshed Model, meters
* CDT, Central Daylight Time
** Urban averages computed from the 6-7, 7-8 and 8-9 a.m. average at
Sites 101, 102, 104, 105, 106, and 107. The reported NMOC/NO ratios
are the averages of the six site-ratios, and thus do not necessarily
equal the ratio of the urban average concentrations.
*** From Reference 14
56
-------
DRY 159
Q312-I 1700-1800 CDT ]
0 198 - 11600-nOO CDT)
PERCENT DIFFERENCE = 58X
O 174 - I 1600-1700 CDT )
0223 - I 1600-nOO CDT)
PERCENT DIFFERENCE = -22X
0232 - I HOO-1500 CDT)
0244 (1500-1600 CDT)
PERCENT DIFFERENCE = -5X.
oRlRSHEO nODEL PREDICTED PERK OZONE. PPB
oOBSERVED PERK OZONE. PPB
Figure 4-2. Airshed Predictions Versus Observations of Regional Peak Ozone,
57
-------
involves comparing the observed ozone peak at each site with the ozone level
predicted by the Airshed Model precisely at the time of the observed peak.
The results of this test are. illustrated graphically in the left hand portion
of Figure 4-3. When the timing criterion is relaxed, the Airshed peaks
predicted at each monitoring site are compared to the observations, regardless
of the time of the peak. These results are shown in the right hand portion of
the same figure. In each of the individual graphs in Figure 4-3, the predicted
ozone (y-axis) is plotted versus observed ozone (x-axis). For reference, the
45 degree line (solid) indicates perfect agreement between predictions and
observations, while the other two dashed lines indicate over-and underpredictions
of ±30%. The results for Day 159 indicate that most predictions are within
±30% of the observed level, with or without the timing criterion. However, a
tendency does exist for the model to overpredict ozone levels, especially at
those sites with higher measured levels. On the other hand, the Airshed Model
tends to underpredict ozone levels on Day 195, even though most predictions
are within 30% of the observed levels. On Day 275, the most rigorous test
indicates a tendency to overpredict at the observed low ozone levels, and
underpredict at the higher levels. When the timing criterion is relaxed, the
agreement is improved somewhat, but underpredictions at locations with high
measured ozone still persist..
A detailed evaluation of model performance should, to the extent possible,
address how well the model reproduces concentrations of precursor patterns as
well as ozone patterns. While such an analysis is beyond the scope of this
study, particular attention was focused on the Airshed Model's predictions of
early-morning, urban core concentrations of NMOC and NO because of the impli-
X
cations for applying EKMA. Airshed Model predictions of these precursors are
compared to observed levels in Table 4-2. General patterns of precursor
levels were reproduced by the Airshed Model (e.g., high precursor levels on
Days 159 and 275, low levels on Day 195), although a tendency for underpre-
diction was found. (Note that in every case, NO is underpredicted, ranging
J\
from 14% to 42%, while on two of the days, NMOC levels were underestimated.)
The urban core, 6-9 a.m. NMOC/NO^ ratios predicted by the Airshed Model agree
reasonably well with the ratios derived from ambient measurements.
58
-------
[Rl WITH TIHt CRITttlON
IB) NI7HOUT Tin! CRITERION
0 40 BO 120 160 200 240 2BO
OBKRVEO OZONE, m
40 BO 120 1*0200 240 280
OWERVEO 010*1. m
0 40 80 120 160 200 240 280
OBSERVED DIME. PPB
40 80 120 180 200 240 280
OBSERVED OIONE. PPB
0 40 60 120 1BO 200 240 2BO
OBUIVEO OZONC. PfB
40 BO ISO 1BO 200 140 260
OC8EKVEO OIOMC. PfB
Figure 4-3. Airshed Predictions Versus Observations of Peak Ozone
59
-------
Table 4-?. 6-9 a.m. Urban Core Precursor Predictions Versus Observations.*
Day 159
Predicted Observed
6-9 a.m. CDT urban average NMOC, ppm 1.2 1.8
6-9 a.m. CDT urban average NO , ppm .152 .205
X
6-9 a.m. CDT urban average NMOC/NOv 8.1:1 7.6:1
/\
Day 195
6-9 a.m. CDT urban average NMOC, ppm 0.2 0.2
*
6-9 a.m. CDT urban average NOV, ppm .028 .048
X
6-9 a.m. CDT urban average NMOC/NOV 8.9:1 7.0:1
X
Day 275
6-9 a.m. CDT urban average NMOC, ppm 1.5 1.9
6-9 a.m. CDT urban average NO ppm .203 .236
/\
6-9 a.m. CDT urban average NMOC/NOv 7.7:1 8.1:1
X
Urban average NMOC and NO levels are computed from the 6-7, 7-8, and
8-9 a.m. averages at Sites 101, 102, 104, 105, 106, and 107 and the
cells corresponding to those sites. The NMOC/NO ratios are the average
of 6-9 a.m. ratios at the individual sites (cells), thus, they do not
necessarily equal the ratio of the urban average concentrations.
60
-------
The elementary analyses just described were not intended to assess
conclusively the accuracy of the model, nor do they provide a complete picture
of model performance. However; they do give s.ome insight into how well the
model is reproducing observed ozone levels. In most instances, predicted
ozone levels at the locations of monitoring sites are within ±30% of the
observations. Except for Day 159, the model tends to underpredict the higher
ozone levels. On Day 159, the model is biased towards overprediction at most
monitoring sites. Finally, a tendency exists for the model to underpredict
the early-morning precursor levels within the urban core.
4.2.2 Simulations of Changes in VOC Emissions
A number of sensitivity tests were conducted to assess how the
Airshed Model predictions of ozone respond to changes in VOC emissions. These
tests were designed primarily to indicate the potential effectiveness of
emission reduction programs. To establish this sensitivity over a wide range,
reductions in VOC of 17%, 42% and 75% (from base levels) were tested. These
reductions were applied uniformly to area and point source hydrocarbon emissions,
as well as to initial NMOC concentrations.* Boundary conditions were held
constant in all simulations.
Figure 4-4 summarizes the sensitivity results with respect to
changes in the predicted peak ozone levels. The large boxes represent the
modeling region, with the inner ones indicating the urban core. One diagram
is included for each simulation. On each, the location, time and magnitude of
the predicted peak ozone are shown. The first apparent result is that ozone
levels are lowered by reductions in VOC emissions, although the relative
changes are different. For example, a 75% reduction in VOC gives ozone reduc-
tions between 41 and 64%, depending on the particular day. Secondly, the
results indicate that, as VOC's are reduced, the peak ozone level is likely to
Clearly, a realistic control program would not produce uniform emission
reductions. However, the uniformity assumption makes the task manageable
and eliminates the need to make some arbitrary assumptions regarding
point versus area reductions.
61
-------
DRY
a
a
312 PPB
17-18 COT
»HC=OZ
a
D
289 PPB
n-ie COT
*MCr-17Z
DRY
a
a
174 PPB
16-17 COT
iHCzOZ
a
a
159 PPB
17-18 COT
»HCs-l7Z
DRY
a °
252 PfB
14-16 COT
*HC=OZ
D a
209 PPB
16-17 COT
»HC=-I7Z
159
a
D
216 PPB
18-18 COT
*HCc-42Z
195
a
D
199 FPB
17-18 COT
«HC--42I
275
a
D
149 PPB
le-n CDT
»MC=-42Z
3
D
139 PPB
15-16 COT
*HC:-7SZ
a
D
103 PPB
17-18 CDT
*HCr-7SZ
a
a
83 PPB
18-17 COT
»HC:-76I
0 - LOCRT10M OF PREDICTED PERK Q20NC
Figure 4-4. Summary of Airshed Model Simulations,
62
-------
of.tur -.lightly farther from the city (i.e., farther downwind). Although not
universally true, the peak often occurs later in the day under an emission
reduction scenario.
Figure 4-5 illustrates the simulation results in another way.
First, regional peak ozone concentrations are graphed against VOC reductions,
followed by a graph showing the relative change (i.e, percent change) in ozone
accompanying the reductions. In this last figure, a line of unit slope is
included as reference to indicate equally proportional changes between VOC and
ozone. The figure clearly indicates that the 03 response to changes in VOC is
different for each day. The Airshed Model predicts the largest incremental
changes in 03 on Day 275, while Day 195 exhibits the least response. In all
cases, the percent change in ozone is less than the change in VOC input to the
model. These findings are not entirely surprising since boundary conditions
remained unchanged in all simulations, and one might expect the importance of
boundary conditions to vary from day to day.
In examining Figure 4-5, one may be tempted to estimate the emission
reduction necessary to achieve the National Ambient Air Quality Standard
(NAAQS) for ozone. Several caveats are in order. Recall that in the base
case simulations, the Airshed Model predictions of regional peak ozone did not
agree precisely with measured peaks. In fact, a large difference existed on
Day 159. Whether or not the Airshed Model's predicted peak would be accepted
as the basis for a regulatory application has not been resolved. Secondly, as
described above, boundary conditions were not changed in any simulation. In a
regulatory application, some estimate of future boundary conditions might be
factored into the model applications (e.g., the level of ozone aloft might be
altered to reflect the implementation of upwind control programs in the
future).2 Finally, the modeling results provide information on how the 03
levels change on particular days. To establish the control needed to meet the
ozone standard, this information must be related to the statistical form of
standard. The latter allows, on average, one maximum daily value to be
greater than 0.12 ppra at each site during a year.17 Procedures for relating
the model output to the form of the standard are being explored, but have not
63
-------
10 20 30 40 60 60 70 80 90 100
HYDROCARBON REDUCTION. X
O DflY 159 A DRY 395
ORY 275
10 20 30 40 SO M 70 60 90 100
HTOROCMtMN REDUCTION. X
Figure 4-5, Airshed Model Sensitivity of Regional Peak Ozone to
Hydrocarbon Reductions.
64
-------
y»'t been prescribed. Thus, the information presented in Figure 4-5 could not
necessarily be used to estimate the reduction in VOC emissions needed to
achieve the ambient ozone standard.
4.3 Comparison of EKMA with Airshed
The previous section described the results of the Airshed Model
simulations in which VOCs were reduced uniformly from base conditions. These
simulations have been replicated, to the extent possible, with Level III EKMA.
Two different sets of comparisons, along the lines described in Section 4.2,
are made. The first involves comparing the models exactly as they would be
applied, in practice, without trying to compensate for known model differences.
For example, the base case ozone peaks used in EKMA are measured levels, while
base case peaks used with the Airshed Model are the regional peaks predicted by
Airshed. Other known differences in model inputs exist between Airshed and
EKMA also (e.g., NMOC/NO ratio, dilution rate, etc.). This is a very severe
/\
test and provides information on the model predictions obtained under independent
applications.
The second set of comparisons entails resolving known differences in model
inputs to eliminate these differences as potential sources of discrepancy
between models. In effect, EKMA was made to replicate, as closely as possible,
the conditions in the Airshed Model. Another way of looking at the second set
of comparisons is that EKMA is being used to model the more complex Airshed
Model. Such tests may provide some information regarding potential discrepancies
found in the first set of comparisons.
4.3.1 Independent Model Tests
Figure 4-6 shows the models' sensitivity to changes in hydrocarbons
for each day. The first apparent result is that relationships between model
responses differ according to the day being modeled. First, consider Day 159.
EKMA is less sensitive over the entire range of emission reductions than the
Airshed Model (i.e., EKMA predicts a smaller reduction in ozone for a given
hydrocarbon reduction). On Day 195, Airshed and EKMA agree reasonably well up
to a 40% reduction in emissions. At reductions greater than 40", EKMA becomes
65
-------
"9
E
53
DRY 159
<9 20 40 BO 60 100
MrOROCWWON REDUCTION. X
8
DflY 196
20 40 BO 60 100
HYDROCARBON REDUCTION. X
•* 20 40 «0 80 100
HTWOCRBBON RIOUCTION. X
fllHSHtO nOOEL
CNNR
Figure 4-60 Comparison of Airshed and EKMA (Independent Tests)
66
-------
substantially more sensitive than Airshed to incremental changes in VOCs. On
Day 275, the agreement between Airshed and EKMA is reasonably good up to a
reduction of about 40%; but again, EKMA exhibits a greater sensitivity to HC
changes at higher reductions. These findings demonstrate that EKMA and the
Airshed Model do not agree under all conditions.
While the overall model responses are an important consideration, a
major function of any of these models is to determine how much control is
needed to achieve the NAAQS for ozone. As discussed earlier, this question
cannot be directly answered here. However, it is possible to estimate for
each day the VOC emission reduction needed to lower a daily peak to 120 ppb.
Presumably, data such as these would be used in determining the precise
degree of control needed to achieve the ozone NAAQS.*
Table 4-3 summarizes the results, assuming an independent application
of models. Two sets of Airshed Model predictions are presented in the table.
In the first set, the Airshed Model results are used as is, with the base case
peak corresponding to the model predicted peak (not the observed). In this
case, EKMA predictions, when compared to those of the Airshed Model, are
lower, about the same, and lower on Days 159, 195 and 275, respectively. For
those days in which EKMA yielded lower controls than Airshed, the differences
are on the order of 10%.** The second set of predictions corresponds to a
simplistic method for calibrating the Airshed Model to base case measurements.
* It should be emphasized that in all simulations in which VOC emissions
were reduced, boundary conditions were kept at the base levels. In a
true regulatory application, some consideration may be given to altering
boundary conditions to reflect the impact of upwind control programs.
This is most important with respect to ozone aloft. Furthermore, the
form of the ozone NAAQS complicates the process of using the Airshed
Model results to estimate the level of control needed to achieve the
NAAQS. The NAAQS allows for one maximum daily one-hour level to be above
0.12 ppm per year, per site. The three days that have been modeled
correspond to peaks observed at three different sites. Thus, the number
of site/days that have been modeled is insufficient to estimate the
lowest level of control that is just needed to achieve the NAAQS.
** On Day 159, the control needed to reduce the Airshed predicted peak of
120 ppb is outside the range of tests, and thus the difference could be
somewhat greater.
67
-------
Table 4-3» Emission Reductions Needed to Reduce Peak 03 to 120 ppb.
Day
159
195
275
Base case peak 03, ppb
03 reduction necessary to
lower peak to 120 ppb
HC reduction necessary to
lower peak to 120 ppb
Base case peak 03, ppb
03 reduction necessary to
lower peak to 120 ppb
HC reduction necessary to
lower peak to 120 ppb
Base case peak 03, ppb
03 reduction necessary to
lower peak to 120 ppb
HC reduction necessary to
lower peak to 120 ppb
Airshed*
312
62% (39%)
| >75% (52%) |
I I
174
31% (46%)
| 60% (>75%) |
I I
232
48% (51%)
[ 57% (59%) j
1 1
EKMA
198
39%
66%
223
46%
62% ;
1
244
51%
47% ,
The HC reductions in parenthesis refer to the control needed to lower
ozone by the same percent as the other two models.
68
-------
Here, the Airshed Model response curve (Figure 4-6} is used to estimate the
hydrocarbon reduction that will give the percent change in ozone that is
necessary to lower the measured peak to 120 ppb. For this simplistic cali-
bration of the Airshed Model to base conditions, the comparisons lead to
somewhat different results. In this case, EKMA estimates of control are
higher, lower, and lower than those of the Airshed Model. Once again, the
differences between Airshed predicted controls and those of EKMA are on the
order of 10%, or less.
The results just described are similar to the findings regarding the
comparisons of overall model response, i.e., the comparisons reveal day-to-day
variations in results. However, differences between EKMA and the Airshed
Model were usually on the order of ±10%. EKMA gave moderately higher esti-
mates of emission reductions in only one case (14% higher). Thus, based on
the Airshed Model results, EKMA would not appear to consistently overestimate
the VOC emission reduction necessary to lower peak ozone to the level of the
standard.
4.3.2 Common Basis Tests
As described above, the comparisons of Section 4.3.1 were made on
the basis of independent applications of two models. In this section, com-
parisons of the models are made using the Airshed Model simulations as the
basis, i.e., the EKMA simulation is made to replicate the Airshed base case
simulation as closely as possible. In these tests, the 6-9 a.m. NMOC/NO
A
ratio predicted by Airshed for the "urban core" (see Table 2-1) and the
Airshed Model predicted peak ozone are both used to establish the EKMA
starting point. Additionally, the OZIPP program was modified to reflect more
precisely the growth in mixing height as treated in the Airshed Model, and
NMOC and NO concentrations aloft corresponding to Airshed Model inputs were
X
used in the EKMA modeling.* Presumably, these modifications put the models on
The mixing height growth in Airshed is based on a piecewise linear fit to
measured mixing heights. NMOC and NO levels aloft were on the order of
.05 ppmC and .003 ppm, respectively. x
69
-------
a more common basis and allow for a more direct comparison of model results by
removing possible discrepancies in model inputs.
Figure 4-7 depicts the relative changes in peak ozone as a function
of changes in hydrocarbons. The results are somewhat similar to those in the
preceding section. However, the overall agreement between Airshed and EKMA is
slightly improved. In most cases, EKMA appears to be more sensitive to
changes in VOC emissions than the Airshed Model over the entire range of
evaluation. On Day 159, however, EKMA's sensitivity is less than Airshed's,
up to a control level of about 75%.
Table 4-4 summarizes the individual emission reductions predicted by
each model to be necessary to lower the Airshed Model's base case peak ozone
to 120 ppb. EKMA gives lower control estimates than Airshed in all three
cases. The largest difference is 9%, although the possibility exists for a
larger difference on Day 159. Once again, the comparisons suggest that EKMA
does not consistently overestimate the controls needed to reduce peak ozone
levels to 120 ppb.
These findings suggest that the model differences found in the
previous sections are not entirely due to discrepancies in model input. Even
when the models are put on a common basis, differences do occur. EKMA cannot
exactly reproduce the Airshed Model predictions over all conditions considered,
but it does not consistently overestimate the VOC reductions necessary to
reduce peak ozone levels to 120 ppb.
70
-------
"s
I
52
DRY 159
JO 40 60 BO 100
hrOROCARBON REDUCTION. X.
20 40 60 80 100
HYDAOCARBOW REDUCTION. X
•9 20 40 80 60 LOO
MTOIOCHH80N REDUCTION. X
AIRSHED nOOEL
Figure 4-7c Conparlson of Airshed and EKI1A (Connon Basis Tests)
71
-------
Table 4-4. Emission Reductions Needed to Lower Peak 03 to 120 ppb.
Day
159 Base case peak 03, ppb
03 reduction necessary
to lower peak to 120 ppb
•
HC reduction necessary to
lower peak to 120 ppb
195 Base case peak 03, ppb
03 reduction necessary to
lower peak to 120 ppb
HC reduction necessary to
lower peak to 120 ppb
275 Base case peak 03, ppb
Reduction in 03 necessary to
lower peak to 120 ppb
HC reduction necessary to
lower peak to 120 ppb
Airshed
312
62%
>75%
174
31%
60%
232
48%
57%
EKMA
312
62%
72%
174
31%
51%
232
48%
50%
72
-------
5.0 CONCLUSIONS AND RECOMMENDATIONS
Three methods for evaluating EKMA have been considered. The first two
consist of comparing absolute model predictions of peak ozone with observations.
The major differences between these two methods, are in the complexity of the
data base, and the assumptions used in deriving the model inputs. Again,
these two methods do not address the key questions of how well EKMA predicts
changes in peak ozone due to changes in precursor emissions. However, the
first two approaches do provide some indication of how well the model simu-
lates the ozone formation processes. The third approach more directly
addresses the major question of interest. Here EKMA predictions of changes in
ozone are compared to those of a state-of-the-art photochemical air quality
simulation model. The major findings of each phase of the study are summarized
below, followed by recommendations for additional study.
The predictions of peak ozone using OZIPP/EKMA in a Level II mode are not
good. Substantial underpredictions of ozone resulted in almost every case. A
detailed evaluation of each simulation revealed that ambient NMOC levels may
also have been underpredicted. The same pattern was also found with the
relatively inert pollutant CO, suggesting a problem with the emissions/ambient
precursor relationship. However, the findings of the sensitivity analysis
suggest that the lack of agreement between predictions and observations could
possibly be explained for some days. Of critical importance is the definition
of the air parcel trajectory. Using different mathematical techniques to
calculate a trajectory will often lead to different trajectories, and some-
times these differences can result in significantly different model predictions,
The foregoing findings indicate that a Level II approach used in a
regulatory framework may have some serious problems. Model predictions of
peak ozone appear to be critically sensitive to particular model inputs that
are difficult to estimate, even with a comprehensive data base. How the
predictions of changes in ozone resulting from changes in precursors would be
affected by poor absolute predictions of ozone is not fully understood, but
one study has suggested that estimates of controls needed to meet the ozone
NAAQS would be underpredicted for a case in which base case peak ozone
73
-------
concentrations arc underpredicted.<:) While some tests could have been
umcluU.ed l.o address this problem, it is clear that using different techniques
to derive the model inputs will lead to different results. It is beyond the
scope of this s.tudy to examine several various alternatives that could be
employed. In summary, then, the model does r.ot replicate peak ozone levels in
a reasonable manner, but it is not clear whether this is a failure of the
model itself or comes about from problems in developing model inputs. In
either case, the poor model performance does not lead to confidence in using
the Level II approach to make estimates of changes in ozone that would result
from changes in precursors.
Somewhat surprisingly, when the data base complexity is reduced to that
commensurate with a Level III analysis, model performance in predicting peak
ozone is markedly improved. Clearly, this improvement is the result of starting
the trajectory in an area with the highest precursor initial loading. Because
of the assumed trajectory, the agreement between predictions and observations
may be somewhat fortuitous. Nevertheless, one would expect that the high
initial loading would eventually lead to an ozone peak, although the timing
and location of the peak may not correspond precisely with an ozone monitoring
site. Furthermore, the general pattern of the Level II trajectories corre-
spond to the assumed Level III trajectory in many instances, although dif-
ferences in timing and location occur (i.e., most Level II trajectories pass
near the urban area by the late morning). Thus, a distinct possibility exists
that the "smoothing" of the data attendant with a Level III analyses may
remove some data anomalies inherent in the development of the Level II inputs.
While the Level III approach is less intuitively appealing from a technical
point of view than the Level II analysis, peak ozone predictions do agree
reasonably well with observations. Furthermore, the model predictions were
found to be relatively insensitive to model inputs as compared to the Level II
approach. Because of these two factors, greater confidence would normally be
placed in the Level III predictions of changes in ozone than in those found
from a Level II approach.
74
-------
The most direct means investigated for evaluating the accuracy of EKMA in
predicting changes in ozone accompanying changes in precursors was to compare
EKMA's predictions with those of a complex PAQSM. Even though no absolute
guarantee exists that the PAQSM predictions are accurate, these complex models
represent the state-of-the-art in modeling ozone formation, and as such,
should provide the best available estimate. Comparisons of EKMA wth the PAQSM
were complicated by differences in their mode of application. Considering all
of the tests, EKMA did not reproduce the PAQSM results precisely. However,
when estimates of the degree of control needed to reduce peak ozone to the
level of the standard were made for each of three days with the two models,
differences in the estimated VOC emission reductions were less than 10%. EKMA
was not found to over or underestimate the required reduction, as compared to
the PAQSM reduction. One apparent finding of the study was that the VOC
emission reductions needed to reduce peak ozone to the level of the standard
is dependent on the day being modeled. For example, the PAQSM predicted that
the level of control needed to reduce peak ozone to 120 ppb was greater for
Day 195 than for Day 275, even though the latter had the higher ozone peak
(both predicted by the PAQSM and observed). The extent to which this is a
result of meteorological effects or background conditions is not known at this
time, but the fact that EKMA replicated these results is encouraging.
The above findings lead to the following recommendations for future
study.
0 The approaches for estimating Level II model inputs need more
detailed study. In particular, techniques for estimating air
parcel trajectories need further investigation. Techniques for
considering surface level winds in conjunction with the winds
measured higher in the mixed layer should be evaluated.
0 The trajectory model underlying EKMA should be compared to a
more complex photochemical trajectory model. Differences in the
treatment of specific physical phenomena should be identified,
and their effects should be isolated to identify possible limita-
tions and potential improvements in the trajectory model underlying
EKMA.
75
-------
0 Further studies, are needed to identify potential differences -that
may result from the use of different chemical mechanisms. In
particular, EKMA should be compared to the PAQSM using the same
chemistry.
0 Comparisons should be made in which boundary conditions to the
PAQSM and EKMA are altered in accordance with current regulatory
guidelines.2
76
-------
6.0 REFERENCES
1. Uses, Limitations and Technical Basis of Procedures for Quantifying
Relationships Between Photochemical Oxidants and Precursors^
EPA-450/2-77-021a, U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, November 1977.
2. G. L. Gipson, W. P. Freas, R. F, Kelly, and E. L. Meyer, Guideline for
Use of City-specific EKMA in Preparing Ozone SIP's, EPA-450/4-80-027,
U.S. Environmental Protection Agency, Research Triangle Park, North
Carolina, March 1981.
3. J. Trijonis, Verification of the Isopleth Method for Relating Photochemical
Oxidant to Precursors, EPA-600/3-78-019, U.S. Environmental Protection
Agency, Research Triangle Park, North Carolina, February 1978.
4. J. Trijonis and Marilyn Marians, "Historical Emission and Ozone Trends in
the Houston Area," Proceedings, Ozone/Qxidants Interaction with the
Total Environment II, Specialty Conference, Air Pollution Control Association,
October 1976.
5. "Data Collection for 19S2 Ozone Implementation Plan Submittals," Federal
Register, November 14, 1979, 44_ (221). 65669-65670.
6. F. A. Schiermeir, Air Monitoring Milestones: RAPS' Field Measurements,
Environmental Science and Technology, 12, 1978.
7. Ozone Isopleth Plotting Package (OZIPP), EPA-600/8-78-014b, U.S. Environmental
Protection Agency, Research Triangle Park, North Carolina, July 1978.
8. G. Z. Whitten and H. Hugo, User's Manual for Kinetics Model and Ozone
Isopleth Plotting Package. EPA-600/8-78-014a, U.S. Environmental Protection
Agency, Research Triangle Park, North Carolina, July 1978.
9. H. E. Jeffries, et al., Effects of Chemistry and Meteorology on Ozone
Control Calculations Using Simple Trajectory Models and the EKMA Procedure,
EPA-450/4-81-034, U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, November 1981 (in press).
10. M. W. Chan, D. U. Allard and I. Tombach, Ozone and Precursor Transport
Into an Urban Area - Evaluation of Approaches, EPA-450/4-79-039, U.S.
Environmental Protection Agency, Research Triangle Park, North Carolina,
December 1979.
11. Procedures for Quantifying Relationships Between Photochemical Oxidants
and Precursors: Supporting Documentation, EPA-450/2-77-021b, U.S.
Environmental Protection Agency, Research Triangle Park, North Carolina,
February 1978.
77
-------
12. F. L. Ludwig, Assessment of Vertical Distributions of Photochemical
Pollutants and Meteorological Variables in the Vicinity of Urban Areas,
EPA-450/4-79-017, U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, August 1979.
13. R. Haws and R. Paddock, The Regional Air Pollution Study (RAPS) Grid
System. EPA-450/3-76-021, U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina, December 1975.
14. K. L. Schere, "Evaluation of the Urban Airshed Model Using Data of the
Regional Air Pollution Study," presented at Symposium on RAPS Results,
St. Louis, Missouri, October 1980.
15. K. L. Schere and K. L. Demerjian, "A Photochemical Box Model for Urban
Air Quality Simulation," Proceedings, 4th Joint Conference on Sensing of
Environmental Pollutants, American Chemical Society, November 1977.
16. S. D. Reynolds and L. E. Reid, An Introduction to the SAI Airshed Model
and Its Usage, SAI Publication EF 78-53R, Systems Applications, Incorporated,
101 Lucas Valley Road, San Rafael, California, May 1978.
17. Code of Federal Regulations, "National Primary and Secondary Ambient Air
Quality Standards," Title 40, Part 50.9.
78
-------
APPENDIX A
The following figures and tables summarize the Level II inputs for
each test case that was modeled. Three items are included for each day:
(1) a data sheet summarizing the model inputs;
(2) a map showing the derived air parcel trajectory;
(3) a figure showing the day-specific mixing height profile
and the associated dilution rate.
A-l
-------
MODLL INPUT DATA
DATE: /Ot
JULIAN DAY: 2.73"
SITE: X02.
Simulation Start Time: OfOO LOT
Simulation End Time: l8QO
Initial Concentrations:
NMHC 0. 73
03 Aloft:
ppm
LOT
; N0y .12.7
A
ppm; 03
ppm; N02/NOX
Post 8 a.m. Emissions:
Hour
r
NMOC Emission
Densi ty ,
kg-moles/km2 hr
•£727
3
_!*_
5-
4
/7/J
1,7 tl
I,VL
1.793
7 i»is
NMOC Emission
Fraction
NO Emission
Density,
kg-moles/km2 hr
NO Emission
Fraction
-------
10/1/76
Day 275
Peak Ozone = .24 ppm
Time of Peak = 1500-1600 LOT
4330
4250
424C
CRYSTAL
1
i
i
\
\
\
\
*>
21
700 710 720 730 740 750 /faU
UTM EASTING
A-3
-------
October 1, 1976
Day 275
M
I
X
I
H
G
H
E
I
G
H
T
M
E
T
E
R
S
MIXING HEIGHT DATA
-2
X10
1.00
8
17
D
I
L
Li
T
I
0
H
R
A
T
E
M
I
N
TIME. LOT
A-4
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SITE:
Simulation Start Time: QfOO LPT
Simulation End Time: /fOO LPT
Initial Concentrations:
NMHC ^. 10 ppmC; NO
Aloft: 0. Of
ppm
ppm; 03 0.01 S ppm; N02/NOX
Post 8 a.m. Emissions:
Hour
\
2
3
fr
NMOC Emission
Density,
kg-moles/km2 h_r_
O.tfl
o.tfl
2./07
NMOC Emission
Fraction
0.071
Q.orf
^536
A/12.
270/
NO Emission
Density,
kg-moles/km2 hr
g.Q/7
O.OtS
o.nt
0.3X,
NO Emission
Fraction
0.601
7.1*7
A-5
-------
Day 195
Time'o/Peak =* 1600-1700 LOT
4330
4250
4240
I
CRYSTAL
CITYV
/
\
m
\
s
i
WATERL
»
?o
21
0800 LOT
700 710 720 730 740 750 760 7/0 /«u
DIM EASTING
A-6
-------
July 13,1976
Day T95
M
I
y
I
H
G
H
E
I
G
H
T
M
E
T
E
R
MIXING HEIGHT DATA
0.063
-0.806
-0.904
-0 892
D
I
L
U
T
I
0
N
R
H
T
E
M
I
N
17
TIME, LOT
A-7
-------
MODEL INPUT DATA
DATE:
JULIAN DAY-:
SITE:
Simulation Start Time: £>gQO LPT
Simulation End Time: /gOQ LPT
Initial Concentrations:
NMHC
PpmC; NOV O.Ofte ppm; 0, 0.QZ3 ppm; N09/NOV 0,6?
03 Aloft: llt
O.lll
NO Emission
Density,
kg-moles/km2 hr
0.H1
O.IYI
0.190
1.55V
o.m
NO Emission
Friction
O.US
0.3V3
A-8
-------
Day 160
Time of Peak = 1700-1300 LOT
4330
4320
700
710
720
730 740 750
UTM EASTING
770
780
A-9
-------
June 8, 1976
Day 160
MIXING HEIGHT DATA
M
I
X
I
N
G
H
E
I
G
H
T
E
T
E
R
8.926
8
TIME, LD1
r.r
A-10
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SITE:
Simulation Start Time: ^frOQ LPT
Simulation End Time: tfOO LPT
Initial Concentrations:
NMHC 0.39 ppmC; NO £
°3 Aloft: Q.fl ppm
ppm; 00 0.O3J ppm; N09/N0
j L
Post 8 a.m. Emissions:
Hour
J__
NMOC Emission
Pensity,
kg-moles/km2 hr
QJTI
0.073
NMOC Emission
Fraction
O.OS1
o,oK
0.071
&.
0.017
NO Emission
.Density,
kg-moles/km2 hr
t.ori
0.013
O.Oil
Q.OQ*
0,00?
NO Emission
Fraction
Q.oK
0.057-
0.93*
0.03L
A-11
-------
6/7/76
Day 159
Peak Ozone = .20 ppm
Time of Peak * 1600-1700 LOT
4330
4250
4240
CRYSTAL
CITYsj
1
^
\
s
1
\
WATERL
®
30
24
700 710 720 730 740 750 760 770 /BU
UTM EASTING
A-12
-------
June 7, 1976
Day 159
2800
MIXING HEIGHT DATA
0.03
-0 82
-0.01
8.06
D
I
L
U
T
I
0
H
R
A
T
E
M
I
N
8
16
1?
TIME, LOT
A-?:
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SITE: /03
Simulation Start Time: 0&QO LPT
Simulation End Time: /JOO LPT
Initial Concentrations:
NMHC
Aloft: ^. /O
ppmC; N0w O.Otj ppm; 0
~" - -
ppm
ppm; N09/NO
Post 8 a.m. Emissions:
Hour
NMOC Emission
Density,
kg-moles/km2 hr
Uil
1,757
2.7VJ
NMOC Emission
Fraction
f.WO
2.36?
NO Emission
Density,
kg-moles/km2 hr
0,115
a
A 7^7
NO Emission
Fraction
o.ts*
A-14
-------
6/8/76
Day 160
Peak Ozone
Time of Peak
.19 ppm
1400-1500 LOT
4330
4320
4310
4300
C3
429Q
a:
o
4280f
4270
4260
4250
4240
I
CRVSTAL!
ciTY^p
700 710 720
£
(
V
.'i A , 1 •
tf.
JIJ
2i
\
730 740 750 760 770 780
UTM EASTING
A-15
-------
June 8, 1976
Day 160
M
I
,-'-,
1
H
G
H
E
I
G
H
T
H
E
T
E
R
NIXING HEIGHT DATA
8 028
TIME, LDT
A-16
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SITE:
Simulation Start Time: 0%OO LPT
Simulation End Time: XfOQ
LPT
Initial Concentrations:
NMHC 0.1? ppmC; N0v 0.02S ppm; 0~ O. 0/3 ppm; N09/N0 0.37
X ' ,3 L. X
03 Aloft:
ppm
Post 8 a.m. Emissions:
Hour
J
Z.
3
NMOC Emission
Pensity,
kg-moles/km2 hr_
0.05$
0.17S
NMOC Emission
Fraction
0.07S
0.075
0. 37*
NO Emission
Defisity,
kg-moles/km2 hr
0.0/1
O.QJ7
j.oyj
D.O
-------
8/25/76
Day 238
Peak Ozone = .19 ppm
Time of Peak = 1400-1500 LOT
4330
4240
700
720
730 740 750
UTM EASTING
760
770
780
A-18
-------
August 25, 1976
Day 238
MIXING HEIGHT
M
I
X
I
H
a
H
E
I
G
H
T
M
E
T
E
R
c
0.0;
h-0.01
0.00
15 16
D
I
L
U
T
I
0
N
R
H
T
E
M
I
N
'IME, LOT
A-19
-------
MOULL INPUT DATA
DATE:
JULIAN DAY:
SITE: /AT
Simulation Start Time: 0fOO LPT
Simulation End Time: 1$ OQ LPT
Initial Concentrations:
NMHC # f V ppmC; NO £
03 Aloft:
ppm
ppm; QS 0.005" ppm; N02/NOX #23
Post 8 a.m. Emissions:
Hour
6_
7
NMOC Emission
Density,
kg-moles/km2 hr
o.m
Emission
o.w
O.ltf
NMOC Emission
Fraction
O.IQ7
OJJ1_
O.US
0.0V3
&W3
O.W3
Defisity,
kg-moles/km2 hr
0,131
0.IS*
0,0*1
0.021
&w
0,K!
NO Emission
Fraction
o.i si.
0./11
ft/25"
ft&7J
O.&lf
ffi.u?
A-20
-------
10/2/76
Day 276
Peak Ozone = .19 ppm
Time of Peak = 1700-1300 LOT
4330
4320
4310
4300
C3
O
•z.
700
710 720
r30 740 750
UTM EASTING
A-21
-------
October 2, 1976
Day 276
MIXING HEIGHT DATA
M
I
X
I
N
G
H
E
I
G
H
T
M
E
T
E
R
S
2838
1800-
1600-
1400-
1200-
1000-
1 I I I
10 11 12 13
0 63
^-0.02
-0.01
0.00
D
I
L
U
T
I
0
N
T
E
s
M
I
N
8
TIME, LOT
A-22
-------
MOULL INPUT DATA
DATE: ?//7/76 JULIAN DAY: 2.6 / SITE: //f
Simulation Start Time: OSOO LPT
Simulation End Time: /flQ LPT
Initial Concentrations:
NMHC _/ 73 ppmC; NO 0./5~? ppm; 00 O.OOS" ppm; N00/N0 # J0
-" X se— 0 —*
-------
9/17/76
Day 261
Peak Ozone = .15 ppm
Time of Peak = 1300-1400 LOT
4330
700
710
7?0
730 740 750
UTM EASTING
760
770
A-24
-------
September 17, 1976
Day 261
MIXING HEIGHT
M
I
v
I
N
G
H
E
I
G
H
T
M
E
T
E
R
.^
o
I I I I I
e 3 10 n 12 1:1
1?
TIME, LOT
A-25.
-------
MODEL INPUT DATA
DATE
: 7//9 /?(*
JULIAN DAY:
SITE:
Simulation Start Time: QfQO LPT
Simulation End Time: XfQO LPT
Initial Concentrations:
NMHC
03 Aloft:
ppmC; NO,, ^.00? ppm; 0, 0.06
ppm
?/NOY # 70_
A
Post 8 a.m. Emissions:
Hour
I
2.
J
V
NMOC Emission
Density,
kg-moles/km2 hr
0J'V
3.773
j,o$i
NMOC Emission
Fraction
0.17*
_fe_
2.S7t
0.Q9S
NO Emission
Defisity,
kg-molgs/km2 hr
A027
6,57*
0./ZZ.
&03I
NO Emission
Fraction
A-26
-------
7/19/76
Day 201
Peak Ozone = .15 ppm
Time of Peak = 1300-1400 LOT
-4330
700
710
720
730 740 750
UTM EASTING
760
A-27
-------
July 19, 1976
Day 201
M
I
X
I
H
G
H
E
I
G
H
T
M
E
T
E
R
S
MIXING HEIGHT DATA
6.63
-0.82
-6 01
9.00
0
I
L
U
T
T
6
N
R
H
T
E
M
I
H
8
17
TIME.. LOT
A-28
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SITE:
Simulation Start Time:_0fOQ LPT
Simulation End Time: /fOO LUT
Initial Concentrations:
NMHC ^,/2. ppmC; NO, 0.QO* ppm; 0Q
-------
8/8/76
Day 221
Peak Ozone = »12 ppm
Time of Peak = 1800-1900 LOT
4330
4320
4310
4300
429Q
0£
O
4280
4270
4260
4250
4240
700
710 720
730 740 750
DIM EASTING
.760 770
780
A-30
-------
August 8, 1976
Day 221
MIXING HEIGHT DATA
M
I
X
I
N
G
H
E
I
G
H
T
M
E
T
E
R
S
2000
1800-
0.03
-0 01
0.0Q
D
I
L
II
T
I
0
N
F:
A
T
E
M
T
i
N
16 17
TINF, LOT
A-31
-------
APPENDIX B
The following figures compare Level II predicted with observed temporal
patterns of ozone, CO, NO, NMHC, NO, N02, NMOC/NO. and N02/N0 . All pol-
s\ s\ /\
lutant concentrations are in points of ppm (ppmC for NMOC), and the time ii>
Local Daylight Time starting at 8 a.m.
B-l
-------
October 1, 1976
Day 275
9.25
6.20-
0.15-
0.10—
0.05-H
0.00
03
LOT
B-2
-------
October 1, 1976
Day 275
CO
LOT
B-3
-------
October 1, 1976
Day 275
NOX
NMHC
NO
6.5-
8.0-
2.5-
13 18
LOT
0.8-
0.25-
6 13 18
LOT
e.ee-
I
8 13
LOT
IS
N02
0.2-
8.0-
13 18
LOT
NMHC/NOX
29-
8-
I
8 13 18
LOT
N02XHOX
e-
8
I
13
LOT
13
B-4
-------
July 13, 1976
Day 195
0.25-
0.20-
0.15-^
0.10-
0.05—
03
T 1
I
12
I
14
LOT
I
16
18
J-5
-------
July 13, 1976
Day 195
1.0
0.8-
0.6-
0.4-
0.2
CO
LOT
B-6
-------
July 13, 1976
Day 195
NOX
WHC
NO
0.5
e.e
0.01
0.00
N02
8 13 18
LOT
NMHC/NOX
Z0 >••»
0-
8
13
LOT
18
0-
111 • .
8
I
13
LOT
13
B-7
-------
June 8,
Day 160
0.25—1
6.28—
0.15-
0.10-
0.05-
9 66
LOT
B-8
-------
June 8, 1976
Day 160
CO
2.0-
1.5-
1.0-
0.5-
0.0-
8
I
10
I
12
!
16
18
LOT
B-9
-------
June 8, 1976
Day 160
NOX
0.1-
0.0-
- S
HMHC HO
l-i 1 0.02-
Jl
0.00
1
8 13 18 8 13 18 8 13 13
LDT LDT LDT
N02
0.05-
6 08-
I
8 13 18
LDT
NNHC/HOK
50-
I
8 13 18
LDT
I
8 13 18
LDT
B-10
-------
June 7, 1976
Day 159
6.20
0.15-
0.10-J
6.05-
0.00
03
LOT
B-ll
-------
June 7, 1976
Day 159
.25-
1.00-
0.75-
0.50-J
0.25-
0.66-
I
18
CO
I
12
1
14
I '
16
13
LOT
B-12
-------
June 7, 1976
Day 159
NOX
0.0
0.00
6 13 18
8
N02 NMHC/NOX
0.05—1 1 50-
NOS/'NOX
9 00
8 13 18
8
8-
8
I
13 18
LOT
B-13
-------
June 8, 1976
Day 160
8.20
8.15-^
0.18-
8.85-
8.88
03
LDT
B-14
-------
June 8, 1976
Day 160
1 25-r-
CO
i.ee-
0.75-
0.50-
0.25-
9.00
LOT
B-15
-------
June 8, 1976
Day 160
0.1
0.8-
8
NOX
13
LOT
18
C.5-
e.e-
8
NilHC
13
LOT
16
0.82
0.88
HO
LOT
0.05-
0.90-
N02
8 13 18
LOT
58-
e-
NWC/NOX
8 13 18
LOT
NQ2/NGX
B-16
-------
August 25, 1976
Day 238
0.28-
0.15-
0.10-
0.05-
0.06-
03
1
I t 1
I
12
14
LOT
I
16
18
i-17
-------
August 25, 1976
Day 238
6.8-
CO
8.6-
0.4-
0.2-
0.0-
T I li III
8 9
I I I
16 11 12
LOT
i i i
13
t i I i
14
15
B-18
-------
August 25, 1976
Day 238
0.65-
8.8Q-
8
NOX
I
13
LOT
18
NMHC
0.5-
60-
•\
v. * • * • •
8 13 18
LOT
HO
0.02
8
8
LOT
8.025
N02
0.006
NMHC/NOX
ise
0-
8
13
LOT
18
N02/NOX
ti
I
13
LOT
18
B-19
-------
October 2, 1976
Day 276
0.15-
I
I1
0.05-
6.00-
I
19
I
14
LOT
16 18
B-20
-------
October 2, 1976
Day 276
CO
LOT
B-21
-------
October 2, 1976
Day 276
NOX
NMHC
NO
0.0
0.25
0.00
N02
0.2
0.0
NMC/NOX
50-
0-
8 13 18
LOT
H02/NOX
B-22
-------
September 17, 1976
Day 261
0.20-
0.15-1
0.10-
0.05-
0 00-
03
r I '
3 10
\
14
I '
16
LOT
18
B-23
-------
September 17, 1976
Day 261
CO
LOT
B-24
-------
September 17, 1976
Day 261
0 5
0.0
NOX
LDT
0-
8
NMHC
13
LOT
18
0.2
8.0
8
HO
0.2-
0.0-
N02
8
I
13
LDT
18
20-
8
NMHC/NOX
13
LDT
18
0-
N02XNOX
I
13 18
LDT
B-25
-------
July 19, 1976
Day 201
e.i5-
0.10-
0.05-
0.00-
03
I
14
LOT
16
18
B-26
-------
July 19 , 1976
Day 201
1.25-
1.00 —
0.75-
0.50-
8.25 —
CO
II T T I I I 1 I I I I
S1 HI
I I I I
11
LOT
12
13
14
B-27
-------
July 19, 1976
Day 201
NOX
6.1
0.0-
NMHC NO
0.5 ii i 0.92-
0.0-
8 13 16
LOT
8
13
LOT
0.00-
18 8
13
LOT
IS
N02
NMHC/NOX
NQ2/NQX
0.05-
8.00-
160-
8 13 18
LOT
8
13
LOT
e-
18
rrn
1
8 13 18
LOT
B-28
-------
August 8, 1976
Day 221
03
9.125
0.100—
LOT
B-29
-------
August 8, 1976
Day 221
3 —
2-
CO
LOT
B-30
-------
August 8, 1976
Day 221
NOX
NMHC
HO
0 Q5
0.80
0-
13
LOT
8.96-
18
8
i
13
LOT
18
N02
NWC/NOX
N02.--NOX
6 02
0.00
0-
N
6
13
LOT
16
8-
8
~
13
LOT
18
B-31
-------
APPENDIX C
The following figures contrast the trajectories calculated from
surface wind measurements using two different procedures. The solid
line shows the back trajectories calculated from the vector average of
all 25 RAMS sites. The dashed line corresponds to back trajectories
calculated on the basis of the nearest three stations, using a l/r2
weighting factor. In both cases, 10-minute average wind data were used,
but only hourly intervals are shown on the trajectories. All trajectories
start at 0800 LOT and end at the time of the observed peak.
C-l
-------
10/1/76
Day 275
Peak Ozone = .24 ppm
Time of Peak = 1 SOD-1 60,0J.DT
4330
4320
SfWILLE
• PA
* RAI
S SIM IONS
*
2
r"
f>
<&\r<
431 D
4300
^. 4290.
£ 428C
4270-
4260
4250 L
4240
,HO°So
700
710
720
730 740 750
UTM EASTING
60
770
780
C-2
-------
7/.13/76
'Day 195
Peak Ozone = .22 ppm
Time of Peak OAQO-J 700J.DT
4330
700
710 720
730 7^0 750
UTM EASTING
760 770
780
-------
6/8/76
Day 160
Peak Ozone = .22 ppm
Time of-teak =. 1700-1800_LDT_
4330
700
710
720
730 740
UTM EASTING
750
-760
780
C44
-------
6/7/76
Day 159
Peak Ozone = .20 ppm
Time of-Peak-=.a600J700J_DT
-4330
700
710
720
'30 740 750
UTM EASTING
C-5
-------
6/8/76
Day 160
Peak Ozone = .19
Time of Peak = UnD-O 500 _LDT.
4330
4320
431G
4300
42901
4280:
427Cf
4260
4250
4240
(^S-^'S^ ty . /»i
^^mm^^ 'x^l/
700
710
720
730 740 750
UTM EASTING
770
780
CI6
-------
8/25/76
Day 238
Peak Ozone = .19 ppm
Time of-E£ak O400-"1500 LOT
4330
700
710 720
730 740 750
UTM EASTING
760 770 780
-------
10/2/76
Day 276
Peak Ozone = .19 ppm
Time of Peak = 1700-1800 LOT
4330
700
710
720
730 740' 750
DIM EASTING
760
780
C-8
-------
9/17/76
Day 261
Peak Ozone = .15 ppm
Time of-Peak = -1300-14QQ_LDT
-4330
760 7/0 /bU
700
710
720
730 740 750
UTM EASTING
-------
7/19/76
Day 201
Peak Ozone = .15 -ppm
Time of Pp*Jcf 1300-1400
-4330
4320
-431
4300
•429
c:
o
428
-4-27
-4260
4250
-4240
700
710 720
730 740 750
UTM EASTING
• 760
770 780
Ci-10
-------
8/8/76
Day 221
Peak Ozone = .12 ppm
Time of-Reak =J800-1900 LOT
4330
4240
700
710 720
730 740 750
UTM EASTING
760
770
780
on
-------
APPENDIX D
The following tables summarize the Level III inputs for each
test case0
D-l
-------
MODEL INPUT DATA
DATE: /0/QtJ?t> JULIAN DAY: 2?S SITE:
V ^T~^~
Simulation Start Time: Q&OO LPT
Simulation End Time: /?00 LPT
Initial Concentrations:
NMHC _/,99 PpmC: NO, I, £36 ppm: 0, Q.V ppm: N07/N0 0.Z-S
' ' * m " ' "'"' A "*~ ' _j " """" ^- A
C. Aloft: ^, g^ ppm
0800 LPT Mixing Height, meters: ZSQ
Maximum Afternoon Mixing Height, meters:
Post c £ .r,. Emissi ons :
K'MOC Emission K'0v Emission
Deris iiy. NMOC Emission "Den's ity, NO Errnssior
KC-mo'ies/km2 hr Fraction ko-mol es/ktn2 hr Friction
_L_ 2.7/fc ^./37 ^4S
2. 2.7/^> ^>/3f £Hfc5"
_^_ ^'7^ X9./T? ^./6S ^./f^
JL_ <
2.7/fc ^./^7 &KS 0./91
Z.7IL P.'tf
2.7/4 0./3J
^./^2.
D-2
-------
MODEL INPUT DATA
DATE: 7/A5&/74 JULIAN DAY: /f5~
f
Simulation Start Time: 0fOQ LPT
Simulation End Time: ffOO LPT
Initial Concentrations:
NMHC #26 ppmC: NOV. 0.0^8
ppm; 0- 0.Q
0, Aloft: Q.V& ppm
0800 LPT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters:
Pos i £ a.m. Emissions:
SITE: ///
PDTTU NO-/NO
L- S
10 U!
y
9
h'NOC Emission
Oensi ty .
KG- moles /km2 nr
2LZ/C
2.7 I (.
2M
NMOC Emissior
Fracti on
/.otj
Dens i t.v.
kc-rnol es/krr>' h;
NO Emission
Froction
QMS
0,m
O.ttfS
QMS
0.3Q1
D-3
-------
MODEL INPUT DATA
DATE: &/08/7C* JULIAN DAY: /£O
SITE:
Simulation Start Time: QtOd LPT
Simulation End Time: ffOO LPT
Initial Concentrations:
NMHC 1.0$ PpmC; NO Q, 107 ppm; 0
X -" — -~ - - —
C7 Aloft: 0A/O pprr;
0800 LOT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters:
Post c c.rr, . Emissions:
0.0
ppm: NO-/NO
£
HOL
NMOC Emi ss i on
Density.
KC-mo'ies/km2 hr
OJ<>3
NMOC Emi ss ion
Fraction
3
y
T
0.'C3
O.tU
O.H3
0.Q15
QMS
NO . Emi s s i on
Density.
kg-mol es/krr;2 hr
O.Q
NO Emission
Fraction
0.0*0
0. OVQ
&OV0
D-4
-------
MODEL INPUT DATA
DATE: 4/07/76
JULIAN DAY:
SITE:
Simulation Start Time: O#OO LPT
Simulation End Time: tfc O LPT
Initial Concentrations:
NMHC /. £3 ppmC; NO, Q»\05 ppn"'1
//
ppm
0800 LOT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters: /fEO.
Post o £ . rr,. £missi ons :
ppm;
NHOC Emission
De n s i ty .
NO-nol es/'km2 hr
NMOC Emission
Fraction
O.OtV
0.001
NO Emission
Density
ko-mol es /KITT^ n;
.003
NO Errnssion
Fraction
z.
3
^
6
_7
2.7/C
2.7/6
2.7/fc
^?.S33
«^
^,/63
o.m
0.02X
o.ooi
O.otf
O.otf
Q.ttS
0.VLS
O.KS
o.on
0.0M
t.UI
0.070
O.oW
0.001
D-5
-------
MODEL INPUT DATA
DATE: 6/01/76, JULIAN DAY: f&O SITE:
Simulation Start Time: QfQO LPT
Simulation End Time: /?OQ LPT
Initial Concentrations:
NMHC /.Of ppmC: NOX 0JO? ppm; QS 0.O ppm; N02/NOX
(j~. Aloft: 0,fO ppm
0800 LDT Mixing Height, meters:
Maximum' Afternoon Mixing Height, meters:
Post & c.rr. Emissions:
IvMOC Emission NO Emission
Density. IvMOC Emission Density. NDv Err.issior
ttc-rnoles/km2 hr Fraction ko-mol es/km2 hr Friction
40/5 ft0VV Q.090
D-6
-------
MODEL INPUT DATA
DATE:
JULIAN DAY: 2.3^
SHE:
Simulation Start Time: QfOQ LPT
Simulation End Time: t?OO LPT
Initial Concentrations:
NMHC 2.0/
N0
0, Alofi: 0.O1 ppm
0800 LDT Mixing Height, meters:
ppm; 0,
Maximum Afternoon Mixing Height, meters:
Post 8 c .rr,. Emissions:
ppm; r,'02/NOx
jjOLir
I
z
J
y
y
£
NMOC Emission
Dens i T>P .
ko-moles/km2 nr
ft/43
NMOC Er.ission
Fracti on
NO Emissi or.
Dens i ty.
i: G-mol es /krrr- nr
0
NO Emission
Fraction
0.02JL
O.QU.
D-7
-------
MODEL INPUT DATA
DATE:
JULIAN DAY:
SHE:
Simulation Start Time: Q?OO LPT
Simulation End Time: /ffOQ LPT
Initial Concentrations:
NMHC 3.V7 ppmC: NOV O.ZL'J Ppm;
'~ ™ —-11 •- . .
&, Aloft:
ppm
0800 LDT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters: Jfr/O,
Post £ c.rr,. Emissions:
ppm; NO-/NO . 4 IS
10 ur
j^
3
K'MOC Emission
der.s 1 ty.
ko-mo'les/km2 hr
5,7/6
NMOC Emissior
Fraction
O./U
0./J3
NO Emission
Density.
kc-mol es/krrr2 hi
O.oW
o.m
NO . Emission
Fraction
0.0/t,
fi.Olb
0.0&
go*
am
D-8
-------
MODEL INPUT DATA
DATE: &?//?/?(, JULIAN DAY: 2&I SITE: ///
Simulation Start Time: QltOO LPT
Simulation End Time: /fOO LPT
Initial Concentrations:
NMHC 2. 0V PPmC; NO,. 0.'?Y ppm; C0 0, O ppm; h'00/NO . O.
A U £ X
C, Aloft: 0f&& ppm
0800 LPT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters:
Posi & c.rr,. Emissions:
K'MOC Emission NO, Emission
Density. K'MOC Lmission Density, NO Emission
nour \'. o-mo'l es/km: hr Fraction ko-mol es/krr*-2 hr Fraction
0,i3O 0.V&S Q.1*/
Q.OK 0.0Z3 Q.OIl
O,i2,3 OtOfit> ^.QZ3 Q.Oi3
Q.QiX, 0.023 0.013
OtOO(o 0.Q13 0,0ft
D-9
-------
MODEL INPUT DATA
DATE
: Q7//f/7(, JULIAN DAY: 201
SITE: /22.
Simulation Start Time: 08OO LPT
Simulation End Time: // 00 LPT
Initial Concentrations:
NMHC ^.?£ ppmC; NO
0, Aloft:
ppm
0800 LPT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters:
Post 8 a.m. Emissions:
0.O ppm; N02/NOX g, £5*
Hour
IIMOC Emission
Density,
kg-moles/km2 hr
A//7
O ij
NMOC Emission
Fraction
0.0*
0.° I?
0,001
NO Emission
Density,
kg-moles/kre2 hr
QMS
NO Emission
Fraction
D-10
-------
MODEL INPUT DATA
DATE: ?Ot?L JULIAN DAY: 2ZI
Simulation Start Time: &&OO LPT
Simulation End Time: /fOO LPT
Initial Concentrations:
NMHC ft 23 ppmC; NOX 6.033 ppm; 03 &Q ppm; N02/N0x Q. 1$
03 Aloft: ^.07 PPm
0800 LOT Mixing Height, meters:
Maximum Afternoon Mixing Height, meters:
Post & a.m. Emissions:
NMOC Emission NO Emission
Density, NMOC Emission Density, NO Emission
Hour kg-moles/km2 hr Fraction kg-moles/km2 hr Fraction
I 2.7/4 0.*&
f 1,333 O.IS* OStt
_!__ Z3&
10 0.°7*
3 Z.7/C 0X^3 f>.KS 1,375
i 2.7/4 0X03 0.V
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before cornplclinf:/
REPORT NO.
EPA 45Q/4-K?-nnQ
2.
4. TITLE AND SUBTITLE
An Evaluation of the Empirical Kinetic Modeling
Approach Using the St. Louis RAPS Data Base
3. RECIPIENT'S ACCESSION NO.
5 REPORT DATE
June 1982
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT
Gerald L. Gipson
9. PERFORMING ORGANIZATION NAME AND ADDRESS
10. PROGRAM ELEMENT NO.
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards/MDAD/AMTB
Mail Drop 14
Research Triangle Park, North Carolina 27711
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
Same
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The use of three different approaches for evaluating the Empirical Kinetic
Modeling Approach (EKMA) are described. The first two approaches consist of using the
model underlying EKMA to make predictions of peak ozone for comparison with observation
In one ase, the detailed St. Louis RAPS data base was used to develop model inputs.
In the second, a much more simplified data base was used as the basis for model inputs.
In general, the simplified approach produced better results. The final approach
consisted of comparing EKMA predictions of changes in ozone with those of a more complec
photochemical air quality simulation model. EKMA did not agree with the complex model
over the entire range of evaluation. However, when estimating the degree of control
necessary to lower peak ozone to the level of the standard, differences between EKMA
and the complex model and EKMA were usually less than 10%.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Ozone
Control strategies
Photochemical pollutants
Photochemical models
EKMA
OZIPP
b.IDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
8. DISTRIBUTION SI
Unlimited
. SECURITY CLASS (This ReporTj
Unlimited
21. NO. OF PAGES
177
20 SECURITY CLASS (This page)
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS E 01 TION i s OBSOL E T t
------- |