United States
Environmental
Protection
Agency
Office of Air duality
Planning and Standards
Research Triangle Park, NC 27711
EPA-450/4-90-006C
APRIL 1990
AIR
»EPA
URBAN AIRSHED MODEL
STUDY OF FIVE CITIES
Evaluation of Base Case Model Performance
for the Cities of St. Louis and Philadelphia
Using Rich and Sparse Meteorological Inputs
-------
EPA-450/4-90-006C
URBAN AIRSHED MODEL
STUDY OF FIVE CITIES
Evaluation of Base Case Model Performance
for the Cities of St. Louis and Philadelphia
Using Rich and Sparse Meteorological Inputs
By
Ralph E. Morris
Thomas C. Myers
Edward L. Carr
Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, CA 94903
EPA Project Officers:
Richard D. Scheffe, Office of Air Quality Planning and Standards
John C. Chamberlin, Office of Policy Planning and Evaluation
OFFICE OF AIR QUALITY PLANNING AND STANDARDS
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
APRIL 1990
-------
Disclaimer
This material has been funded wholly or in part by the United
States Environmental Protection Agency. It has been subject to
the agency's review, and it has been approved for publication as
an EPA document. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
Contents
List of Figures v
List of Tables viii
1 INTRODUCTION 1
Use of the Urban Airshed Model 1
The "Five Cities" UAM Study 2
2 DESCRIPTION OF THE CB-IV VERSION OF THE URBAN
AIRSHED MODEL 6
Use of the Smolarkiewicz Algorithm to Solve the
Advection Equation 7
Use of the CB-IV to Solve Photochemistry 8
3 DEVELOPMENT OF A PLANR UAM BASE CASE IN ST. LOUIS 11
Definition of the PLANR Base Case 11
Meteorological Conditions 11
Preparation of Inputs 12
Data Availability 12
Wind Field Preprocessor 17
Model Inputs 18
Diagnostic Simulations to Arrive at a Base Case 20
Diagnostic Run 1 22
Diagnostic Run 2 22
Diagnostic Run 3 24
4 EVALUATION OF THE PLANR UAM APPLICATION TO ST. LOUIS 34
Model Inputs 34
Historical UAM(CB-II) 34
RAPS UAM 34
SIMPLE UAM 35
89092r2 1
11
-------
Comparison of Performance 35
Discussion 45
Corrections for Model Bias in Calculations of Ozone Reductions
in Response to Emission Control Strategies 46
Uncorrected Bias 48
Decrement Approach 51
Percentage Approach 51
Summary 51
5 DEVELOPMENT OF A PLANR UAM BASE CASE IN PHILADELPHIA ... 56
Meteorological Conditions 56
Preparation of Inputs 61
Data Availability 61
Model Inputs 62
Diagnostic Simulations to Arrive at a Base Case 66
Diagnostic Run 1 66
Diagnostic Run 2 80
PLANR Base Case 87
6 EVALUATION OF THE PLANR UAM APPLICATION TO PHILADELPHIA 90
Comparison of Model Performance 90
Development of Inputs Using a Rich Data Base (the
Philadelphia Oxidant Study) 90
POS UAM Performance 92
Comparative Performance of PLANR UAM, POS UAM, and
UAM(CB-II) 92
Discussion 97
Corrections for Model Bias in Calculations of Ozone Concentrations
in Response to Emission Control Strategies 98
7 SUMMARY AND RECOMMENDATIONS 103
St. Louis Test 103
Philadelphia Test 105
Recommendations 106
References 108
Appendix A: Model Performance Statistics for Hourly and Daily Maximum
Ozone Concentrations for the RAPS, PLANR, and SIMPLE UAM
Applications to St. Louis
89092r2 1
111
-------
Appendix B: Wind Fields Used in Diagnostic Run 1 Application to
Philadelphia
Appendix C: Hourly Ozone Concentrations in Philadelphia
Predicted from Diagnostic Run 1
Appendix D: Time Series of Predicted and Observed Hourly NO, NO2, and CO
Concentrations in Philadelphia for Diagnostic Run 1
Appendix E: Wind Fields Used in Diagnostic Run 2 Application to
Philadelphia
Appendix F: Hourly Ozone Concentrations in Philadelphia Predicted
from Diagnostic Run 2
Appendix G: Time Series of Predicted and Observed Hourly Ozone, NO,
and CO Concentrations in Philadelphia for the UAM (CB-IV)
Using a Rich Data Base (POS UAM)
Appendix H: Hourly Ozone Concentrations in Philadelphia Predicted by
the UAM(CB-IV) Using a Rich Data Base (POS UAM)
Appendix I: Results of the. UAM Sensitivity Test for Philadelphia Using
Meteorological Inputs Developed from a Sparse Data Base
(PLANR UAM) and Air Quality Inputs Developed from a Rich
Data Base (POS UAM)
Appendix J: Comparison of Instantaneous FAA/NWS Wind Velocity
Observation with Hourly Average Wind Speeds
89092r2
IV
-------
Figures
3-1 Surface weather map at 0700 EST on 13 July 1976 .................. 13
3-2 950 mb constant pressure surface in the vicinity of St. Louis
at 1800 LSI 13 July 1976 showing 950 mb heights and winds .......... 14
3-3 950 mb constant pressure surface in the vicinity of St. Louis
at 1800 LSI 14 July 1976 showing 950 mb heights and winds .......... 15
3-4 The St. Louis modeling domain showing the location of the RAPS
ozone monitors ................................................ 16
3-5 Isopleths of predicted maximum daily ozone concentrations
with superimposed maximum daily observations for PLANR
diagnostic run //I ............................................... 23
3-6 Isopleths of predicted maximum daily ozone concentrations
with superimposed maximum daily observations for PLANR
diagnostic run #2 ............................................... 25
3-7 Isopleths of predicted maximum daily ozone concentrations
with superimposed maximum daily observations for PLANR
diagnostic run #3 ............................................... 26
3-8 Comparison of hourly predicted and observed ozone concentrations
at each ozone monitoring site for PLANR diagnostic runs 1, 2,
and 3 [[[ 28
4-1 Isopleths of RAPS UAM predicted maximum daily ozone
concentrations with superimposed maximum daily observations
for the St. Louis region on 13 July 1976 ........................... 36
4-2 Isopleths of SIMPLE UAM predicted daily maximum ozone
concentrations with superimposed daily maximum observations
-------
4-3 Comparison of predicted and observed hourly ozone concentrations
at each monitoring site for the RAPS, PLANR, and SIMPLE UAM 39
4-4 Region-wide maximum ozone concentrations calculated by the RAPS,
PLANR, and SIMPLE UAM for three scenarios in St. Louis,
13 July 1976 50
4-5 Use of the decrement approach to correct predicted region-wide
maximum ozone concentrations to demonstrate attainment of the
NAAgS 52
4-6 Use of the decrement approach to correct predicted region-wide
maximum ozone concentrations to demonstrate attainment of the
NAAQS 53
5-1 Geographical location of the Philadelphia airshed modeling
region 57
5-2 Philadelphia airshed modeling region 58
5-3 Synoptic surface weather map at 0700 EST on 13 July 1979 60
5-4 Temperature sounding for Dulles Airport on 13 July 1979 64
5-5 Low-level NOX emissions for a typical summer weekday in
the Philadelphia AQCR for 1979 67
5-6 Low-level total VOC emissions for a typical summer weekday in
the Philadelphia AQCR for 1979 68
5-7 Time series of predicted and observed hourly ozone concentrations
in Philadelphia for diagnostic run 1 70
5-8 Isopleths of predicted daily maximum ozone concentrations from
diagnostic run 1 with superimposed observations 76
5-9 Scatterplot, residual analysis, and model performance statistics
for hourly ozone concentrations in Philadelphia and diagnostic
run 1 78
5-10 Time series of predicted and observed hourly ozone concentrations
in Philadelphia for diagnostic run 2 81
5-11 Scatterplot, residual analysis, and model performance statistics
for hourly ozone concentrations in Philadelphia and diagnostic
run 2 88
89092r2 1
VI
-------
6-1 Predicted maximum daily ozone concentrations for the POS UAM with
superimposed daily maximum observations 93
6-2 Scatterplot, residual analysis, and model performance statistics
for hourly ozone concentrations in Philadelphia and POS UAM 94
89092r2
VII
-------
Tables
2-1 The Carbon Bond Mechanism-IV 9
3-1 Respeciation of emissions of hydrocarbons 21
4-1 Comparison of performance statistics for four modes of
application of the UAM to St. Louis for 13 July 1976 38
4-2 Predicted regional maximum ozone concentrations for the RAPS,
PLANR, and SIMPLE UAM applications for different VOC emission
reduction scenarios 49
4-3 Estimated reduction in VOC emissions required to meet attainment
of the ozone NAAQS 54
5-1 Routine surface meteorological sites in the vicinity of
Philadelphia with data available on 13 July 1977 59
5-2 Routine air quality observation in operation on 13 July 1979
in the vicinity of Philadelphia 59
5-3 Hourly varying metscalars used in the PLANR UAM application
to Philadelphia 65
5-4 Emission totals (moles/day) for reactive hydrocarbons and NOX
for the CBM-II and CBM-IV UAM applications for a typical summer
weekday in 1979 in the Philadelphia AQCR 69
6-1 Comparison of performance statistics for the UAM(CB-II), POS UAM,
and PLANR UAM applications to Philadelphia for 13 July 1979 96
6-2 Predicted region-wide maximum ozone concentrations for the
POS UAM, PLANR UAM, and UAM(CB-II) for VOC emission reduction
scenarios 99
89092r2
Vlll
-------
6-3 Percent VOC emission reduction required to reduce the calculated
region-wide maximum ozone concentration to 12 pphm in
Philadelphia based on the POS UAM, PLANE UAM, and UAM(CB-II)
applications and different approaches for correcting model
bias 101
89092r2
IX
-------
1 INTRODUCTION
The job of reducing ozone concentrations to levels below the National Ambient Air
Quality Standard (NAAQS) has proven to be far more difficult than was thought when
the Clean Air Act was passed and amended. The level of ozone precursor emissions
remains too high; either emission reductions have been too small or have been
required of the wrong sources, or both.
A plethora of technical explanations has been offered for failure to attain the ozone
standard. These include perceived weaknesses in the attainment planning process
(Federal Registrar, Vol. 52, No. 226, November 24, 1987; OTA 1988a,b,c), incomplete
understanding or recognition of the anthropogenic and natural factors that cause ele-
vated tropospheric ozone levels (Science, 1988), the failure to consider the effects of
natural emissions (Chameides et al., 1988; Morris et al., 1989), use of a simplistic
modeling approach (OTA, 1988a; Seinfeld, 1988a; Burton, 1988), and failure to reduce
the amount of emissions intended, either through overestimates of the effectiveness
of control technology or failure to account for certain categories of emission
sources. The EPA, after lengthy consideration, has proposed a comprehensive policy
that includes major changes in the planning process for reducing ozone concentra-
tions (Federal Registrar, Vol. 52, No. 226, November 24, 1987). These changes
include improvements in modeling practices and requirements for improving the data
to support improved modeling practices. The EPA is now evaluating public com-
ments on the proposed policy.
USE OF THE URBAN AIRSHED MODEL
The EPA has recommended the Urban Airshed Model (UAM) as the preferred
approach for estimating emission controls needed to attain the ozone NAAQS. How-
ever, in the past there has been a reluctance to use the UAM because of the per-
ception that it requires using data from costly intensive measurement studies and
requires extensive computational resources. Most of the cost of applying the UAM is
attributed to the practice of conducting an extensive evaluation of UAM perform-
ance, which usually entails many diagnostic simulations. This evaluation enables us
to understand why the UAM performs as it does for a particular application and, if
deemed necessary, to take actions to improve model performance. Historically, it
has been expected that the UAM will calculate hourly ozone concentrations to within
approximately 15 to 20 percent of the observed peak value (Seinfeld, 1988a; Burton,
1988). More recent applications of the UAM to the Los Angeles basin have used
89092r2 2
-------
routinely available meteorological data and predicted observed ozone levels with a
hi*h degree of skill (Seinfeld, 1988a; Burton, 1988} Hogo, Mahoney, and Yocke
1588). A recent application of the UAM to the New York metropolitan area used
simple inputs, i.e., constant wind fields and mixing depths (Rao 1987).
This simplified use of the UAM, relying on routinely available data and reducing the
requirement for strict evaluations of model performance, offers air quality managers
a practical air quality assessment tool for identifying emission control strategies
that demonstrate attainment of the ozone NAAQS. This simplified approach is cal-
led Practice-for-Low-Cost-Airshed-Application-for-Nonattainment-Regions
(PLANR). The PLANR use of the UAM requires almost the same quantity and
quality of inputs as EKMA, and the overall application cost is substantially reduced.
The possible exception is the emissions inventory, which in PLANR applications
should contain the same spatially (horizontally and vertically) and temporally varying
emissions used in standard UAM applications (such detail is necessary to account for
the differing reactivities of VOC emissions). However, local agencies generally have
emissions inventories at hand; in addition, UAM input inventories can be readily esti-
mated from existing national emissions inventories (e.g., the National Acid Precipi-
tation Program 1980 and 1985 inventories). Knowledge of current emission rates is
needed to estimate the emission controls required to achieve attainment of the
ozone NAAQS.
The PLANR use of the UAM may not be appropriate for all nonattainment regions.
When attainment is expected to be imminent, improved methods for using EKMA
may be adequate. In other, more complex situations, such as the Los Angeles basin,
the Houston region, and the New York Metropolitan area among others, the com-
plexity of meteorological conditions and the emissions distribution and the severity
of the ozone attainment problem probably require a more detailed application of the
UAM. The application of UAM to these more complex situations, called Practice-of-
Airshed-Application-in-Complex-Regions (PACR), would involve more extensive
model performance requirements and hence more diagnostic simulations, and a
resultant increase in costs. However, even for a complex nonattainment region, the
PLANR approach would probably be more comprehensive and reliable than EKMA for
estimating the controls needed to achieve ozone attainment.
THE "FIVE CITIES" UAM STUDY
The EPA has funded a study of the PLANR approach in five urban areas in the U.S.
(New York, St. Louis, Atlanta, Philadelphia, and Dallas-Ft. Worth). The main objec-
tives of this "Five Cities" study are to:
(1) Demonstrate the usefulness of PLANR for air quality planning;
(2) Determine the effects of alternative fuels and alternative Reid vapor
pressure values for fuels on urban ozone concentrations;
89092r2 2
-------
(3) Demonstrate the use of PLANR to evaluate SIP control strategies and
compare results with those obtained with EKMA; and
(4) Transfer the UAM model, modeling data bases, and applications tech-
nology to the states for use in future SIPs.
In addition, the study includes two city-specific analyses:
(1) For the St. Louis and Philadelphia areas, comparison of the PLANR use of
the UAM (i.e., using only routinely available data) with applications of
the UAM that use an extensive data base; and
(2) The effects of biogenic emissions on anthropogenic emission reductions in
the Atlanta area.
Previous reports on the "Five Cities" study have documented the evaluation of alter-
native fuel emission scenarios for the New York metropolitan area and the city of St.
Louis (Morris et al., 1989a), the use of the UAM to evaluate the effects of biogenic
emissions for the Atlanta area (Morris et al., 1989b), and the demonstration of the
PLANR use of the UAM for the city of Atlanta and the Dallas-Fort Worth metroplex
region (Morris et al., 1989c). This report describes evaluations of PLANR applica-
tions of UAM to the St. Louis and Philadelphia regions.
St. Louis
For the application to St. Louis, an ozone episode on 13 July 1976 was selected for
modeling. During this period the Regional Air Pollution Study (RAPS) collected data
from an intensive measurement network, thus providing a basis for evaluating the
PLANR results. Four UAM applications to St. Louis, using different types of inputs,
were compared:
UAM(CB-II) The EPA exercised the 1978-1980 version of the UAM (which
incorporates the CB-II chemical mechanism) and its preproces-
sors using the RAPS data base (Schere and Shreffler, 1982;
Cole et al., 1983).
RAPS UAM The updated UAM (which incorporates CB-IV) was exercised
using the RAPS data base. Wind fields, mixing heights, and
boundary conditions were prepared using data from RAPS and
current UAM preprocessors.
PLANR UAM The UAM(CB-IV) was exercised using only routinely available
meteorological and air quality data and current UAM prepro-
cessors; initial and boundary conditions were assumed to be
89092r2 2 clean (25 ppbc VOC and 1 ppb NOX).
-------
SIMPLE UAM The UAM(CB-IV) was exercised using only RAMMET preproces-
sed meteorological data (i.e., meteorological observations from
a single surface site and mixing heights from the closest upper-
air site), resulting in spatially constant winds and mixing
heights. Boundary conditions were assumed to be clean.
A comparison of results from these four UAM simulations will provide insight into
how data availability affects model performance and how well the model simulates
>: mission -~o .-;.: o I ^canaries.
Philadelphia
For the Philadelphia Air Quality Control Region, an ozone episode on 13 July 1979
was selected for modeling. During this period there were several special surface
meteorological and air quality monitors in operation as part of the Philadelphia oxi-
dant study (POS). However, since the POS was not in full operation until 18 July
1979, no special upper-air sites were in operation on 13 July 1979.
Considerable effort went into the preparation of input files for the UAM (CB-II) in a
past study (Haney and Braverman, 1985). After extensive analysis of the data collec-
ted from the POS, several unique interpolation schemes (e.g., urban heat island
effects) were used in the preparation of the UAM modeling inputs. These UAM
modeling inputs offer a basis for evaluating the PLANR procedures for applying the
UAM where only routine data are used in the input preparation. Three different
UAM applications to Philadelphia were compared:
UAM(CB-II)
POS UAM
PLANR UAM
The UAM (CB-II) was exercised using inputs prepared by Sys-
tems Applications, Inc. and the EPA using routine data and
data from the extensive POS data base (Haney and Braverman,
1985). Initial and boundary conditions were interpolated from
all available air quality data.
The UAM (CB-IV) was exercised using the same meteorological
inputs as for the UAM (CB-II). Emissions, initial concentra-
tions, and boundary conditions were similar to those used in the
UAM (CB-II) run only for the CB-IV chemical mechanism. Rou-
tine data and data from the POS were used.
Modeling inputs for the UAM (CB-IV) were developed using only
routinely available meteorological and air quality data and
using current UAM preprocessors.
89092r2 2
-------
A comparison of the results from the UAM (CB-II) and POS UAM, which use the same
meteorological and air quality inputs, gives a comparison of two different models
(the UAM (CB-II) and UAM (CB-IV)) using the same inputs. A comparison of results
from the POS UAM and PLANR UAM will give a comparison of two different inputs
using the same model (UAM(CB-IV)).
89092r2 2
-------
DESCRIPTION OF THE CB-IV VERSION OF THE
URBAN AIRSHED MODEL
The Urban Airshed Model (UAM) is a three-dimensional grid model designed to calcu-
late the concentrations of both inert and chemically reactive pollutants by simula-
ting the various physical and chemical processes that take place in the atmosphere.
The basis of the UAM is the atmospheric diffusion or species continuity equation.
This equation represents a mass balance in which all of the relevant emissions, trans-
port, diffusion, chemical reaction, and removal processes are expressed in mathe-
matical terms. Based on the grid concept, the model is generally employed to simu-
late an 8- to 72-hour period during which episodic meteorological conditions persist.
Because the model can resolve both spatial and temporal features of the concentra-
tion field, it is well suited to the analysis of future control strategies and their
effects on air quality in various parts of the modeling region. Before the model is
used for such an analysis, its ability to replicate measurements from an historical
ozone episode is tested. Model inputs are prepared from observed meteorological,
emission, and air quality data for a particular day or days. Once the model inputs
have been adjusted within the range of their uncertainty so that the model performs
within prescribed levels, the emission inventory can be changed to represent assump-
tions about future emission scenarios. The model is then re-run with the forecasted
emissions, and the resulting hourly ozone patterns are what the model predicts is
likely to occur under meteorological conditions similar to the historical episode.
The UAM is the only air quality model recommended by the EPA for photochemical
or reactive pollutant modeling applications involving entire urban areas (EPA,
1986). The EPA guidelines refer to the 1978-1980 version of the UAM; the formula-
tion of that version is discussed by Ames and others (1985a,b). Many improvements
to the UAM have been made over the last 10 years. The two most significant are:
Incorporation of the latest version of the Carbon-Bond Mechanism, the CB-IV
(Gery, Whitten, and Killus, 1988)
Use of the Smolarkiewicz algorithm for advection (Smolarkiewicz, 1983)
89092r2 2
-------
USE OF THE SMOLARKIEWICZ ALGORITHM
TO SOLVE THE ADVECTION EQUATION
Grid-based air quality simulation models require a numerical approximation of the
horizontal advection terms in the species conservation equations. The 1978-1980
version of the Urban Airshed Model (UAM) used a variant of the Sharp and Smooth
Transport Algorithm (SHASTA) originally formulated by Boris and Book (1973).
Since 1977 there have been many comparative studies of advection schemes. Exam-
ples of such studies relevant to the UAM are those of Zalesak (1970), Schere (1983),
Chock and Dunker (1983V )). Smolarkiewicz (1983), and Yamertino and
Scire (1985). In each of these studies an idealized scalar function (a cone, block,
ellipse, or cosine wave) representative of a concentration distribution is advected by
a rotating wind field (constant angular velocity). The rotating wind field provides a
range of Courant numbers, depending on the radial distance from the center of the
domain. The degree to which the attributes of the idealized function (total mass,
peak value, mean value, and gradients) are preserved indicates the accuracy of the
scheme.
The above studies showed that a number of advection schemes were more accurate
than SHASTA, as measured by the idealized tests. For the UAM, a number of
specific requirements constrained the selection of alternative to SHASTA. First, it
is important that the scheme be positive definite, i.e., that it not result in negative
concentrations. Second, the scheme should use forward time differencing to mini-
mize storage requirements and to insure compatibility with the chemical mechanism
numerical solution scheme. Third, the ability of an advection scheme to represent
the magnitudes and locations of peak concentrations is of major importance in regu-
latory applications. Fourth, to handle complex airflows, the scheme should display
relatively uniform accuracy over a wide range of Courant numbers (i.e., wind
speeds).
Further review indicated that the advection scheme developed by Smolarkiewicz
(1983) represented the best combination of accuracy and economy. The
Smolarkiewicz scheme is conceptually similar to SHASTA in that a highly diffusive
transport step is followed by an anti-diffusive correction step. The transport step is
essentially the well-known "upstream" finite-difference scheme. The correction step
involves a second exercise of the upstream finite difference scheme, substituting the
anti-diffusive velocity for the actual velocity. The Smolarkiewicz scheme is positive
definite and forward in time, and can be used in either a time-split or multidimen-
sional mode. Smolarkiewicz demonstrated that the scheme was superior in both
accuracy and economy to the multidimensional generalization of SHASTA formulated
by Zalesak (1979). In applications to the Los Angeles area and Kern County, Califor-
nia, the UAM with the Smolarkiewicz algorithm produced results that were more
accurate than those produced in earlier applications using SHASTA (Hogo, Mahoney,
and Yocke, 1988; Whitten et al., 1985).
89092r2 2
-------
USE OF THE CB-IV TO SOLVE PHOTOCHEMISTRY
The latest version of the Carbon-Bond Mechanism (CBM-IV) was recently implemen-
ted in the UAM (Gery, Whitten, and Killus, 1988). Whenever a new chemical kinetics
mechanisms is merged into a complex air quality simulation model, the predictive
capabilities and solution speed of the new computer code require optimization and
evaluation. This process is even more important now because the recent gas-phase
chemical kinetics mechanisms (CAL, RADM, and the CBM-IV) are larger than pre-
vious mechanisms, and therefore require significantly more computing time. The
CP.W.-IV reactions are shown in Table 2-1. There are some minor differences
between the original CB-IV and the version implemented in the UAM. For the
examination of ethanol (ETOH) blended fuels, ETOH was added as a species. The
ethanoi reaction is ETOH + OH * ALD2 + HO2 with a rate constant of 4,300 1/ppm-
min. Both numerical and chemical improvements were made to the CB-IV implemen-
tation in the UAM so that computational speed could be increased and solution
uncertainty diminished.
The implementation of the CB-IV in the UAM used a modified Crank-Nicholson
algorithm for the simultaneous solution of the differential equations that represent
the chemical changes for each species. This numerical integration scheme produced
results that were within a few percent of those obtained with the previous (Gear)
algorithm over a wide range of atmospheric conditions (Morris et al., 1989a,c).
89092r2 2
-------
TABLE 2-1. The Carbon Bond Mechanism-IV.*
Reaction Rate Data
Number
1)
2)
3)
4)
i)
6)
7)
8)
9)
10)
11)
12)
13)
14)
15)
16)
17)
1»)
19)
20)
21)
22)
23)
24)
2t>)
26)
27)
26)
29)
30)
31)
32)
33)
34)
35)
36)
37)
38)
39)
40)
41}
42)
43)
44)
45)
46)
47)
48)
49)
SO)
51)
Reaction1
03
0
0
0
03
03
03
N03
N03
N03
NO
NO
OH
OH
HONO
OH
OH
H02
H02
OH
H02
H02
OH
OH
FORM
FORM
FORM
ALD2
ALD2
ALD2
C203
C203
C203
C203
N02
0
NO
N02
N02
NO
N02
03
03
010
010
OH
H02
N03
NO
N02
N02
N205
N205
NO
N02
NO
HONO
HONO
HONO
N02
HN03
NO
N02
PNA
PNA
H02
HOZ
H202
H202
CO
OH
FORM
FORM
0
N03
0
OH
N03
ALD2
NO
N02
PAN
C203
H02
OH
H20
+ H20
+ H20
H20 >
.....>
-h\a->
.....>
.....>
.....>
.....>
.....>
.....>
.....>
.....>
.....>
.....>
-hv3->
.....>
.....>
.....>
-h\3->
.....>
-....>
.....>
.....>
_....>
-h\6->
_....>
.....>
.....>
.....>
.....>
NO + 0
03
N02
NO
N02
0
01D
0
2.000H
H02
OH
0.89N02 + 0.890 + 0.11NO
NO + N02
2.00HN03
N03 + N02
2.00N02
2.00HONO
HONO
OH
N02
NO
HN03
N03
OH
PNA
H02
N02
H202
+ NO
N02
+ N02
N02
2.000H
H02
H02
CO
CO
OH
HN03
CO
2.00H02
HN03
X02
X02
H02
+ OH
C203
C203
FORM
N02
PAN
C203 + N02
2.00FORM +
0.79FORM +
0.790H
X02
CO
CO
CO
FORM
2.00X02
0.79X02
2.00H02
H02
2.00H02
0.79H02
Pre-factor
(ppm"nimn-l)
Temp. Factor Rate Constant (? 298K
exp((-E/R)/T)
(ppnTnmin-l)
8.383 E+04
2.643 E+03
1.375 E+04
2.303 E+02
3.233 E+02
1.760 E+02
5.300 E-02
*EXP( 1175/T)
*EXP(- 1370/T)
*EXP( 687/T)
*^XP{ 602/T)
-TX,P(~ .'
1.
3.
2.
2.
3.
1.
147
260
344
100
390
909
3.660
7.
1.
2.
2.
1.
6.
1.
9.
1.
1.
7.
5.
1.
2.
1.
8.
7.
2.
4.
3.
1.
849
900
110
600
600
554
975
770
500
537
600
482
640
876
909
739
690
550
720
220
500
4.302
9.
300
1.739
1.
037
E+05
E+03
E+01
E+01
E+04
E+01
E+02
E-06
E+16
E-05
E-ll
E+02
E-01
E+03
E-05
E+03
E+03
E+02
E+15
E+03
E+01
E-10
E-01
E+03
E+02
E+04
E+04
E-01
E+04
E+04
*EXP(
*EXP(-
*EXP{-
*EXP(
*EXP(-
*EXP{
390/T)
940/T)
580/T)
250/T)
1230/T)
256/T)
*EXP(-10897/T)
*EXP(
*EXP(
*EXP(
*EXP(
*EXP(
*EXP(
530/T)
806/T)
713/T)
1000/T)
240/T)
749/T)
*EXP(-10121/T)
*EXP(
*EXP(
*EXP(
*EXP(-
*EXP(-
*EXP(-
*EXP(
380/T)
1150/T)
5800/T)
187/T)
1550/T)
986/T)
250/T)
3.700
7.915 E+03 *EXP( 250/T)
1.180 E-04 *EXP( 5500/T)
5.616 E+18 *EXP(-14000/T)
3.700 E+03
9.600 E+03
6.521 E+03
*EXP(- 1710/T)
see notes
4.323 E+06
2.664 E+01
1.375 E+04
2.309 E+03
2.438 E+03
4.731 E-02
5.300 E-02xk1
see notes
4.246 E+05
3.260
1.000 E+02
2.999"
3.390 E+Olxk!
4.416 E+04
5.901 E-01
1.853 E+03
1.900 E-06
2.776
1.539 E-04
1.600 E-ll
9.799 E+03
1.975 E-Olxk,
9.770 E+03
1.500 E-05
1.682 E+04
2.179 E+02
1.227 E+04
2.025 E+03
5.115
6.833 E+03
4.144 E+03
2.181 E-01
2.550 E-01xk39
2.520 E+03
3.220 E+02
1.500 E+04
see notes
see notes
2.370 E+02
9.300 E-01
6.360 E+02
2.400 E+04
3.700
see notes
1.831 E+04
1.223 E+04
2.220 E-02
3.700 E+03
9.600 E+03
2.100 E+01
*As currently implemented in the UAM (CB-IV), isoprene is not explicitly
treated as a separate species, and ethanol has been added to the CB-IV.
89092
870<+8r
88097
88151
(Continued)
-------
TABLE 2-1 Concluded.
Number
52)
t>3)
54)
55)
»6)
57)
58)
5*)
60)
61)
62)
63)
64)
6b)
66)
67)
68)
«»)
70)
71)
72)
73)
74)
75)
76)
77)
78)
79)
80)
81)
PAR * OH
ROR
ROR
ROR 4 N02
0 4 OLE
OH 4 OLE
03 4 OLE
N03
0
OH
03
OH
T02
OH
ORES
CrtO
OPEN
OPEN
OH
OH
0
OH
03
* OLE
4 ETH
4 ETH
4 ETH
4 TOL
4 NO
T02
* CRES
4 N03
4 N02
OPEN
4 OH
4 03
4 XYL
4 MGLY
HGLY
4 ISOP
4 ISOP
4 ISOP
H03 4 ISOP
X02 4 NO
X02 4 X02
X02N 4 NO
Reaction1
Reaction Rate Data
-h\£->
-h\2->
0.87X02 * 0.13X02N
0.11ALD2 4 0.76ROR
1.10ALD2 4 0.96X02
0.04X02N 4 0.02ROR
H02
0.63ALD2
0.30CO
0.22PAR
FORM
H02
0.50ALD2
0.44H02
- PAR
0.91X02
O.U9X02N
FORM
1.70H02
X02
0.22AL02
FORM
0.08X02
0.56T02
0.90N02
CRES
0.40CRO
0.300PEN
CRO
C203
X02
C203
0.03ALD2
O.U3X02
0.76H02
0.70H02
0.80MGLY
X02
C203
0.60H02
0.50X02
0.90PAR
X02
0.40MGLY
0.20ALD2
FORM
0.20MGLY
0.44H02
X02N
N02
0.38H02
0.20FORM
0.200H
AL02
PAR
0.74FORM
0.22X02
4 X02
0.i3CO
O.iOOH
4 FORM 4 ALD2 *
4 N02 - PAR
4 0.70X02 4 CO 4
4 0.300H
4 1.56FORM 4 H02 4
4 0.42CO 4 0.12H02
4 0.36CRES 4 0.44H02 4
4 0.90H02 4 0.900PEN
4 H02
4 0.60X02 4 0.60H02 4
4 HN03
H02
2.00CO
FORM
0.62C203
0.69CO
0.20MGLY
0.50X02
1.10PAR
C203
H02
0.80AL02
0.50CO
4 FORM
4 0.20C203
4 0.13X02N
4 0.40AL02
4 0.10PAR
4 U.100H
Pre-factor
(ppm~nntin~l)
Temp. Factor Rate Constant 9 298K
exp((-E/R)/T) k98 (pp
0.11H02
0.11PAR
0.94H02
2.10PAR
0.2«X02 4
0.02X02N 4
4 CO
4 2.00H02 4
4 0.70FORM 4
4 0.080H 4
4 0.20CRES 4
4 0.30T02
4 CO
4 0.550LE 4
4 0.45ETH +
4 0.67H02 4
4 l.OOETH 4
4 0.55ETH 4
4 0.06CO 4
1.203 £403
6.250 £416
9.545 £404
2.200 £404
1.756 £404
7.740 £403
2.104 £401
1.135 £401
1.540 £404
3.000 £403
1.856 E4Q1
3.106 £403
1.200 £404
2.500 £402
6.100 £404
3.250 £404
2.000 £404
9.040
4.400 £404
8.030 £-02
2.453 £404
2.600 £404
9.640
2.700 £404
1.420 £405
1.800 £-02
4.700 £402
1.200 £404
2.550 £401
1.000 £403
*EXP(- BOOO/T)
*EXP(-
*EXP(
324/T)
504/T)
*EXP(- 2105/T)
EXP(- 792/T)
*EXP(
*EXP(.
*EXP(
411/T)
2633/T)
322/T)
*EXP(.
*EXP(
500/T)
116/T)
EXP( 1300/T)
1.203 E+03
1.371 £+05
9.545 E+04
2.200 E*04
5.920 E+03
4.200 E+04
1.800 E-02
1.135 E*01
1.080 E*03
1.192 £404
2.700 E-03
9.150 E*03
1.200 £404
2.500 £402
6.100 £404
3.250 £404
2.000 £404
9.040 «
4.400 £404
1.500 £-02
3.620 £404
2.600 £404
9.640 x
2.700 £404
1.420 £405
1.800 E-02
4.700 £402
1.200 £404
2.000 £403
1.000 £403
89092
88097
88151
10
-------
DEVELOPMENT OF A PLANR UAM
BASE CASE FOR ST. LOUIS
DEFINITION OF THE PLANR BASE CASE
One of the key components of the PLANR use of the UAM is a limitation on the
number of diagnostic simulations used to arrive at a base case. This is achieved by
relaxing the strict model performance standards expected of the UAM in the past.
Although the goal is to achieve a satisfactory level of performance with as few
diagnostic simulations as possible, the model must show some skill in predicting
ozone observations in order to have confidence that the model will respond properly
to changes in emissions.
The minimal performance goal in the past was to have the predicted regional maxi-
mum ozone concentration be within 30 percent and in the general location of the
peak observed value. Model performance has been considered good if the predicted
peak ozone is within 15 percent. Clearly, when model inputs are based only on
sparse routine data, rather than intensively measured data as in the past, model per-
formance cannot be expected to always be as good in the past. However, there
should be some minimal expectations of model performance since incorrect charac-
terization of base case ozone concentrations may lead to incorrect calculations of
ozone reductions due to alternative emission inputs.
The protocols for several recent UAM studies of the impacts of California offshore
drilling emissions (Haney et al., 1986; Yocke et al., 1985) defined a minimal model
performance standard as follows: (1) the UAM-predicted regional maximum ozone
concentration should be within 20 percent and in the general location of the observed
maximum, and (2) the UAM predicted maximum at the location of the observed
maximum should be within 30 percent of the observed value.
We adopted this model performance standard for the PLANR use of the UAM for
St. Louis. It should be emphasized, however, model evaluation should always address
the issue of whether the right answer is being obtained for the right reason.
METEOROLOGICAL CONDITIONS
The ozone episode on 13 July 1976 was chosen as the base case. During the early
morning hours on 13 July a weak warm front extended across the Upper Mississippi
89092r2 2
11
-------
Valley through southwestern Iowa and across central Missouri. This is illustrated in
the daily weather map depicted in Figure 3-1. To the northeast of the warm front a
high pressure system was slowly moving eastward. Circulation about the high-pres-
sure system, which extended to the St. Louis area, was from the northeast during the
morning hours (see Figure 3-2, the 950 mb map at 0600 CST). The winds became
more southeasterly as the high pressure to the north moved eastward (see Figure 3-3,
the 950 mb map at 0600 CST). An examination of the surface winds at Lambert
Field reveals a significant wind shift from the northeast to the west at around 1130
CST, marking the passage of the warm front through St. Louis. Surface winds at
Lambert Field remained out of the west to southwest the remainder of the day. By
0600 CST the warm front had passed St. Louis but had not yet reached Salem, IL (see
Figure 3-3). Thus it appears on 13 July 1976 that air parcels originating early in the
morning at St. Louis would travel west of St. Louis, but then come back over the city
later in the afternoon.
PREPARATION OF INPUTS
An important component of the PLANR use of the UAM is consistency in the proce-
dures used to prepare model inputs for different cities. Procedures to be followed in
preparing UAM inputs from limited data must be flexible. These procedures are cur-
rently being refined and evaluated to determine the optimum methodologies for
generating UAM inputs from limited data.
The preprocessor programs supplied with the 1978-1980 version of the UAM generally
rely on intensively measured data; the additional data needed to obtain gridded fields
of input parameters for the UAM are interpolated from these measurements. Over
the last 10 years many of these programs have become outdated. More recent appli-
cations of the UAM have prepared input data using techniques that have been
developed on a case-by-case basis and are tailored to the available data. In the fol-
lowing paragraphs we discuss the procedures used to prepare meteorological and air
quality inputs for the PLANR study of St. Louis.
Data Availability
Routine data collection in or near St. Louis for 13 July 1976 consists of six surface
meteorological observation sites located at airports (i.e., Federal Aviation Admini-
stration sites) surrounding St. Louis (see Figure 3-4). There were no routine air
quality or upper-air measurement sites. For the PLANR application of UAM to St.
Louis, data from the six surface sites were used along with twice-daily upper-air
observations from two sites outside of St. Louis: Salem IL, approximately 120 km
east of St. Louis, and Monett, MO, approximately 340 km west-southwest of St.
Louis. Since only routine data are being used, transport conditions in this area may
be incorrectly characterized because of the absence of surface meteorological data
within the urban core.
89092r2 2
12
-------
TUESDAY, JULY 13, 1976
S'JRFACE WEATHER MAP CHJ
JN- STATION WEATHEP '
A' I CO * M £ S T t
FIGURE 3-1. Sxirface weather map at 0700 EST on 13 July 1976.
89092
13
-------
.39 739 839 939 1039 1139 1239 1339 1439
QI~I 11111111111111111111 1111111111111111111111111111111111
10 20 30 40 50 60 70 80
ffm4808
- 4708
- 4608
4508
4408
4308
- 4208
4108
- 4008
- 3908
- 3808
3708
FIGURE 3-2. 950 millibar constant pressure surface in the vicinity
of St. Louis at 1800 LST 13 July 1976 showing 950 nfo heights (m msl)
and winds.
89092
14
-------
739 839 939 1039 1139 1239 1339 1439
Little Rock
576
I ! I I I I I I I I I t I I I I i I I I I I I 1 i I I I I I i I i i i i i i i i i I i i i i i > i i M. i I i i i i > i i i i I i i i r
10 20 30 40 50 60 70 80
- 4008
- 3908
- 3806
3708
FIGURE 3-3. 950 millibar constant pressure surface in the vicinity
of St. Louis at 1800 LST 14 July 1976 showing 950 nb heights (m msl)
and winds.
89092
15
-------
o
'
! 1
! .! !
CTTY >^v
7
§ !
^ 1
KM* 1
JWATIKLOO
. |
, t
!
1
i
FIGURE 3-4. The St. Louis nodeling doitain showing the location of the
RAPS ozone monitors.
88151
89063
89092
16
-------
Wind Field Preprocessor
One of the most important UAM inputs is the wind field. A key component in the
PLANR use of the UAM will be use of the Diagnostic Wind Model developed by Sys-
tems Applications, Inc. for the EPA (Douglas and Kessler, 1988). This model was
developed to calculate wind fields for regions of complex terrain for which wind data
were sparse (Morris et al., 1987, 1988).
The DWM was used to generate hourly gridded wind fields for both the PLANR and
RAPS UAM simulations. This model incorporates local surface and upper-air obser-
vations, where available, and provides some information on terrain-induced airflow in
regions where local observations are unavailable. The DWM is formulated in terrain-
parallel coordinates. Wind fields are generated in a two-step procedure.
In Step 1, a domain-mean wind, which is obtained from a representative upper-air
observation, is adjusted for terrain effects. These include the kinematic effects of
terrain (the lifting and acceleration of the airflow over terrain obstacles), thermo-
dynamically generated slope flows, and blocking effects. Step 1 produces a spatially
varying gridded field of u and v for each vertical layer within the model domain.
Defining the most representative hourly upper-air sounding from twice a day upper-
air observations from the routine NWS network is a critical step of the PLANR use
of the UAM. As will be seen, this usually requires interpretation of the meteorologi-
cal conditions of the modeling period.
In Step 2, observational information is added to the Step 1 (u,v) field. Using an
objective analysis procedure, observations are used within a user-specified radius of
influence while the Step 1 (u,v) field dominates in subregions where observations are
unavailable. This procedure consists of four substeps: (1) interpolation, (2) smooth-
ing of the analyzed field, (3) computation of a vertical velocity field, and (4) mini-
mization of the three-dimensional divergence. The following modified inverse-dis-
tance-squared weighting scheme (Ross and Smith, 1986) is used for the interpolation:
(u,v)' = { I k[rk-2(u0,v0)kj + R-^M I krk-2 * R'2}
where (u^v^ denotes an observed wind at station k, rk is the distance from station
k to a given grid point, (u,v)j is the Step 1 wind field, and (u,v)' is the updated wind
vector. The radius R determines whether observations or the Step 1 wind field will
be used.
Following the interpolation, a five-point smoother is applied to the analyzed wind
field to reduce discontinuities that may result from the interpolation. An initial
vertical velocity, W1, is calculated from (u,v)' by integrating the incompressible con-
servation-of-mass equation. It has been noted that vertical velocities obtained from
89092r2 2
17
-------
an objectively anlayzed field may be unrealistically large near the top of the domain
(Godden and Lurmann, 1983). In the DWM, W is modified using a procedure sugges-
ted by O'Brien (1970):
w2(z) = w(z) - (z/ztop)wtop
where Z is the height in terrain-following coordinates and Ztop is the height of the
model top. Note that w^ is zero at the top of the model.
After tha vertical velocity profile is adjusted, it is necessary to adjust the objective
analysis product (u,v)' so that it is mass-consistent with W2« An iterative adjustment
of the horizontal (u,v) field is performed to minimize the three-dimensional diver-
gence within each layer. The adjusted horizontal wind field (u,v)2 is the final pro-
duct of the diagnostic model.
For both the PLANR and RAPS UAM applications to St. Louis the DWM was exer-
cised with 13 to 14 vertical levels using available meteorological measurements.
These wind fields were then vertically averaged to the five vertical layers used in
the UAM applications. The differences in the PLANR and RAPS wind fields were in
the amount and frequency of surface and upper-air data used.
Model Inputs
The following paragraphs describe how the 10 main UAM input files were prepared
for the PLANR application of the UAM to St. Louis.
DIFFBREAK; This file contains the daytime mixing height or nighttime inversion
height for each column of cells at the beginning and end of each hour of the simula-
tion. Hourly mixing heights were estimated at five routine surface meteorological
sites through use of the hourly surface measurements of temperature and the twice
daily upper-air observations from a representative upper-air site using the RAMMET
meteorological preprocessor. The resultant hourly mixing heights at each of the sur-
face sites were then spatially interpolated using the UAM preprocessor program
DFSNBK specifying the 1/r interpolation option.
Due to the presence of a warm front to the east of St. Louis in the afternoon that
separated St. Louis from Salem, IL the Monett MO upper-air observations were used
to define the mixing heights for all of the PLANR evaluation runs. The wind and
temperature observations over Monett were thought to be more representative of the
air over St. Louis despite the fact that the Salem upper-air site is closer to St. Louis
(120 km) than is Monett, MO (340 km). The maximum daily mixing height produced
using the Monett upper-air observations and St. Louis surface observations was
approximately 1,950 m agl. Note that the Salem IL, upper-air sounding produced a
maximum daily mixing height of 2,350 m agl.
89092r2 2
IS
-------
REGIONTOP; This file contains the height of each column of cells at the beginning
and end of each hour of the simulation. If this height is greater than the mixing
height, the cell or cells above the mixing height are assumed to be within an inver-
sion. For the PLANR study of St. Louis a constant 2,000 m agl region top was used.
This value was picked because it is 50 meters above the maximum mixing height.
Thus all five vertical layers of the UAM are contained within the well mixed layer,
offering the maximum vertical resolution possible with the five-layer configuration.
WIND; This file contains the x and y components of the wind velocity for every grid
cell for each hour of the simulation. There are two steps in creating the wind fields
for St. Louis: (1) exercising the Diagnostic Wind Model (DWM) using 14 vertical
levels and data from the six routine surface and a representative upper-air meteoro-
logical observation sites; and (2) vertical interpolation of the 15-layer hourly wind
fields into the five layers used in this application. As noted above, the main differ-
ence between the three diagnostic runs was in the wind inputs into the DWM, namely
the definition of the representative upper-air sounding, and, in run #3 the magnitudes
of the surface wind observations were changed.
METSCALARS; This file contains the hourly values of the meteorological
parameters that do not vary spatially. These scalars are the NO2 photolysis rate
constant, the concentration of water vapor, the temperature gradient above and
below the inversion base, atmospheric pressure, and exposure class. The NO2
photolysis rates were calculated for the CB-IV mechanism using procedures described
by Schere and Demerjian (1977) and actinic flux data collected by Bass and co-
workers (Bass et al., 1980; Gery, Whitten, and Killus, 1988). The concentration of
water was based on measurements of temperature and dew point at the surface
meteorological observation sites. The observed water concentrations at all of the
surface sites were averaged to obtain the hourly input for the UAM. Exposure class
was assigned based on the solar intensity: a value of -2 at night; and daytime values
of either 0 (one hour day/night transition period) to 3. The temperature gradients
below the inversion (TGRADBELOW) and above (TGRADABOVE) were based on the
twice-daily upper-air soundings at Monett, MO.
AIRQUALITY; This file contains the initial concentrations of each species for each
grid cell at the start of the simulation. For the PLANR study of St. Louis we used
the following "clean values" for initial concentrations for all diagnostic simulations:
VOC = 25 ppbc (using EKMA default speciation)
ISOP = 0.001 ppb
NOX = 1 ppb (3/4 NO2, 1/4 NO)
O3 = 40 ppb
CO = 200 ppb
89092r2 2
19
-------
BOUNDARY and TOPCONC; These files define the location of the modeling region
boundaries and specify the concentration of each species that is used as the boundary
condition along each boundary at each vertical level and above the region top. For
the PLANR study of St. Louis, the clean values listed above were used for boundary
conditions.
TEMPERATUR; This file contains the hourly temperature for each surface layer
grid cell. Hourly spatial varying temperatures were obtained by using 1/r interpola-
tion from the surface meteorlogical observations,
EMISSIONS and PTSOURCE; The original area and point source emissions inven-
tories used in UAM/CB-II simulations (Schere and Shreffler, 1982) were converted to
correspond with the CB-IV mechanism. The conversions are described in Table 3-1.
DIAGNOSTIC SIMULATIONS TO ARRIVE AT A BASE CASE
The PLANR use of the UAM limits the number of diagnostic simulations, which are
designed to improve model performance for a particular application to an acceptable
level. Many past UAM applications involved several diagnostic simulations.
Diagnostic simulations usually involve varying uncertain inputs, such as wind fields
and mixing heights, within their ranges of uncertainties. For the PLANR use of the
UAM the number of diagnostic simulations is limited, and the most representative
simulation is used as a base case for modeling analyses of control scenarios.
For St. Louis, three diagnostic simulations were performed before an acceptable base
case was achieved. The three simulations differed in how the observed surface and
upper-air meteorological observations were interpolated and adjusted in the UAM
preprocessor. In the PLANR use of UAM for Atlanta, four diagnostic simulations
were needed (Morris et al., 1989b).
The three diagnostic simulations for the PLANR UAM study of St. Louis differed as
follows:
Diagnostic Run //I: Because of the warm front between St. Louis and Salem on
the afternoon of 13 3uly, the upper-air soundings from Monett, MO were
believed to be more representative of the air over St. Louis than those from
Salem, IL. Thus the Monett upper-air observations were used to define the
mixing heights and domain-mean wind input into the DWM. The twice-daily
upper-air wind observations were linearly interpolated to the hour in question
for input into the DWM. Other meteorological observations (e.g. surface winds
and temperatures) were used with no adjustments.
Diagnostic Run #2; An examination the surface wind observations indicated
that the warm front passes through St. Louis around noon. Thus the Salem
upper-air sounding at 0600 LST was used as input into the DWM until noon; a
89092r2 2
20
-------
TABLE 3-1. Respeciation of emissions of hydrocarbons.
CBM-IV Species As a Function of CBM-II Species
OLE OLE
PAR PAR - ARO * 0.432 - ARO * 2 * 0.568
TOL ARO * O.M32
XYL ARO * 0.568
FORM CARB * 0.288
ALD2 CARB * 0.712
ETH ETH
CRES 1 X 10"6 ppm
MGLY 1 X 10~6 ppm
OPEN 1 X 10"6 ppm
89092 3 21
-------
linear interpolation of the Monett soundings at 0600 and 1800 LSI was used for
the afternoon. Mixing heights were defined from the Monett upper-air sound-
ing and the surface temperature observations.
Diagnostic Run #3; It has been noted in studies in the South Coast Air Basin
that hourly averged wind observations and collocated NWS or FAA instantane-
ous wind observations generally do not agree. During periods of stagnation and
low wind speeds, the hourly surface wind speeds reported by the NWS or FAA
tend to have a positive bias, generally 1 to 2 times the values reported by the
hourly averged monitors. Therefore, in this diagnostic run all FAA surface
wind speed observations were reduced by 50 percent for input into the DWM.
Other inputs remained as in diagnostic run 2.
Diagnostic Run 1
In the first diagnostic run, the meteorological observations were used with no
adjustments in the UAM preprocessors. Observations from Monett, MO were chosen
as the representative upper-air sounding for calculating mixing heights and the
domain-mean wind input for the DWM. Figure 3-5 shows isopleths of the predicted
maximum daily ozone concentrations along with the daily maximum observations.
The predicted region-wide maximum ozone concentration for diagnostic run 1 is 16.2
pphm, compared to the observed maximum of 22.2 pphm, and occurs approximately
20 km north-northeast of the observed maximum. The predicted region-wide maxi-
mum ozone concentration is within 27 percent of the observed maximum when the
two are compared unmatched by location or hour. The predicted maximum ozone
concentration at the location of the observed maximum was half (11.1 pphm) of the
observed peak (22.2 pphm).
An examination of Figure 3-5 reveals two reasons why run //I is not an acceptable
base case: (1) model performance is poor (the predicted region-wide peak ozone is
only within 27 percent of the observed peak, and at the location of the observed peak
the predicted peak is within only 50 percent) and (2) the predicted region-wide peak
ozone occurs too close to the boundary. Reducing VOC emissions in this run would
tend to move the predicted peak ozone concentrations further downwind. It is very
likely the model would locate this peak ozone outside of the modeling domain if the
run 1 inputs were used, resulting in incorrect ozone reductions.
Diagnostic Run 2
As noted previously, a warm front passes through St. Louis around noon on 13 July
1976. Thus the upper-air sounding at Salem, IL is more representative of the air over
St. Louis in the morning, whereas in the afternoon, after the warm front moves to
the east of St. Louis, the upper-air sounding at Monett is more representative. For
89092r2 2
22
-------
Time : 800 - 2000 CST
706 726
NORTH
746
Maximum Value = 16.19
Minimum Value = 2.85
766
SOUTH
- 4316
- 4296
4276
4256
4236
FIGURE 3-5. Isopleths of predicted maximum daily ozone concentrations
(pphm) with superiirposed maximum daily observations for PLANE diagnostic
run #1.
89092
-------
diagnostic run 2 the Salem upper-air winds at 0600 CST were used as input into the
DWM until noon; a linear interpolation between the 0600 and 1800 CST wind observa-
tions from Monett was used for the hours in the afternoon.
The predicted daily maximum ozone concentrations for diagnostic run 2 are shown in
Figure 3-6. Model performance is slightly improved over run 1. The predicted
region-wide maximum ozone concentration is within 25 percent (16.6 pphm) of the
observed peak (22.2 pphm) and occurs approximately 22 km northeast of the location
of the observed peak. At the location of the observed peak run 2 calculates a daily
maximum ozone concentration (matched by location but not hour) to within approxi-
mately 40 percent of the observed value.
Although the results of run 2 are marginally better than run 1, overall model per-
formance is still poor and the predicted peak ozone concentrations still occur too
close to the boundary. Thus run 2 is also considered unacceptable as a base case.
Diagnostic Run 3
The chief problem with runs 1 and 2 is that the wind speeds are too high, which
results in excessive dilution of the ozone peak and formation of the ozone peak too
far downwind. When UAM inputs were prepared for modeling analyses for the South
Coast Air Basin Air Quality Management Plan, differences were noted between wind
speeds reported at colocated hourly average wind monitors and the instantaneous
values reported by Federal Aviation Administration (FAA) sites (Hogo, Mahoney, and
Yocke, 1988). During periods of slow wind velocities, which are typical during eleva-
ted ozone episodes, the hourly average wind speeds were 50 to 75 percent of the wind
speeds reported at the FAA sites. These differences are not surprising because the
primary purpose of the FAA sites is to provide information to pilots on adverse wind
conditions for take-offs and landings. During stagnation and low wind speeds the
FAA wind observer will report the presence of gusts to the pilots; if the hourly
average wind is calm, this information is of no use to the pilots. However, the UAM
requires charaterization of the hourly average wind flows. Thus use of FAA wind
observations in modeling may overstate the surface wind speeds. Further explana-
tion concerning the use of FAA/NWS wind data is provided in Appendix 3 (see also
Morris et al., 1989c).
Thus for diagnostic run 3 the FAA surface wind observations (which included all six
surface sites) were reduced by 50 percent for input to the DWM along with the same
upper-air wind observations used in run 2. Other meteor logical inputs were the same
as in runs 1 and 2.
Figure 3-7 shows the daily maximum ozone concentrations predicted in run 3 as well
as the observed daily maximums. In runs 1 and 2 the UAM predicted a region of ele-
vated ozone concentrations in the northeast portion of the modeling domain, whereas
89092r2 2
24
-------
Time : 800 - 2000 CST
706 726
NORTH
746
Maximum Value = 16 58
Minimum Value = 4.32
766
- 4316
- 4296
- 4276
4256
4236
SOUTH
FIGURE 3-6. Isopleths of predicted naxinum daily ozone concentrations
(pphm) with superimposed nmxiinam daily observations for PLANR diagnostic
run #2.
-------
Time : 800 - 2000 CST
706
20
10
NORTH
726
746
Maximum Value = 19.45
Minimum Value = 4.44
766
T r
1 t t I 1 11 I I ' I
t i :j' r. ! ' t i:
10
SOUTH
4316
4296
4276
4256
4236
FIGURE 3-7. Isopleths of predicted maximum daily ozone concentrations
(pphm) with superimposed maximum daily observations forPLANR diagnostic run #3
(PLANR UAM) .
89092
26
-------
the observations indicate the highest ozone concentrations should be slightly north of
the center of the modeling domain. The lack of any observations in the northeastern
portion of the modeling region does not discount the possibility that elevated ozone
concentrations existed in that region. However, diagnostic run 3 does a much better
job of placing the predicted cloud of elevated ozone concentrations in the upper cen-
ter of the modeling domain (see Figure 3-7).
The predicted region-wide maximum (19.5 pphm) is within 12 percent of the observed
peak (22.2 pphm) and is located approximately 11 km south-southeast from the
observed peak ozone. At the location of the observed peak, run 3 predicts a daily
maximum concentration of 16.7 pphm, about 25 percent of the observed value.
Model performance for run 3 is remarkably better than for runs 1 and 2. Run 3
satisfies the two model performance goals and the predicted peak ozone is located
well within the interior of the region. Thus run 3 is deemed an acceptable base case.
Figure 3-8 compares the predicted and observed ozone concentrations at each moni-
toring site for all three runs. The results from run 3 are in better agreement with
the observations than are the other runs except for sites 108, where run 2 is better,
and site 116, where run 1 is better. The performance of the UAM in the base case
for the PLANR application to St. Louis (diagnostic run 3) is discussed further in the
next section.
89092r2 2
27
-------
30(
O\J
20
^ X
^.
a.
a^
m
O
10
C
A
\J
(
>n
tj\J
20
*-^
s
X
Q.
Q.
^^
1*1
O
10
(f
) 6 12 18 2
i i i i i | i i i i i | i i i i i | i i i i i
- 101
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed Q
-
_ » _
/ \ Q] CD
"~ « * ""
: /W--P :
/ \ »
D-7 \\ \
- / V'1-- Em
. / ^--o- CD
/cd/ \... n
ro^^ ^ d
}-D l-J CO jjf^ -
-I 1 1 Uii'l 1 1 1 I 1 1 1 1 I 1 1 1 1 1 1 UjM
3 6 12 18 2
TIME (HOURS)
5 6 12 18 2
! 1 I 1 1 | 1 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1
- 103
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed Q
-
*»
m +
* --
/_rV- 'a
JF \£m
-fflP^ X D -
_ mm^fjx am _
J-
4
430 30
Ow OW
20 20
s-^
X
Q-^
KJ
O
10 10
Off
l>
.4
3 6 12 18 2
1 1 1 1 1 | 1 1 ! 1 | 1 1 I 1 1 | 1 1 1 1 1
-102
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
- Observed UJ
_ _
-
%
\
/ -. \
- / -.. \ -
\
.-"Z''1*/
J«g] [JJ | ^-litf^T |T]lllllllllllll|T](TjlI!
3 6 12 18 2
TIME (HOURS)
D 6 12 18 2
I I l 1 l | l l l I | l 1 1 1 1 | i l l l l
- 104
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed [D
- -
-
/C'"\
* / '*^v *
/' ^**^s**. *
~ /i/'^ CD ~
»oT] rn i ^i*f I i i i i i I i i i i i I i i i i M
0 6 12 18 1
TIME (HOURS)
4
TA
oU
20
10
430
ou
20
10
.4
FIGURE 3-8. Comparison of hourly predicted and observed ozone concentrations at
each ozone monitoring site for PLANR diagnostic runs 1, 2, and 3.
SYSTEMS APPLICATIONS, INC.
89092
28
-------
30
20
X
CL
IL
^^
r)
O
10
ff
} 6 12 18 2
i i i i i I i i i i i I i t i i i I i i i i i
- 105
PLANR Run 1
PLANR Run 2 - -
PLANR Run 3
Observed D
-
«
,* *»
_* \
_ t» ** __
*" /'\1' %'«
/,/ XX ^
tJ^
"jr
~ /s CD ~
uTi i i rttin 1 i i i i i 1 i i i l i 1 l l i I m
30 30
20 20
X
CL
CL
*
n
O
10 10
[
r\ r\
) 6 12 18 2
i i i i i I i i i i i | i i i l i | i i i i i
- 106
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed [3
-
CD
Q n
CD
-
ul _
_ _
m . n
/ "-., n
VP&S ^ j
n n '-' ^ n i-j .y^^ n-.
T fjr
i 1 i l-J*'! i 1 l l i 1 i l i l l 1 i l i m i
4
30
20
10
r\
0 6 12 18 24 0 6 12 18 24
TIME (HOURS) TIME (HOURS)
30C
20
^
X
CL
CL^
to
O
10
r
L
) 6 12 18 2
i i i i i I i i i i i I i i i i i | i i i i i
- 107
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed D
_ _
m Q
L- 1 m
m
/.
°" X- m
r / N....D -
frS^S^V
$ / N,
"anaC]c] JJ^ n[D -
^j i i L^i*^T i i i i i I i i i i i I i i i m i
*30 3o'
20 20
^^
X
CL
Q^
K1
o
10 10
f\ f\
) 6 12 18 2
! 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1 { 1 1 1 1 1
- 108 '
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed n
«.
.' »
/
/ " -
\
// '^.jga
I ^m\^n
//m \
/P n
" mQ2>5^ ° n -
V .''/' CO ~
rffe'i'tj/i 1 i l i i j 1 i l i i i 1 i i i i LU
4
30
20
10
r\
0 6 12 18 24 0 6 12 18 24
TIME (HOURS) TIME (HOURS)
FIGURE 3-8 Continued.
m
SYSTEMS APPLICATIONS, INC. ^= :
a SE
i
^
89092
-------
0 6 12 18 24 0 6 12 18 24
\J\J
20
x N
i
Q.
Q._
ff)
O
10
C
f\
1 I 1 1 1 | I 1 I 1 I | 1 1 1 1 ! | 1 1 1 I 1
- 109
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
- Observed m
_
_
_ _
mm
_
mmm
_ m
m jf^^a
nj j/]... -\ "»
Q->^" \ m m
Lm m /s/^ Q-
Nii'l'lV'i/i I i i i t i I i i i i i I i i i t i
*j\j *i\j
20 20
5T
i
Q.
£L_
r;
O
10 10
r. ri
I I I I I | I I I I I | I I I 1 I | I 1 I 1 I
-110
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed m
-
~ ~
_ _
^_1
L-.LD
m
mm m -
m ^m
Q ,i^»-«*iC"''%>, n
m t,^!^:"^"' ^sN.[ ^
mm.EkP«tjSi::V''^ ° "
n .- j^^ n-
vroJT! i ^'i I i t i i i I i i i i i t i i i t i
ovy
20
10
r\
0 6 12 18 24" 0 6 12 18 24
TIME (HOURS) TIME (HOURS)
30
20
^
i
Q.
to
O
10
(f_
5 6 12 18 2
! 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1
-111
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed m -
-
" "
_ _
-
_
.. «.,
- ^'^~^-^
/\X^^ ^\^'"
.^itti^^^
^***^^ -i
;₯ m
*jn m i u<*T 1 i i i i i 1 i i i i i 1 i i i m i
30 30
20 20
S"
T
a.
a.
n
0
10 10
rt
3 6 12 18 2
I I ! 1 1 | I 1 1 1 1 | 1 1 I 1 1 | 1 I 1 1 1
-112
PLANR Run 1
PLANR Run 2 -
PLANR Run 3
Observed m -
-
_ _
D
mmrrj
m _
- nj ji..i,(
* \
/ / - \
" /yjC//^sv *"*- D
$'~J/ ^*****. ^
- _ CQi^^ O
-m ^ m m'5^1^ m m-
S^r LJ
4^1 1 l t^fXT i 1 l l i 1 l l i l l 1 1 l l 1 l
4
30
20
10
0 6 12 18 24" "0 6 12 18 24
TIME (HOURS) TIME (HOURS)
FIGURE 3-8 Continued.
SYSTEMS APPLICATIONS, INC.
89092
30
-------
2
I
Q.
O.
ro
O
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed
30
12
18
II r i ) r i i i i | i i i i i i i i i i i
- 114
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CD
20
X
a.
c.
12
TIME (HOURS)
18
24
30
CD
CD
m
a
CJ_
20
10
12
TIME (HOURS)
18
24
30
i l r r
- 115
20
2
Q-
CL
O
10
12
18
2430
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CD
l l I i
30
12
18
24
- 116
20 20
Q.
CL
K3
O
10
12 18
TIME (HOURS)
24
| I I I I I | I I I T
i i i i r
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CD
30
20
10
6 12 18
TIME (HOURS)
24
FIGURE 3-8 Continued.
89092
SYSTEMS APPLICATIONS. INC.
-------
30
12 18
24
- 117
20
c_
CL
10
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CD
rrjO Dmn
30r-r
12
18
- 118
20 20
^
I
a.
£L
o .
10 10
6 12 18
TIME (HOURS)
24
(0
24
in30
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CD
m'ffi'T' i i I i i i i i I i i i i i I i i i i
20
10
6 12 18
TIME (HOURS)
24
30
12
18
24
i- 119
20
I
CL
CL
10
cCuSUOffl.
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed CO
mCQ
u
CDQ
30
12
18
24.
20 20
O
10 10
12
TIME (HOURS)
18
24
i i 1 I i | i I I I I 1 I I 1 I I
- 120
d3*
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed UJ
m
30
20
10
12
TIME (HOURS)
18
24
FIGU3E 3-8 Continued.
89092
SYSTEMS APPLICATIONS. INC.
32
-------
12
18
24,
i i i i i ( i r
h 121
20
i
CL
K)
o
PLANR Run 1
PLANR Run 2
PLANR Run 3
Observed Q
20
10
6 12 18
TIME (HOURS)
24
FIGURE 3-8 Concluded.
SYSTEMS APPLICATIONS. INC.
-------
EVALUATION OF THE PLANR UAM APPLICATION TO ST. LOUIS
In this section we evaluate the PLANR UAM application -c St. Louis by comparing
the model performance of UAM when four types of input uata are used: RAPS,
PLANR, SIMPLE UAM(CB-IV), and UAM(CB-II); and then compare how ozone con-
centrations calculated by the RAPS, PLANR, and SIMPLE UAM respond to emission
control scenarios. The purpose of this inter comparison is twofold: (1) to determine
whether the use of only routinely available data (PLANR UAM) produces acceptable
model performance; (2) to determine the minimum input data needed to obtain ade-
quate UAM performance; to determine how model performance effects emission con-
trol strategies designed to show attainment of the ozone NAAQS; and to analyze dif-
ferent procedures for using the UAM to show attainment.
MODEL INPUTS
Historical UAM(CB-II)
Inputs for the UAM(CB-II) were prepared from the RAPS intensive measurement data
using and the old UAM preprocessor programs (Schere and Shreffler, 1982; Cole et
al., 1983; Ames et al., 1985a,b). The RAPS data base included surface wind mea-
surements at over 20 sites (see Figure 3-4) and hourly upper-air measurements at up
to three sites within the city of St. Louis. The maximum afternoon mixing height
was approximately 1,500 m agl. Initial concentrations and boundary conditions were
interpolated from the dense air quality measurement network. Although the same
intensive data base (RAPS) was used to prepare inputs for both the UAM(CB-II) and
RAPS UAM, there still are slight differences in the meteorological inputs (e.g., wind
fields) because different preprocessors were used.
RAPS UAM
The inputs for the RAPS UAM were based on the RAPS data base. The procedures
for creating the model inputs generally followed the procedures for the PLANR
application (see Section 3) except the RAPS dense surface measurement network and
the RAPS hourly upper-air soundings within St. Louis were used for the wind fields
and mixing heights. The maximum afternoon mixing height was approximately
1,500 m agl. Initial concentrations and boundary conditions were interpolated from
the dense air quality network.
89092r2 2
34
-------
SIMPLE UAM
Meteorological data for St. Louis on 13 3uly 1976 were preprocessed by RAMMET
(CRSTER) and turned into UAM wind files by assuming that the single hourly surface
wind direction represented wind directions aloft and wind speeds increased with
height following the stability-dependent power law wind profile. The mixing heights
were assumed to be spatially constant but varied hourly. The maximum afternoon
mixing height (from Salem, IL) was 2,250 m agl. The same "clean" concentration
values used in the PLANR UAM were used for initial and boundary conditions.
COMPARISON OF PERFORMANCE
We compared the performance of the UAM(CB-II) and the RAPS, PLANR, and
SIMPLE applications of UAM to St. Louis. Isopleths of predicted and observed daily
maximum ozone concentrations from the RAPS and SIMPLE UAM applications are
shown in Figures 4-1 and 4-2 (isopleths for the PLANR application are shown in
Figure 3-7). Table 4-1 presents model performance statistics for the four applica-
tions. (Scatterplots and additional residual analysis plots of model performance sta-
tistics for the RAPS, PLANR, and SIMPLE applications are given in Appendix A.)
Figure 4-3 shows time series plots of predicted and observed ozone concentrations at
each monitoring site for the RAPS, PLANR, and SIMPLE applications.
As seen in Table 4-1 and Appendix A, calculated and observed values are very close
for the RAPS simulation. The predicted region-wide maximum (24.2 pphm) over-
states the observed maximum (22.2) by 9 percent and occurs approximately 9 km
southwest of the observation. The predicted maximum at the location of the peak
observation (21.9 pphm) is within 2 percent and occurs 2 hours later than the obser-
ved value. The RAPS UAM overpredicts the hourly ozone observations for all hours
of the simulation by 10 percent and overpredicts the average of the daily maximum
observations at all sites by 7 percent. The RAPS UAM with CB-IV chemistry
generally produced better agreement with observations than did the RAPS UAM with
CB-II chemistry.
When only routine data are used for model inputs, model performance is somewhat
degraded. The PLANR UAM predicts the hourly and daily maximum observations to
within approximately 20 percent on average. The predicted region-wide maximum
ozone concentration (19.5 pphm) is within 12 percent of the observed maximum and
occurs approximately 12 km south-southeast of the observation. The predicted
maximum ozone at the location of the peak observation is within 25 percent of the
observed peak ozone and occurs at the same hour as the observed peak.
89092r2 2
35
-------
Time : 800 - 2000 CST
706 726
NORTH
7*6
Maximum Value = 24.35
Minimum Value = 7.73
766
SOUTH
- 4316
- 4296
1
- 4276
- 4256
4236
FIGURE 4-1. Isopleths of RAPS UAM predicted maximum daily ozone concentrations
(pphm) with superinposed naxinum daily observations for the St. Louis region
on 13 July 1976. (* denotes location of iraximum concentration value.)
-------
Time : 800 - 2000 CST
706 726
NORTH
746
Maximum Value = 1 1 83
Minimum Value = 2.85
766
SOUTH
FIGURE 4-2. Isopleths of SIMPLE UAM predicted daily iraxLmum ozone
concentrations (pphm) with superimposed daily naxLiTum observations
for the St. Louis region on 13 July 1976.
- 4316
- 4296
- 4276
- 4256
4236
89092
37
-------
TABLE 4-1. Comparison of performance statistics for four modes of application
of the UAM to St. Louis for 13 July 1976.
UAM
Performance Measure (CB-II)
Hourly Ozone Concentrations (matched by
Number of pairs
Average observed (pphm)
Average predicted (pphm)
Bias (pphm)
Average percent difference
Average absolute (gross) error
Gross error percent difference
Correlation coefficient
time and
184*
8.3
7.4
0.9
11*
N/A
N/A
0.95
RAPS
UAM
location)
265
6.8
7.5
-0.7
10%
1.7
25%
0.91
Daily Maximum Ozone Concentration (matched by location but
Number of pairs
Average observed (pphm)
Average predicted (pphm)
Bias (pphm)
Average percent difference
Average absolute (gross) error
Gross error percent difference
Correlation coefficient
Peak Ozone Concentration
Peak observed (pphm)
Unmatched by time or location:
Predicted region-wide maximum (pphm)
Ratio of prediction to observation
Matched by location but not time:
Predicted maximum (pphm)
Ratio of prediction to observation
Hours difference in prediction
to observation
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
22.2
17.4
0.78
16.8
0.76
N/A
14
15.0
16.0
-1.0
7%
2.4
16%
0.68
22.2
24.2
1.09
21.9
0.99
+2
PLANR
UAM
265
6.8
5.4
1.4
21%
2.02
30%
0.90
not time)
14
15.0
12.3
2.7
18%
3.9
26%
0.55
22.2
19.5
0.88
16.7
0.75
0
SIMPLE
UAM
265
6.8
4.3
2.5
37%
2.9
43%
0.79
14
15.0
8.2
6.8
45%
6.8
45%
0.20
22.2
11.8
0.53
10.0
0.45
-2
* Due to differences in sample sizes between the historical UAM(CB-II) and the
current UAM performance statistics, statistics for non-peak ozone results
cannot be directly compared.
89092rl 3 ,0
-------
30
12
18
24
20
x
Q.
a
to
o
I I I T
101
i i i i i i i i i i i i i i r
Simple -
PLANR
Observed
30 30
12
18
24
1 I I I I I I I I I I I I I I I I I I I I I I
-102
Simple
20 20
a.
£L
10 10
12
TIME (HOURS)
18
PLANR
.Observed
/ \
/ .-
30
20
10
6 12 18
TIME (HOURS)
24
X
0.
0.
30
20
10
rT
) 6 12 18 2
i i i i i I i i i i i | i i i i i I i i i i i
- 103
Simple
RAPS -
PLANR
Observed d -
-
i-4°
/m \
/ ?.
/-£ ° '-'\
/ C?"x~-
- -^^s mm _
5-m m i L»'i I i i i i i I i i i i i I i i i i m
430 30
20 20
S
0.
Q^
K3
O
10 10
n rC
) 6 12 18 2
i I i I i | I I i I 1 | I I i I i | i 1 i I i
- 104
Simple
RAPS -
PLANR
Observed tTJ
- -
-
/ . \
* JT\ *
* J' \ %
* / \ *
- s^S^ ^- I
=-_ ^^ / Q] -
ijTi m i (.'f 1 i i i i i 1 i i i i i 1 t i : i 9
4
30
20
10
n
6 12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
24
FIGURE 4-3. Comparison of predicted and observed hourly ozone concentrations
(pphm) at each monitoring site for the RAPS, PLANR, and SIMPLE UAM.
SYSTEMS APPLICATIONS. INC.
89092
39
-------
30*
20
I
0-
CL
O
10
*,
5 6 12 18 2
I I I I I I I I I I I I I I I I I I I I I I I
- 105
Simple
RAPS -
PLANR
Observed Q
-
/ ,
/ "'"V
/./
J^T"! (UQ I I I I I I I I I I I I I I I I 1 [T]
) 6 12 18 2
4
30 30
20 20
S
0.
K3
O
10 10
[
0 0
4 (
3 6 12 18 2
i l i i i | i i i I i | i l i i i | i i i l i
-106
Simple
RAPS -
PLANR
Observed D
-
a
9° °
I ,--E" "'"-.? J
/v \nmn i
'" Q» xv ^
^.».f...^....^X^»'| i i i i i 1 i i i i i 1 i i i m i
3 6 12 18 2
4
30
20
10
0
4
TIME (HOURS) TIME (HOURS)
C
20
S
I
o.
Q.
O
10
[
°(
) 6 12 18 2
i l i l i | i l l i i | i i i i i | i i l i i
- 107
Simple
RAPS -
PLANR
- Observed UJ
.-" *"-.
-? rri \
- /A °\ I
/ El* *\
/ ^xx "'^
- /i/ ^^o
rn m CD CD rpft??SS
^\l ^ Y\-*' 1 1 1 1 1 1 1 1 1 ! 1 1 1 1 1 1 (T| 1
5 6 12 18 2
4 (
30 30
20 20
S
0-
o^
O
10 10
0 0
4 (
) 6 12 18 2
i i i l i | i i i l i | i l i l l | i i i l i
-108 J
Simple
RAPS -i
PLANR
Observed D -
- / \ -
/ ! \
* - "^
«
* :
/_/ nmm\
^ -*" ^1 "=
..._ .-* ^^ ^^ _^
/JP ^~~\
/JsP \mCD
^I^ffTrnS^^ Q
r>., .»' CD
r^'Qi^' l i 1 i l i l i 1 i l i l i 1 i l i i CD
3 6 12 18 2
4
30
20
10
0
4
TIME (HOURS) TIME (HOURS)
FIGURE 4-3 Continued.
§H
SYSTEMS APPLICATIONS, INC. ^. =
&OE
*
FS
-------
30
12
18
24
1 I I I I I I I \ f I I I I ] I \ IT I I I
I- 109
Simple
RAPS
PLANR
Observed Q
20
x
0.
Q-
.con
30 30
12
18
24
20 20
CL
£L
10 10
12
TIME (HOURS)
18
24
1 I 1 I I I I I T
- 110
Simple
RAPS
PLANR
Observed ED
.-.m
Eld" "-EK
..'d Da
a x^< a
m
30
20
10
6 12 18
TIME (HOURS)
24
S
I
a.
0.
30
20
10
rf
) 6 12 18 2
i I i I i | I I i I I | I I i I i | i 1 i I i
-111
Simple
RAPS -
PLANR
Observed Q -
-
-
-
~~ .'" " ~~
L / S X
\
/xlXV^'
jS>*^~^**^'
^~~~^^^^ / CD"
?jn m i \.*{ I i i i i i I i i i i i I i i i m i
430 30 '
20 20
s"
^.
a.
a.
o
10 10
o A
> 6 12 18 2
I I I I i | I I i I i { i i i I I | I I i I I
-112
Simple
RAPS -
PLANR
- Observed CD
-
~ ,P.aDm ~_
/m/.....>""'-,n
/ / \""m
~ / \ ~
//n&^~^"^ °°
~ mm ^£Hy^ ^^" O
-D p^j5- QJ D-
^"""[""Y^-i''1 i i i i i 1 i i i i i 1 i i i i i
4
30
20
10
o
12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
24
FIGURE 4-3 Continued.
89092
SYSTEMS APPLICATIONS. INC.
41
-------
12
18
i r iii|iiiii|iiiir|iirii
h 113
Simple
RAP
PLANR
Observed
20
2430 30°
12
i i i i i | i i i i i | i i i i i | i
- 1 14
Simple
RAP
PLANR
Observed
12
TIME (HOURS)
18
12 18
TIME (HOURS)
24
30
12
18
24
- 115
20
I
a.
a.
to
o
10
Simple
RAPS
PLANR
Observed El
30
12
20 20 -
6 12 18
TIME (HOURS)
24
Simple
RAP
PLANR
Observed CD
l i i l i li i i i i I i i i i i I i l i l l
10 10 -
- 10
6 12 18
TIME (HOURS)
24
FIGUEE 4-3 Continued.
SYSTEMS APPLICATIONS, INC.
89092
42
-------
12
18
I I I I \ I I I I I | I I I I I I I I I I I
h 117
Simple
RAPS
PLANR
Observed CD
20
S
I
Q.
r")
O
24 0
30 30
12
18
24
20 20
2
I
CL
Q.
10 10
12
TIME (HOURS)
18
24
o eta CD m"
I T 1 I \ 1 T I I I | I I I I I | I I I I I
- 118
Simple
RAPS
PLANR
Observed CD
*».
i rn'm'T' i i I i i i i i I i i
I I I I I I I
20
10
12 18
TIME (HOURS)
24
12
18
24
- 119
20
n
O
10
Simple
PLANR
Observed
CO
0
12
18
24
20 20
I
a.
Q.
10 10
1 I I I I | i I I 1 I | I I I I I I I I I I T
- 120
Simple
RAPS
PLANR
Observed Q
(ffiSliSlSI-l-'f' I i I I I ' I ' I I I I I ' ' ' I I I o (f&
0 6 12 18 24 0
TIME (HOURS)
20
10
6 12 18
TIME (HOURS)
24
FIGURE 4-3 Continued.
STSTEMS APPLICATIONS, INC.
89092
43
-------
30
12
18
24
1 I I I I I I I I I ] I I I I I I 1 I I 1 T
- 121
Simple
PAP5
PLANR
Observed d
20
0-
0-
ro
O
12
TIME (HOURS)
18
20
10
24
FIGURE 4-3 Concluded.
89092
44
SYSTEMS APPLICATIONS. INC.
-------
Model performance is further degraded in the SIMPLE UAM simulation. The predic-
ted region-wide maximum ozone concentration (11.8 pphm) is barely within 50 per-
cent of the maximum observation and occurs approximately 14 km southeast of the
observation. The model tends to underpredict most of the high observations; the
daily maximum ozone observations are underpredicted by 45 percent on average.
The performance of the UAM(CB-II) simulation is similar to that of the PLANR
UAM. The observed peak (22.2 pphm) is reproduced better in the PLANR application
(19.5 pphm) than in the UAM(CB-II) application (17.4 pphm). However, at the loca-
tion of the observed peak the UAM(CR-II) and PLANR simulations predict almost
exactly the same daily maximum ozone value; both are within 25 percent of the
observation. Note that the sample sizes for predicted and observed hourly ozone
concentrations are different in the UAM(CB-II) (N = 184) and the RAPS, PLANR, and
SIMPLE UAM (N = 265). Comparison of performance statistics for hourly ozone con-
centrations between the UAM(CB-II) and UAM(CB-IV) simulations is complicated by
this difference.
In summary, adequate model performance was obtained with the UAM(CB-IV) using
only routinely available meteorological and air quality data as input. Based on the
peak ozone performance measures, the PLANR performance statisfied the model
performance goals and produced results that were more accurate that those obtained
with the historical UAM(CB-II), which used intensively measured data. However, the
PLANR use of the UAM should be evaluated for different cities before definitive
conclusions can be drawn concerning the use of the UAM with limited data.
DISCUSSION
Differences in model performance between the RAPS, PLANR, and SIMPLE applica-
tions of UAM (CB-IV) and the UAM(CB-II) can be attributed primarily to differences
in meteorology (winds and mixing heights), initial and boundary conditions, and, in
the case of the UAM(CB-II), chemistry. The PLANR UAM predicts a lower peak
ozone than the RAPS UAM, primarily because the RAPS UAM uses higher boundary
conditions, but also because PLANR UAM uses higher mixing heights and wind speeds
(i.e., more dilution). Even though the UAM(CB-II) also used the higher boundary con-
ditions, the chemical mechanism (CB-II) is less reactive than the one used in the
PLANR UAM application (CB-IV), which might explain why the PLANR UAM pre-
dicts a higher peak than the UAM(CB-II). The SIMPLE UAM predicts the lowest peak
because it has low boundary conditions (same as PLANR UAM) and the highest mix-
ing height and wind speeds.
The emission inputs were based on the 1976 NEDS emissions data converted to
UAM(CB-II) formats as part of the St. Louis Ozone Study. In the UAM(CB-IV) simu-
lations (RAPS, PLANR, and SIMPLE) the chemical speciation of the emissions may
be improperly characterized since the VOC emissions in CB-IV were estimated from
89092r2 2
45
-------
their CB-II counterparts rather than from the original compositions of individual
species. It is unclear whether such improper characterization results in an over- or
underestimate of the emissions reactivity in the UAM(CB-IV) simulations.
Since the development of the 1976 emission inventory for UAM modeling, many new
VOC emissions sources have been recognized. These emissions include natural emis-
sions, motor vehicle running loss emissions, and previously unaccounted evaporative
sources, such as solvents, paints, and coatings. The underprediction of the observed
peak ozone concentrations by the PLANR UAM and UAM(CB-ll) could be due entirely
to the absence of these emissions. Conversely, the seemingly good agreement
between observed ozone concentrations and concentrations predicted by the RAPS
UAM may be because other UAM inputs were incorrectly specified, thus offsetting
the understatement of the emission inventory. The most likely UAM inputs in the
RAPS UAM that may be offsetting the understatement of the emissions include
overestimates of the initial and boundary conditions and overestimates of the reac-
tivity of the emissions.
The RAPS UAM initial and boundary conditions were based on the dense surface air
quality network from the RAPS data base, whereas the PLANR and SIMPLE UAM
used "clean" values (25 ppbc VOC and 1 ppb NOX). The use of surface air quality
data to characterize concentrations aloft invariably results in overestimates of the
initial pollutant mass and the pollutant mass flux entering the region. If the initial
or boundary concentrations in the RAPS UAM are overspecifled, then the model will
estimate higher emission control requirements to meet attainment than are actually
needed. Since the RAPS UAM does not tend to systematically overpredict the obser-
ved concentrations at ozone sites upwind of the urban core (see Figure 4-3), it is
more likely that the RAPS initial concentrations have been overstated.
This discussion is speculative at best, but it does illustrate that good model perform-
ance does not always indicate that the model is operating correctly. Because of the
underestimation of emissions in the 1976 emission inventory we would expect the
UAM to underpredict the peak observed ozone concentration. Thus the seemingly
good performance of the RAPS UAM is questionable. This exercise does illustrate
two important points. First, adequate UAM model performance evaluation is a
necessary but not a sufficient condition for defining a proper UAM base case. An
evaluation of whether the model is getting the right answer for the right reason is
also necessary. Second, an accurate emissions inventory is essential for using any
photochemical model to calculate emissions control strategies to demonstrate
attainment of the ozone NAAQS.
CORRECTIONS FOR MODEL BIAS IN CALCULATIONS OF OZONE REDUCTIONS
IN RESPONSE TO EMISSION CONTROL STRATEGIES
Although the performance of the PLANR UAM application to St. Louis is adequate
using just routine data, it is unclear at this time how the UAM will be used to
89092r2 2
46
-------
evaluate alternative emission control strategies when the observed ozone peak is not
well replicated. Model bias (either over- or underprediction of the observed peak)
needs to be accounted for when using a model to demonstrate attainment of the
ozone NAAQS. This report is not the forum to propose a policy for the use of the
UAM to demonstrate NAAQS attainment. However, the recent development of
several sets of UAM inputs (RAPS, PLANR, and SIMPLE UAM) allows us to compare
several methodologies for accounting for model bias and what effects data availa-
bility will have on the calculated emission reduction requirements.
When using just routine data and a limited number of diagnostic simulations (P.-.
UAM) we cannot expect to replicate the peak ozone observations, thus some measure
of accounting for the model bias needs to be incorporated into the PLANR proce-
dures for demonstrating ozone attainment.
We have initially identified three methodologies for using the UAM to demonstrate
attainment of the ozone standard, two of which will correct for model bias:
(1) Uncorrected bias approach. The model predictions are used as they are. That
is, regardless of the difference between the observed peak and the predicted
region-wide maximum in the base case, an emission control scenario is con-
sidered to demonstrate attainment of the ozone NAAQS when the predicted
region-wide maximum in response to emission reductions is reduced to the
NAAQS level (12 pphm).
In the next two methods the difference between the predicted region-wide maximum
ozone concentrations in the base case and in an emission reduction scenario is com-
pared with the difference between the observed peak and the ozone NAAQS.
(2) The decrement approach. The emission reduction needed to reduce the predic-
ted regional maximum ozone to the NAAQS level is the same incremental
reduction as that needed to reduce the observed peak. For example, if the
observed and predicted peaks are respectively 22 and 20 pphm, an emission
control strategy demonstrates attainment of the ozone NAAQS when the pre-
dicted region-wide maximum is reduced by 10 pphm (i.e., the increment from
the observed peak of 22 pphm to the ozone NAAQS, 12 pphm).
(3) The percentage adjustment approach. The emission reduction needed to reduce
the predicted peak ozone is the same percentage reduction needed to reduce
the observed peak. For the example of observed and predicted peaks of 22 and
20 pphm, an emission control scenario demonstrates attainment when it
reduces the predicted region-wide maximum by 45 percent, i.e., to 10.9 pphm,
the same percentage required to reduce the observed peak to 12.0 pphm.
Variants of these approaches include taking into account an irreducible amount of
background ozone, which is unaffected by anthropogenic emission reductions, and
89092r2 2
47
-------
using the maximum predicted ozone concentration at the location of the observed
peak instead of the region-wide predicted maximum ozone. In the following sections
we demonstrate how each of these approaches would work for the RAPS, PLANR,
and SIMPLE UAM applications with across-the-board VOC emission reductions.
For this illustrative example we have neglected any effects of emission changes to
NOX, which will greatly effect the control strategies.
Three VOC emission scenarios were simulated by the RAPS, PLANR, and SIMPLE
UAM: the base case and a 60 percent and 80 percent reduction in VOC emissions.
Table 4-2 shows the predicted region-wide maximum ozone concentrations for the
different emission scenarios. In the following sections we discuss how to estimate
the VOC emission reduction required to reduce the observed peak ozone to 0.12 ppm
using the methodologies discussed above.
Uncorrected Bias
Figure 4-4 shows the predicted peak ozone reductions for different VOC emission
reduction scenarios as calculated in the RAPS, PLANR, and SIMPLE UAM applica-
tions. If the model predictions were used as is, the SIMPLE UAM estimates that no
emission reduction is required to reduce the peak ozone concentrations to 12 pphm
because of its gross underprediction of the observed peak. The PLANR UAM esti-
mates that a 55 percent VOC emission reduction is required, and the RAPS UAM
estimates that a 91 percent VOC emission reduction is required.
The difficulty in interpreting the modeling results is complicated by the fact that a
region is considered in nonattainment of the ozone NAAQS when the fourth highest
measured daily maximum ozone concentration in three years for a region exceeds
12.4 pphm. The RAPS UAM does a very good job of predicting the peak observed
ozone concentration at the location of the observed peak (within 2 percent) but it
predicts a higher region-wide maximum than the peak observation. It is impossible
to determine whether such a higher ozone concentration actually existed in St. Louis
on 13 3uly 1976 because no ozone monitors existed at the location of the predicted
peak.
The EPA recommends that emission control requirements should not be based on
ozone predictions constrained to a particular monitoring site (Layland and Cole,
1983). However, if the RAPS UAM predicted region-wide maximum ozone concen-
tration is an overprediction of actual ozone concentrations for this day, then the cal-
culated VOC emission reductions needed to eliminate exceedances of the ozone
NAAQS may be overstated. If VOC emission reductions are based on the ozone con-
centration predicted by the RAPS UAM at the location of the highest observed
ozone, a 78 percent reduction would be required to eliminate the ozone exceedance
on this day, in contrast to the 91 percent reduction based on the region-wide maxi-
mum ozone prediction.
89092r2 2
48
-------
TABLE 4-2. Predicted regional maximum
ozone concentrations for the RAPS,
PLANR, and SIMPLE UAM applications
for different VOC emission reduction
scenarios.
Percent
VOC
Emission
Reductions
Predicted Regional Maxi-
mum Ozone Concentration
(pphm)
RAPS
PLANR
SIMPLE
0
60
80
24.35 19.45 11.83
17.73 11.80 8.91
14.27 8.48 7.17
89092rl 3
49
-------
10
20
30 40 50 60 70
Percent Reduction in VOC Emissions
80 90 100
FIGURE 4-4. Region-wide maximum ozone concentrations calculated by the RAPS,
PLANR, and SIMPLE UAM for three scenarios in St. Louis, 13 July 1976 (0, 60,
and 80 percent reductions in VOC emissions).
EEE89092
50
-------
Decrement Approach
In the decrement approach a VOC emission reduction scenario is considered to
demonstrate attainment when the predicted region-wide maximum ozone concentra-
tion is reduced by the amount needed to reduce the peak observed ozone to the
NAAQS level. For the St. Louis episode on 13 July 1976 this would mean that the
peak observed vc.'ue (22.2 pphm) should be reduced by 9.8 pphm to arrive at 12.4
pphm. Figure 4-5 shows the VOC emission reductions required to meet attainment of
the ozone NAAQS when the decrement approach is applied to the RAPS, PLANR, and
SIMPLE UAM results. If the RAPS simulation results are used, a 74 percent reduc-
tion in VOC emissions is required to reduce the peak ozone concentration to 12.4
pphm; a 71 percent reduction is required if the PLANR simulation results are used.
Because of the gross underprediction of the peak ozone concentrations in the SIMPLE
UAM results for St. Louis, the decrement approach cannot be used with these results.
Percentage Approach
In the percentage approach a VOC emission reduction scenario is considered to
demonstrate attainment when the predicted region-wide maximum ozone concentra-
tion is reduced by the percentage required to reduce the observed peak ozone con-
centration to the NAAQS level. For the St. Louis episode on 13 July 1976 a 44 per-
cent reduction in the peak ozone is required to reduce the observed ozone concentra-
tion to 12.4 pphm. Figure 4-6a shows the VOC emission reductions needed for ozone
attainment when the percentage approach is applied to the RAPS, PLANR, and
SIMPLE UAM results (83, 66, and 90 percent respectively).
One potential problem with the percentage approach is the failure to account for
background ozone. Both the decrement and percentage methods are complicated by
background ozone assumptions. As the predicted ozone approaches background
levels, the sensitivity of the predicted ozone concentrations to emission controls is
reduced. Thus when the model underpredicts the peak observed ozone concentration,
the decrement and percentage methods will overstate.
SUMMARY
Table 4-3 summarizes the percentage VOC emission reductions required to reduce
the peak observed ozone concentrations to below the ozone NAAQS for several
methods using the RAPS, PLANR, and SIMPLE UAM. When model calculations are
used without adjustment (uncorrected bias), the emission reduction requirements are
understated if the model underpredicts the observed peak ozone (i.e., PLANR and
SIMPLE UAM) and overstated when the model overpredicts the peak ozone (RAPS
UAM). Thus it is appears that when the model underpredicts, the predicted ozone
89092r2 2
51
-------
15
10
Attainment
PLANRUAM
X
O'RAPS UAM
SIMPLE UAM
20
30 40 50 60 70
Percent Reduction in VOC Emissions
80
90
100
FIGURE 4-5. Use of the decrement approach to correct predicted region-wide
maximum ozone concentrations to demonstrate attainment of the NAAQS.
The predictions are from the RAPS, PLANR, and SIMPLE UAM applications
to three scenarios in St. Louis (see Figure 4-4). The attainment line is the
reduction needed to reduce the observed peak ozone to below the NAAQS.
EEE89092
52
-------
/PLANRUAM .,.,
/ ^9 ..''''"
20
30 40 50 60 70
Percent Reduction in VOC Emissions
80
90
100
FIGURE 4-6. Use of the percentage approach to correct predicted region-wide
maximum ozone concentrations to demonstrate attainment of the NAAQS.
The predictions are from the RAPS, PLANR, and SIMPLE UAM applications to
three scenarios in St. Louis (see Figure 4-4). The attainment line is the reduction
needed to reduce the observed peak to below the NAAQS.
EEE89092
53
-------
TABLE 4-3. Estimated reduction in VOC emissions
required to meet attainment of the ozone NAAQS.
Different estimation methods were applied to the
RAPS, PLANE, and SIMPLE UAM results.
Percent VOC Emission
Reduction Requirement
Method
Uncorrected bias
approach
Decrement approach
Percentage approach
Percentage approach
with background
RAPS
91a/78b
78
83
82
PLANR
55a
73
66
64
SIMPLE
Oa
>100
90
74
a Using predicted region-wide ozone concentrations.
Using predicted ozone concentrations at the
location of the observed peak.
89092rl 3
54
-------
concentrations should not be used as is to demonstrate attainment of the ozone
standard. In contrast, when the model overpredicts, the model calculations can be
used as is to obtain a conservative estimate of the necessary emission reduction.
When model bias is accounted for (using the decrement or percentage method), the
PLANR and RAPS UAM results produce quite similar estimates of the VOC emission
reductions. The estimates vary by less than 10 percent. However, the SIMPLE UAM
produces estimates that are quite variable and not consistent with each other. This
is not desirable and indicates that when the UAM performs extremely poorly, as in
the SIMPLE UAM, then the model should not be used to demonstrate attainment of
the ozone NAAQS.
Comparison of the decrement and percentage methods of estimating VOC emission
reductions reveals that, for the St. Louis episode on 13 July 1976, the RAPS UAM
leads to 5 to 18 percent more reductions than does the PLANR UAM.
89092r2 2
-------
5 DEVELOPMENT OF A PLANR UAM BASE CASE IN PHILADELPHIA
Because we wished to compare the PLANR UAM modeling results for Philadelphia
with those from certain previous studies of this region (Haney and Braverman 1985;
Haney and Burton, 1988), we used the same modeling domain (Figures 5-1 and 5-2)
and modeling period (0000 to 2000 on 13 July 1979) used in the earlier studies. Most
of the remaining PLANR UAM modeling inputs were quite different because the
PLANR UAM inputs were developed using just routine meteorological and air quality
data.
METEOROLOGICAL CONDITIONS
Our characterization of meteorological conditions in Philadelphia on 13 July 1979
was based on daily weather maps, routine surface meteorological data in the region
(Table 5-1), routine twice-daily upper-air observations from John F. Kennedy (JFK)
International Airport (New York City) and Dulles International Airport (Washington,
DC), and ozone concentrations from the routine air quality modeling network (Table
5-2). The winds on 12 and 13 July were very light due to the presence of a high-
pressure system located off of the coast of Florida (i.e., a Bermuda high). As seen in
the daily weather map (Figure 5-3) the high-pressure system dominated the surface
and aloft wind fields. On 13 July the high-pressure system weakened during the
course of the day, due to the movement of a trough that resulted from remnants of
Hurricane Bob, and moved eastward into western Pennsylvania. It is postulated that
the high observed ozone concentrations on 13 July were due to stagnant conditions on
the 12th and the morning of the 13th.
On the afternoon and evening of 12 July surface winds were light and mainly from
the west-northwest. By early morning, on 13 July most of the surface winds were
calm, with some wind monitors recording very light winds from the north. The winds
remained fairly light during the early morning of 13 July and then became southerly
by 1000 EST. Under this flow regime it is assumed that pollutants from Phildelphia
traveled a short distance south during 12 July and early morning on the 13th, and
then north across the city after the wind shift at 1000. Such recirculation of the 12
July emissions on 13 July is most likely the cause of the high ozone concentrations
recorded on 13 July.
The presence of high amounts of ozone and ozone precursors in Philadelphia on the
morning of 13 July is also evident in the routine ozone monitoring data. Observed
89092r2 5
56
-------
PHILRDELPHIR
R1RSHED
FIGURE 5-1. Geographical location of the Philadelphia airshed nodeling
region. (Source: Haney and Bravernan, 1985)
63033r
89092
57
-------
NORTH
20
30
tr>
10
I ' iiiiiii<t
10 20 30 40 SB
KILOMETERS
i i i i
tJ
-------
TABLE 5-1. Routine surface meteorological sites in the vicinity of
Philadelphia with data available on 13 July 1977.
Site
ID
NEAP
PIAP
MCGR
WILL
LAKH
THEN
MILL
WILM
CHES
BRIS
ALGH
UTM Zone 18
Site Name
Northeast Philadelphia Airport, PA
Philadelphia International Airport, PA
McGuire Air Force Base, NJ
Willow Grove Naval Air Station, NJ
Lakehurst Naval Air Station, NJ
Trenton Mercer-County Airport, NJ
Millville Airport, NJ
Greater Wilmington Airport, DL
Chester, PA
Bristol, PA
Alleghany , PA
UTMX
499.0
478.6
534.1
488.7
556.9
515.6
494.3
448.6
467.8
510.0
499.0
UTMY
4436.0
4414.6
4431.3
4449.8
4431.5
4459.0
4357.0
4390.7
4409.3
4439.5
4425.2
UAM Modeling
Domain
X
22.4
18.3
29.4
20.3
34.0
25.7
21.5
12.3
16.2
24.6
22.4
y
19.2
14.9
18.3
22.0
18.3
23.8
3.4
10.1
13.9
19.9
17.0
TABLE 5-2. Routine air quality observation in operation on 13 July 1979
in the vicinity of Philadelphia.
Site
ID
AMS
CAMD
CHES
CLAY
FRAN
sour
BRIS
NEAP
ANCO
Site Number
in Figure 5-2
1
5
6
7
11
18
3
12
2
Location
UTM Zone 18
(km)
Site Name
Air Management Services Laboratory, PA
Camden, NJ
Chester, PA
Claymont, DL
Franklin Institute, PA
South Broad and Spruce Streets, PA
Bristol, PA
Northeast Airport, PA
Ancora, NJ
UTMX
491.6
491.7
469.0
461.5
485.2
486.1
510.0
499.0
511.8
UTMY
4428.5
4419.0
4410.0
4406.4
4422.8
4421.6
4439.0
4436.0
4392.4
89092rl 6
-------
FIGURE 5-3. Synoptic surface weather nap at 0700 EST, July 13, 1979.
(Source: Allardetal., 1981)
B3033r
89092
60
-------
ozone concentrations at two of the routine sites upwind (north) of the Philadelphia
urban core were quite high at 1000 EST on 13 July 1979. Observed ozone concentra-
tions at 1000 EST exceeded 12 pphm at Ancora, New Jersey and exceeded 8 pphm at
Bristol, Pennsylvania. The presence of high surface ozone at such an early time of
the day indicates that a reservoir of elevated ozone concentration exists aloft and is
mixed down as the mixing height rises. Because of these high morning measurements
upwind of Philadelphia and the knowledge that all cities in the northeast corridor are
subjected to transport of ozone and ozone precursors, these elevated pollutants con-
centrations must be accounted for in the PLANR UAM initial and boundary condition
inputs.
PREPARATION OF INPUTS
The procedures used to prepare the model inputs followed the PLANR procedures
discussed by Morris and others (1989c). Because there is evidence that pollutants
were transported into the region and that emissions from 12 July in the Philadelphia
region influence ozone formation on 13 July, a large modeling region should be used
(to minimize the effects of boundary conditions) and 12 July should be simulated as
well (to eliminate the need for initial concentrations). However, because one of the
main objectives of this exercise was the comparison of the PLANR UAM with the
results from a past study, it was felt that the modeling domain and simulation period
should remain the same to simplify the comparison. Differences between the
PLANR UAM applications to Philadelphia and St. Louis were as follows: (1) because
of the transport of pollutants into the Philadelphia region, initial concentrations
(AIRQUAL) and lateral boundary conditions (BOUNDARY) for Philadelphia were
based on interpolation of the routine surface air quality observations whereas "clean"
values were used in St. Louis; and (2) the emissions data were directly speciated into
CB-IV species based on source-specific process codes rather than estimated from the
CB-II species, as was done in St. Louis. In the following sections we discuss the
preparation of meteorological, air quality, and emissions inputs for the PLANR
application of the UAM to Philadelphia.
Data Availability
The routine meteorological surface observations sites in operation on 13 July 1979 in
the vicinity of Philadelphia are listed in Table 5-1. The locations of these routine
sites are shown in Figure 5-2. No routine upper-air observations were available in
the immediate vicinity of Philadelphia. Thus we relied on upper-air observations
from two sites: (1) Dulles International Airport, approximately 225 km west-south-
west of Philadelphia, and (2) John F. Kennedy (JFK) International Airport, approxi-
mately 145 km east-northeast of Philadelphia.
Routine surface air quality observation sites consisted of the nine sites within the
Philadelphia UAM modeling domain listed in Table 5-2. All nine sites measured
89092r2 5
61
-------
hourly average ozone concentrations on 13 July 1979. In addition, three of the sites
also measured NO, NO2» and CO (AMS Laboratory, Camden, and South Broad), and
two sites also measured CO (Ancora and Camden).
Although data from only the nine routine air quality monitoring sites were used in
the development of the PLANR UAM inputs, 21 air quality observation sites (see
Figure 5-2), including 12 from the Philadelphia Oxidant Study (POS) were used in the
evaluation of the PLANR UAM.
Model Inputs
DIFFBREAK; Mixing heights in the Philadelphia area were estimated using upper-
level temperature soundings at Dulles and JFK airports and hourly surface tempera-
ture from the Philadelphia routine surface meteorological monitoring network (Table
5-1) using the method described by Kelley (1981). Calculated maximum mixing
heights were 1775 m AGL based on the Dulles Airport soundings and 1881 m AGL
based on the JFK Airport soundings. Because JFK Airport is on the coast and Phila-
delphia and Dulles Airport are further inland, the Dulles mixing height was felt to be
more representative of conditions in Philadelphia. The nighttime DIFFBREAK was
assumed to be 250 m AGL and the diurnal variation in the mixing height was calcula-
ted using the Dulles Airport sounding and the hourly varying surface temperatures.
The spatially varying mixing height field was then obtained by using the 1/r interpo-
lation in the UAM preprocessor DFSNBK.
REGIONTOP; The region top for the PLANR application of the UAM to Philadelphia
was defined as 1850 m AGL. This value is 75 m above the maximum mixing height;
thus all five vertical layers are assumed to lie within the mixing layer during the
period of the maximum mixing height.
WIND; The Diagnostic Wind Model (DWM) was configured for 14 layers. Surface
wind data were taken from the 11 routine sites and upper-air wind soundings from
Dulles and JFK airports. Because Philadelphia was approximately between the two
upper-air sites, the hourly domain-mean vertical wind sounding needed as input to
the DWM was calculated by averaging the soundings from Dulles and JFK and using
linear interpolation between the average morning and evening soundings to the hour
of interest.
Two wind fields were prepared for the Philadelphia region, one each for diagnostic
runs 1 and 2. In diagnostic run 1 the wind observations were used without adjust-
ments. For diagnostic run 2 the surface wind speeds at the instantaneous FAA/NWS
sites (all six surface sites) were used but reduced by 50 percent. The two wind fields
generated for the diagnostic runs are shown in Appendixes B and E, respectively.
89092r2 5
62
-------
METSCALERS; The NG>2 photolysis rates (RADFACTOR) were calculated from the
solar zenith angle and assuming clear skies; the procedures described by Schere and
Demerjian (1977) and actinic flux data collected by Bass and co-workers (1980) were
used. Exposure class was also estimated from the amount of solar radiation, assum-
ing clear skies. Water vapor concenrations were calculated from the hourly dew
point measurements recorded at Philadelphia International Airport. Water vapor
concentrations were approximately 17,000 ppm in the morning of 13 July 1979. At
around 0600 EST water vapor concentrations started rising and peaked at approxi-
mately 27,000 ppm at 1300 EST. By 2000 EST the concentration of water vapor was
about 20,000 ppm.
The values for temperature gradients below the diffusion break (TGRADBELOW) and
above (TGRADABOVE) were estimated from the JFK Airport temperature soundings
(see Figure 5-4) and the calculated mixing heights. The values used for metscalers
are given in Table 5-3.
AIRQUALITY; The initial conditions at 0000 on 13 July 1979 for the first two layers
of the UAM, which lie below the diffusion break, were estimated by using 1 /r inter-
polation of the measurements (i.e., O?, NO, NO2> and CO) from the nine routine air
quality sites and four hypothetical stations located at the four corners of the model-
ing domain with assumed "clean" values. Since no measurements of VOC were avail-
able, the initial concentration of VOC was assumed to be "clean" (0.025 ppmC) for
the two layers below the diffusion break (250 m AGL). Other species concentrations
in the CB-IV mechanism were set to their minimum value, either 0.000001 or
0.0000000001 ppm. Initial concentration in the three layers above the diffusion
break were assumed to be "clean" as defined in Section 3.
BOUNDARY; Because of the stagnant recirculation conditions of 13 July 1979 and
the short simulation period (0000 to 2000), boundary concentrations do not greatly
influence calculated ozone formation on 13 July 1989. Accordingly, the boundary
conditions used in a previous UAM application to Philadelphia (Haney and Braverman,
1985) were also used here to simplify the comparison with that study (see Chapter 6).
TOPCONC; Because no routine air quality data were measured aloft in the Phila-
delphia region on 13 July 1979, boundary concentrations above the region top were
assumed to be the same "clean" values used for St. Louis (Section 3).
TEMPERATUR; The gridded field of hourly varying surface temperatures was
obtained by using 1/r interpolation from the surface measurements from the routine
network (Table 5-1).
EMISSIONS and PTSOURC; The original emission inventory developed for the EPA
by Engineering-Science (EPA, 1982) was obtained on magnetic tape in the form of
disaggregated raw data files containing information for major point, minor point,
mobile, and area source emissions. These files were then processed with the emis-
sion preprocessor CENTEMS (see SAI, 1989) to obtain a gridded low-level emission
89092r2 5
63
-------
if,
r j
Vn
lr> T
'\« c
IT)
1HQI3H
(SU3J.3U) 1HQ13H
CO
-------
TABLE 5-3. Hourly varying metscalers used in the PLANR UAM
application to Philadelphia.
Hour
Beginning
0000
0100
0200
0300
0400
0500
0600
0700
0800
0900
1000
1100
1200
1300
1400
1500
1600
1700
1800
1900
Exposure
Class
-2
-2
-2
-2
-2
-2
-2
1
1
2
2
2
3
3
3
3
2
2
1
1
N02 Photolysis
Rate
(ppm~' min"')
0.000
0.000
0.000
0.000
0.000
0.018
0.169
0.330
0.444
0.523
0.576
0.608
0.619
0.612
0.586
0.539
0.467
0.363
0.211
0.054
TGRADBELOW
(K/m)
-0.00040
-0.00040
-0.00040
-0.00040
-0.00040
-0.00040
-0.00760
-0.01000
-0.01000
-0.01050
-0.01050
-0.01050
-0.01050
-0.01050
-0.01050
-0.01050
-0.01050
-0.01050
-0.01000
-0.01000
TGRAD ABOVE
(K/m)
-0.00380
-0.00380
-0.00380
-0.00380
-0.00380
-0.00380
-0.00380
-0.00278
-0.00256
-0.00286
-0.00240
-0.00157
-0.00044
-0.02450
-0.02450
-0.02450
-0.05117
-0.14450
-0.13267
-0.13267
89092rl 6
-------
file (EMISSIONS) and elevated emission file for a typical summer weekday with
speciation of hydrocarbons appropriate to CB-IV. No adjustment (e.g., temperature
effects) to the emissions on 13 July 1979 was made. The elevated emission file con-
tains emissions rates and stack parameters for major point sources; this file is used
as input, along with the UAM meteorological inputs, in the point source preprocessor
to obtain the UAM elevated emissions file PTSOURC.
The spatial distributions of the low-level NOX and VOC emissions are shown in
Figures 5-5 and 5-6. Table 5-4 lists the emission totals for NOX and VOC for both
the low-level emissions (EMISSIONS) and elevated emissions (PTSOURC) input files.
Also shown in Table 5-4 are the emission totals used in the earlier UAM(CB-II)
modeling study for Philadelphia (Haney and Braverman, 1985). The emission totals
for the two studies differ because the source-specific or process-specific VOC and
NOX factors used to split the VOC and NOX species are different in the two versions
of the chemical kinetics mechanism (CB-IV or CB-II) and because new updated
speciation splits were used in this study. In addition, we used a different definition
of when a point source is treated as a low-level source.
DIAGNOSTIC SIMULATIONS TO ARRIVE AT A BASE CASE
We performed two diagnostic simulations for the PLANR application of the UAM to
Philadelphia. These diagnostic simulations differed in the observed surface wind
speeds used in the UAM wind preprocessor (the DWM). Because of the large amount
of ozone and ozone precursors transported into cities in the northeastern U.S. like
Philadelphia, and the fairly complex meteorological conditions in Philadelphia on 13
July 1979 (a large amount of wind shear was evident), Philadelphia may not be
amenable to the PLANR use of the UAM. More comprehensive measurement studies
or modeling, such as the Regional Oxidant Modeling for Northeast Transport
(ROMNET) program, are needed to characterize transport in northeastern cities.
Thus it was not surprising that the model performance goalthe predicted region-
wide maximum ozone should be within 20 percent and in the general vincinity of the
observed peak, and the predicted peak ozone at the location of the observed peak
should be within 30 percentcould not be met for Philadelphia in only two diagnostic
simulations.
Diagnostic Run 1
In the first diagnostic run for Philadelphia the routine meteorological and air quality
observations were used in the UAM preprocessors with no adjustments and the model
was exercised from 0000 to 2000 on 13 July 1979. The layer 1 wind fields for diag-
nostic run 1 are shown in Appendix B. Isopleths of predicted hourly ozone concentra-
tions for diagnostic run 1 are shown in Apendix C. Time series of predicted and
observed hourly ozone concentrations are in Figure 5-7 and isopleths of predicted
daily maximum ozone concentrations are shown in Figure 5-8.
89092r2 5
66
-------
V)
O O
CM _J
i i r | i i
O
ro
X
I
O
O
CM
1SV3
o
CN
i i i i i i f
i r i r T i i
iii i ri
i i i I
i i i i i
I t i I i i i i i i i i i I
i i i i i
O
CN
o
CN
fc
(O
o
Jk£
0)
01
s-
o
o; <4-
T o;
3. o
o o-
ID i
O)
O) "O
s- re
3 r
O)'t-
i- .C
U_ O.
1S3M
-------
§
CO
o o
CM _
8
I O
1SV3
o
CN
I I I I I I
I I I 1
I '
o
(N
oc.
o
I I I I I I I I I I
I I I I I I I I I I I I I I I I I
o
CN
o;
o
CN
X
I-
o
to
0) (O
T £
"3. o.
O i
I (U
un o.
3 4J
CD
1S3M
-------
ca
J^
0
CM
tO
C
o
-U
cd
o
H
i-l
a
a.
cd
z
^5
£3
(-H
CQ
O
T3
C
cd
t i
i i
i
z
ca
CJ
cu
±3
J^
o
CM
X
a
§
0}
c
0
JQ
L>
S
O
o
jC
cu
*2
o
(0
cu
trf
o
CM
id1
O
CO
CU
r-l
O
g
to
iH
cd
-U
o
§
H
CO
CO
iH
g
b3
.=r
i
in
u
,_j
CQ
H
OS
^^
^y
4»4
i
z
CQ
O
(1
1
y
CQ
U
CU
o
r:
cu
Li
tu
CM
CM
H
a
i i
^
CQ;
O
J
I 1
14
|
^g
0
U
CU
o
c
cu
CM
cu
CM
^H
«C3
.,
i i
1
CQ
U
4-4
t I
|
?T^
CQ
O
CO
cu
1-1
U
cu
a.
to
in
j.
o
o
o
o"
b-
4»
C^
^
o>
=^
CM"
CO
»
«
t-
^
CM
in
8
o
In
X^
«»
TO
in
in
^^
«»
CO
CO
£^
k
in
CO
1
o
8
cT
»
^
«^
8
o
o
co
CO
k
^
^~
o
^~
^-
^
o
o
in
^~
vO
if
>
^
Od
c^
CO
en
^
^^
t-
-f
8
o
vO
^f
VO
CO
o
b
t-
in
CM
in
i
o
8
vcT
8
«^^
o
o
o
o"
o
^^
CM
3£
in
±
o
o
in
^>
CO
CM
«>
<^
ou
5
vC3^
cT
tn
^y
>,
oo
jgp
in
8
o
^
CM
ON
^
^r
o
vO
^\
**
in
CO
|
O
o
o
vO
s
(\J
v~*
8
o
^^
^y
_qf
«h
CM
X
O
in
CO
in
in
CO
t«^.
O^
».
in
vO
o
^
CM
CO
CO
a*
CO
vO
CO
^
tn
CM
t^
co
ni
^y
*»
CM
CO
vO
^~
CVJ
o
o
«k
^
CO
vO
>
CO
vO
*r
vO
(y\
&\
»~
«»
CM
vO
v3
Q
cc
L,
ev
©
01
-------
24 0
30 30
i i i i I i i i i i i i i i i i i i i i
- AMS LAB
OBSERVED CD -
PREDICTED -
OBSERVED CD -
PREDICTED
12
TIME (HOURS)
12
TIME (HOURS)
18
24
30
12
18
24
i i i i i i i i i i i
- BRIGANTINE
5
X
OBSERVED CD -
PREDICTED
mm.
CD
_a
30 30
I I I I I I i I I I 1 I I I
: BRISTOL
15 15
12
18
24
OBSERVED CD -
PREDICTED
12
TIME (HOURS)
18
24
30
15
12
TIME (HOURS)
18
24
FIGURE 5-7. Time series of predicted and observed hourly ozone concentrations (pphm)
in Philadelphia for diagnostic run 1.
STSTEMS APPLICATIONS. INC.
89092
7n
-------
30
12
18
24
I I I I I | I I I I I | I I I I I | I I I I I
- CAMDEN
OBSERVED CO -
PREDICTED
C-
0.
n CD
30 30
12
18
I I I I | I I I I
CHESTER
15 15
M
O
12
TIME (HOURS)
18
24
OBSERVED UJ
PREDICTED -
m
a
30
15
12
TIME (HOURS)
18
24
30
12
18
i i i i r | i i i i i | i i \ ii | i i r
- CLAYMONT
OBSERVED
PREDICTED
15
24 0
30 30
6
12
18
24
M
O
12
TIME (HOURS)
18
24
I I I i I I I I I I I I I I I I I j i I I I
- CONSHOHOCKEN
OBSERVED [D
PREDICTED
m
m
30
15
12
TIME (HOURS)
FIGURE 5-7. Continued.
89092
SYSTEMS APPLICATIONS, INC.
-------
30
12
18
I I I I l | I I I I I | I I I I I I I I l I l
F DEFENSE SUPP ^^
PREDICTED
Nl
O
2430 30°
12
18
24
6 12 18
TIME (HOURS)
_ 1 f I I T f 1 1 I I I | I I I I I | I 1 ! I I
- DOWNINGTON m
OBSERVED n
PREDICTED
M
O
24
30
15
12
TIME (HOURS)
18
24
30
12
18
24
I I \ I
- FRANKLIN INS
0.
O.
i i r i i i i i i i r i i i i
OBSERVED
PREDICTED
30 30
12
18
24
I I I 1 I 1 I I I I | I I I I 1 j I I T 1
- ISLAND RD Al
OBSERVED CD -i
PREDICTED
15 °r 15
ISI
o
6 12 18
TIME (HOURS)
24
i i i l i i i i i i i i i i
30
15
12 18
TIME (HOURS)
24
FIGURE 5-7. Continued.
89092
SYSTEMS APPLICATIONS. INC.
-------
12
18
24
I I I I | I i
- LUMBERTON
15
N
O
OBSERVED
PREDICTED
12
TIME (HOURS)
0 30
12
18
24
1 I I I I | I I I T
- NORRI5TOWN A
2
X
18
I I 1 I I ] I II I I
OBSERVED UJ -
PREDICTED
12
TIME (HOURS)
18
30
15
12
18
24
5
I
0.
tsl
O
I I I I I [ I \ I I I I I I 1 I i I I I I I I
: ROBBINSVILLE m
OBSERVED [D
PREDICTED
I I i i i I i i
30
12
6 12 18
TIME (HOURS)
24
i i i f i | i i i i i i i i i i i i ; i i i i
h ROXY WATER P
12
TIME (HOURS)
18
24
FIGURE 5-7. Continued.
89092
SYSTEMS APPLICATIONS. INC.
-------
12
18
24
I I I I I | I I I I I | I I l I l
SE SEWAGE PL
N
o
PREDICTED
12
TIME (HOURS)
18
30 30
12
18
24
1 I I I I | I I I I I | I I I I I | I I I I I
- SOUTH BROAD
OBSERVED [0 -
PREDICTED
2
15 fr 15
Kl
O
CD
12
TIME (HOURS)
18
30
15
30
12
18
I I i I (| I I I I I | I 1 I I I | I l 1 I 1
- SUMMIT BRIDG m
OBSERVED [3 -
PREDICTED
15
tsj
o
2430 30°
12
TIME (HOURS)
18
12
18
24
i i i i i | i t i r i
h SW CORNER
X
15 9- 15
M
O
24
OBSERVED CD -
PREDICTED -
12
TIME (HOURS)
18
30
15
FIGURE 5-7. Continued.
89092
SYSTEMS APPLICATIONS, INC.
74
-------
30
12
18
24
I I I I | I T
>- TRENTON
I
Q.
0.
M
O
I I I I 1 I I I I I I
OBSERVED CD
PREDICTED
30 30
12
18
24
M
O
12
TIME (HOURS)
18
24
I I I I I | I I I I T
- VAN HISEVILL
OBSERVED [JJ -
PREDICTED
30
15
12
TIME (HOURS)
18
24
12
18
24
^niiii|iiiii|r
- VINELAND
^
0.
fsl
O
m
CO
OBSERVED CD -
PREDICTED
m
CO
CO
30
15
6 12 18
TIME (HOURS)
24
FIGURE 5-7. Concluded.
89092
SYSTEMS APPLICATIONS. INC.
75
-------
Time : 0 - 2400 1ST
387 407 427
447
NORTH
467 487
Maximum Value = 18.66
.Minimum Value = 8.15
507 527 547 567
30
20
10
I J I } I ! I J I » I } I 4 t J
1 (I I > J I 4. ,
1/lliliflt t I I t 1 t
9 -
4500
4480
4460
4440
4420
4400
4380
4360
4340
10
20
30
SOUTH
FIGUPE 5-8. Isopleths of predicted daily maximum ozone concentrations (pphm)
from diagnostic run 1 with superinposed observations.
76
-------
The predicted region-wide maximum ozone concentration (18.7 pphm) is within 9
percent of the observed peak (20.5 pphm) and occurs approximately 45 km to the east
of the location of the observed peak (see Figure 5-8). At the location of the
observed peak ozone (Conshohocken) the predicted maximum daily ozone concentra-
tion (11.1 pphm) is only within 46 percent of the observed peak.
As seen in the time series plots of predicted and observed hourly ozone concentra-
tions for diagnostic run 1 (Figure 5-7), the model does a good job in replicating the
observed hourly ozone concentrations in areas far southeast of downtown (Ancora),
to the southwest of downtown (Claymont and Summit Bridge), far east of downtown
(Lumberton), and northeast of downtown (Bristol and Trenton). At the downtown
sites in Philadelphia (AMS Lab, Camden, Defense Support, Franklin Institute, Island
Rd., South Broad, and SW Corner) diagnostic run 1 usually underpredicts the peak
observed daily maximum ozone concentrations, although this underprediction is well
within 30 percent. However, diagnostic run 1 underpredicts the observed daily
maximum ozone concentrations at the area of peak observed ozone concentrations
(the three sites north of downtown) by 43 percent (Roxy Water), 46 percent
(Conshohocken), and 40 percent (Norristown).
At most sites in the vicinity of downtown the model predictions do not rise as fast as
the observations; thus predicted peaks are lower than the observations and occur
later in the day. There are many possibilities for the delay in the rise of the predic-
tions including: (1) insufficient ozone and ozone precursors aloft are entrained as the
mixing height rises; or (2) photochemistry early in the day is insufficient because
VOC concentrations are lower than they should be.
At sites northwest of Philadelphia, where the highest ozone concentrations were
observed, the same problems exist as for the downtown sites, except that around
1200 the wind fields used in diagnostic run 1 advect the elevated ozone cloud east-
ward away from the observation sites.
A scatterplot of predicted and observed hourly ozone concentrations and model per-
formance statistics for diagnostic run 1 are given in Figure 5-9. When all hours are
considered, the underprediction in the peak observations is compensated for by the
overprediction of the low nighttime observations, resulting in an overprediction bias
of 13 percent (-0.8 pphm). The average absolute (gross) error is about 45 percent (2.8
pphm). The model replicates some of the diurnal variability of the hourly ozone
observations, with a correlation coefficient of 0.68; however, the predicted daily
maximum ozone concentrations occur later in the day than the observed ozone peaks
(Figure 5-7).
Time series of predicted and observed hourly NO, NC^, and CO concentrations for
diagnostic run 1 are shown in Appendix D. The model usually follows the trends of
the observed NO and N©2 concentrations but has a tendency to underpredict. Of
89092r2 5
77
-------
20.00 -
15.00 -
Q.
Q.
Qi
Q_
5.00 -
5.00
10.00 15.00
OBSERVED (pphm)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE
STANDARD DEVIATION
SKEWNESS
KURTOSIS
OTHER MEASURES
MEDIAN
UPPER QUARTILE
LOWER QUARTILE
MINIMUM VALUE
MAXIMUM VALUE
6.22966
4.56076
0.64853
-0.28793
5.50000
9.30000
2.20000
0. 10000
20.50000
7.02572
3.82526
-0.38259
-0.87783
7.89000
9.90000
4.00000
0.01000
15.25000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.680
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.617 HIGH BOUND 0.734
RATIO OF OVER TO UNDER PREDICTIONS 1.598
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 21.818
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 10.303
FIGURE 5-9a. Scatterplot and nodel performance statistics for hourly ozone concentrations
in Philadelphia and diagnostic nan 1 (N = 330) .
89092
78
-------
0.20 -
-8.00 -4.80 -1.60
RESIDUAL (08S-PRED)
THE BINSIZE EQUALS 1.600
1.60
4.80
8.00
RESIDUAL ANALYSIS
AVERAGE
STANDARD DEVIATION
SKEWNESS
KURTOSIS
OTHER MEASURES
MEDIAN
UPPER QUARTILE
LOWER QUARTILE
MINIMUM VALUE
MAXIMUM VALUE
-0.79613
3.42284
0.31528
-0.07560
-0.81000
1.46000
-3.27000
-8.11000
10.44000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -1.3367
UPPER BOUND -0.2555
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 10.3549
UPPER BOUND 13.3825
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 3.51
THE AVERAGE ABSOLUTE ERROR IS 2.82
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7321
RESIDUAL COEFFICIENT OF VARIATION
0.5494
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.7505
FIGURE 5-9b. Residual analysis plot and nodel performance statistics for hourly
ozone concentrations in Philadelphia and diagnostic nan 1 (N = 330).
-------
particular note is the peak observed NC>2 concentration at Norristown in the after-
noon; this peak is not reproduced by the model. It appears that in the early morning
urban NOX emissions are transported south and then northwest, picking up additional
emissions as the plume passes over the urban area, then passing the air quality moni-
tors northwest of downtown Philadelphia. This transport condition does not appear
to be present in the wind fields used in diagnostic run 1. The model tends to sys-
tematically underpredict CO concentrations at all sites (Appendix D).
Diagnostic Run 2
In diagnostic run 1 it appears that the photochemistry is too slow in the vicinity of
downtown, resulting in predicted ozone peaks that are too low and occur too late in
the day. In addition, the wind fields used in diagnostic run 1 appear to advect the
Philadelphia urban ozone plume too far eastward, as evidenced by the overprediction
at the most eastward air quality monitor (Van Hiseville) (Figure 5-7). Photochem-
istry can be speeded up by increasing VOC concentrations, either through increased
initial VOC concentrations at the surface and aloft, increased VOC emissions, or
other adjustments to UAM inputs that will increase predicted VOC concentrations,
such as lowering the mixing heights or adjusting the wind field so that pollutants are
retained in the vicinity of the downtown area longer.
The most obvious way to improve model performance is to include sources of VOC
emissions that were missing from the 1979 inventory, such as biogenic emissions,
mobile source running loss emissions, temperature effects on evaporative VOC emis-
sions, and previously uninventoried VOC sources. The development of such a detailed
1979 emission inventory for Phildelphia is beyond the scope of this project. Fur-
thermore, because of the need to compare the PLANR UAM with past UAM applica-
tions that used the rich POS data base, the use of drastically different inventories
would overly complicate the analysis.
There is some justification for increasing the initial concentrations since, except for
mixed-layer (lowest 250 m) O-j, NO, NO2> and CO concentrations, all other concen-
trations were assumed to be clean. However, there were no routine upper-level air
quality measurements available from which to base higher initial concentrations.
Given the possibility of bias in the observed wind speeds at some of the sites (see
Morris et al., 1989c), the fact that the urban plume in diagnostic run 1 travels too far
east, and the success in the St. Louis PLANR UAM application, it was decided to
reduce observed surface wind speeds at FAA/NWS sites by 50 percent in diagnostic
run 2. Thus new wind fields were developed (see Appendix E) for diagnostic run 2.
All other inputs were the same as diagnostic run I.
Time series of predicted and observed ozone concentrations for diagnostic run 2 are
shown in Figure 5-10, and isopleths of hourly predicted ozone concentrations are con-
tained in Appendix F. There is very little difference in model performance between
89092r2 5
80
-------
30
12
18
24
12
18
I I I I I | I I I I I I I I I I I ( I I I I I
F AMS ^ OBSERVED
PREDICTED
I
°-
Q.
_ I I I I I | I I I I I | I I I I I | I I I I I _
- ANCORA
OBSERVED
PREDICTED
12
TIME (HOURS)
12 18
TIME (HOURS)
24
12
18
24
fsl
O
ou I 1 I I I I I 1 I I
: BRIGANTINE
15 -
j
Cfe 0°
- , t?mm[Tj i , , , ,
°0 6
i i i i i i i i i i i i
OBSERVED [JJ -
PREDICTED -
m -
mm-[Dj
i i i i i i i i i i i i r
12 18 2
ou
15
4°
12
18
i i i i i I i i i i i i i i i t i
: BRISTOL m
OBSERVED 0] -
PREDICTED
TIME (HOURS)
12
TIME (HOURS)
FIGURE 5-10. Tine series of predicted and observed hourly ozone concentrations (pphm)
in Philadelphia for diagnostic run 2.
SYSTEMS APPLICATIONS, INC.
89092
81
-------
12
18
24
I I I I I I I I I I I I I I I I I I I I I I I
: CAMDEN
OBSERVED
PREDICTED
12
TIME (HOURS)
18
24
12
TIME (HOURS)
18
24
30
i i i i | i r
- CLAYMONT
15
N
O
12
18
24
ri i i i
OBSERVED
PREDICTED
30 30
12
18
24
I I I I I j I I I I I | I I
h CONSHOHOCKEN
x
15 9- 15
ISJ
O
12
TIME (HOURS)
18
24
OBSERVED
PREDICTED
30
15
12 18
TIME (HOURS)
24
FIGURE 5-10. Continued.
89092
SYSTEMS APPLICATIONS. INC.
I
82
-------
30
12
18 24
I I I I I I I I I I I [ I I I I I ) I I I I I
: DEFENSE SUPP m
OBSERVED CD -
PREDICTED
15
I I I i^-f1--1 I I I I I I I I I ! I I I I
30 30
6 12 18
i i r~i | i i i i i | i i i i i | i
- DOWNINGTON
OBSERVED
PREDICTED
X
15 15
M
O
0 0
om-
30
15
6 12 18 24" 0
TIME (HOURS)
6 12 18
TIME (HOURS)
24
30
12
18 24
i I I I I ] F I i I I | I I I I I I I I i I I
- FRANKLIN INS
OBSERVED [JJ -
PREDICTED
30 30
12 18
24
i i i i | i i i i i | i i i i i i i i i i
- ISLAND RD Al
OBSERVED H
PREDICTED
2
15 15
M
O
12
TIME (HOURS)
18 24
30
15
6 12 18
TIME (HOURS)
24
FIGURE 5-10. Continued.
89092
83
SYSTEMS APPLICATIONS, INC.
I
-------
12
18
24
2
I
tSI
O
~I I I I I I I I I I I I I I I I I I I I I \
- NORRISTOWN A m
OBSERVED CD -
PREDICTED
r i i i i | i i i i i i i i I i i i i i i
P LUMBERTON
30
15
12
TIME (HOURS)
12
TIME (HOURS)
30
12
18
24
_niiiiiiiiii|iiiiiiiiiir^
: ROBBINSVILLE m
OBSERVED CD -
PREDICTED
x
°-
0.
ISI
O
30 30
12
18
24
I I I I I I I I I I I I I I I I I I I I I I I
- ROXY WATER P
OBSERVED CO -
PREDICTED
tsj
O
6 12 18
TIME (HOURS)
24
0 0
w
- CO
cocoa
co -
%j i i i^xi i i i i i i i i i i__i I i i j i
30
15
6 12 18
TIME (HOURS)
24
FIGURE 5-10. Continued.
89092
SYSTEMS APPLICATIONS, INC.
84
-------
12
18
i i i i i i i i i i i i i i i i i i i i i i r
: SE SEWAGE PL m :
OBSERVED [JJ
PREDICTED
Nl
o
I I I I
30 30
12
18
24
IIIII|IIIIIIIIIII|IIIIT^
: SOUTH BROAD m
OBSERVED CD -
PREDICTED
15 - 15
Nl
O
12
TIME (HOURS)
18
24
30
15
12
TIME (HOURS)
18
24
30
12
18
I I I I I I I I I I I I
: SUMMIT BRIDG
i i i | i i i r
OBSERVED UJ -
PREDICTED
24
30
12
TIME (HOURS)
18
30
12
18
24
I I i I I
1 I 1 I I | I I I F I \ \ I I I
: SW CORNER m
OBSERVED [JJ -
PREDICTED
0
COD
\L_ I I I
30
15
12
TIME (HOURS)
18
24
FIGURE 5-10. Continued.
89092
SYSTEMS APPLICATIONS, INC.
85
-------
30
12
18
24
12
18
I I I I I | I I I I I | I I I I I | I I I I I
- TRENTON
OBSERVED
PREDICTED
I 15
o
OBSERVED CD -
PREDICTED -
12
TIME (HOURS)
18
12 18
TIME (HOURS)
24
12
18
24
_ T i i i i | i i i r ITT i i i r | i i i i \
^ VINELAND
OBSERVED [JJ -
PREDICTED
I
CL
C_
m
15
12 18
TIME (HOURS)
24
FIGURE 5-10. Concluded.
89092
SYSTEMS APPLICATIONS. INC.
86
-------
diagnostic runs 1 and 2. The predicted region-wide maximum ozone concentrations
(18.2 pphm) is within 11 percent of the observed peak (20.5 pphm) and occurs
approximately 45 km east of the location of the observed peak (Conshohocken). At
the location of the observed peak the predicted daily maximum ozone (11.1 pphm) is
only within 46 percent of the observed peak. The similarity between diagnostic runs
1 and 2 is also seen in the model performance statistics (Figure 5-11). Over all
hours, diagnostic run 2 overpredicts the observed hourly ozone concentrations by
about 12 percent and exhibits a gross error of about 45 percent. There is a slight
improvment in the correlation coefficient in diagnostic run 2 (0.69) over diagnostic
run 1 (0.68) but the difference is not significant.
PLANR Base Case
Neither of the two diagnostic runs met the model performance goal. Although the
predicted region-wide maximum ozone concentrations were fairly close to the
observed maximum (within 9 and 11 percent for diagnostic runs 1 and 2, respec-
tively), the predicted peak occurs too far east (45 km) of the observed peak. In
addition, the predicted daily maximum ozone at the location of the observed peak is
only within 46 percent.
Even though the diagnostic simulations did not meet the performance goal, time con-
straints dictate that one of the simulations be selected as a PLANR UAM base case
for further analysis. Thus, use of one of these simulations served as a stringent test
of the effects of model performance on calculated ozone in response to emission
reductions. Because diagnostic run 1 used observed data as is with no adjustments,
and it has not been demonstrated that reducing wind speeds at FAA/NWS sites in the
Philadelphia is justifiable, we selected diagnostic run 1 as the PLANR base case.
89092P2 5
-------
20.00 -
15.
a.
a.
10.
a;
a.
5.00
5.00 10.00 15.00
OBSERVED (pphm)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE
STANDARD DEVIATION
SKEWNESS
KURTOSIS
OTHER MEASURES
MEDIAN
UPPER QUARTILE
LOWER QUARTILE
MINIMUM VALUE
MAXIMUM VALUE
6.22966
4.56076
0.64853
-0.28793
5.50000
9.30000
2.20000
0.10000
20.50000
6.98781
3.86320
-0.34437
-0.87080
7.86000
10.00000
3.63000
0.01000
15.00000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.690
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.629 HIGH BOUND 0.743
RATIO OF OVER TO UNDER PREDICTIONS 1.461
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 22.121
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 10.303
FIGURE 5-lla. Scatterplot and irodel perfontance statistics for hourly ozone concentrations
in Philadelphia and deiagnostic nan 2 (N = 330).
89092
-------
0.20 -
-8.00 -4.80 -1.60
RESIDUAL (OBS-PRED1
THE BINSIZE EQUALS 1.600
1.60
4.80
8.00
RESIDUAL ANALYSIS
AVERAGE -0.75822
STANDARD DEVIATION 3.37841
SKEWNESS 0.26279
KURTOSIS -0.03347
OTHER MEASURES
MEDIAN -0.73000
UPPER QUARTILE 1.41000
LOWER QUARTILE -3.35000
MINIMUM VALUE -8.93000
MAXIMUM VALUE 10.47000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -1.3011
UPPER BOUND -0.2154
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 10.0878
UPPER BOUND 13.0373
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 3.46
THE AVERAGE ABSOLUTE ERROR IS 2.78
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7321
RESIDUAL COEFFICIENT OF VARIATION
0.5423
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
FIGURE 5-llb. Residual analysis plot and model performance statistics for hourly ozone
concentrations in Philadelphia and diagnostic run 2 (N = 330).
89092
-------
6 EVALUATION OF THE PLANR UAM APPLICATION TO PHILADELPHIA
The PLANR UAM base case simulation for Philadelphia was evaluated in two ways:
(1) by comparing performance with applications of two versions of the model
(UAM/CB-II and UAM/CB-IV) in which inputs were developed from a rich data base
(the Philadelphia Oxidant Study, POS); and (2) by comparing the reductions in ozone
concentrations calculated in the three applications in response to emission control
scenarios.
COMPARISON OF MODEL PERFORMANCE
The performance of the PLANR UAM base case simulation (diagnostic run 1, dis-
cussed in Section 5) was compared with the performance of UAM simulations that
used data from the Philadelphia Oxidant Study (POS). The 13 July 1979 oxidant epi-
sode in Philadelphia was used for all simulations. Routine and special surface
meteorological data were available from 16 sites and surface air quality data were
available from over 21 sites in the Phildelphia modeling domain (see Figure 5-2).
However, because the POS was not in full operation until 18 July, no special upper-
air soundings were available for 13 July 1979.
Meteorological data were developed for input to the UAM(CB-II); many diagnostic
simulations were completed before a satisfactory base case was obtained (Haney and
Braverman, 1985). The final base case was used in our comparison and is referred to
here as UAM(CB-II). The same meteorological inputs developed from the POS data
for the UAM(CB-II) were also used for the UAM(CB-IV) application that is compared
here; this application is referred to as POS UAM. In the following paragraphs we
briefly discuss how the modeling inputs were prepared from the POS data for the
UAM(CB-H) and POS UAM; the input preparation is discussed in full by Haney and
Braverman (1985). Note that the CB-IV emission inputs developed in this study dif-
fered in quantity and composition from the CB-II emissions developed previously.
Development of Inputs Using a Rich Data Base (the Philadelphia Oxidant Study)
The development of inputs for the data-intensive application of the UAM to Phila-
delphia required extensive data analysis and many diagnostic simulations (Haney and
Braverman, 1985). Wind data were examined and questionable observations were
deleted from the analysis. Surface wind data were then interpolated to create the
89092r2 7
-------
UAM layer 1 wind field. The winds in the top layer (layer 4 in both the UAM(CB-II)
and POS UAM) were based on upper-air soundings from Dulles and JFK airports and
were assumed to be constant. Following recommendations from the EPA, wind fields
in the layers between layer 1 and layer 4 were obtained by linear interpolation
between the wind vectors in layers 1 and 4.
A very elaborate scheme for creating mixing height fields for Philadelphia was
developed to account for the urban heat island effect. As in the PLANR UAM inputs
for Philadelphia, daytime mixing heights were based on the upper-air soundings at
JFK and Dulles airports. At night and early morning (before 0700 EST) the nighttime
minimum diffusion break was assumed to be 250 m AGL over urban areas (approxi-
mately the stipled area in Figure 4-2) and 100 m AGL over rural areas. The urban
and rural mixing heights were forced to match by 0900 EST.
The UAM(CB-II) and POS UAM used the same initial concentrations (AIRQUAL) and
boundary conditions (BOUNDARY) in the two layers above the mixing height (layers
3 and 4), the same boundary conditions (except for the southwest boundary) in the
two layers below the mixing height, and the same concentrations above the region
top (TOPCONC). These concentrations were as follows: 3 ppb NOX; 80 ppb O3; 58
ppbC VOC; and 200 ppb CO.
Initial concentrations in the lowest UAM layer (layer 1) were obtained by interpola-
ting the two-hour (2300 12 July to 0100 13 July) average concentrations from the
dense measurement network (see Figure 5-2). Initial concentrations in layer 2 were
obtained by linear interpolation between the values in layer 1 and layer 3. Boundary
conditions for the southwest boundary below the mixing height were based on mea-
surements.
Gridded emission inventories for minor and elevated point sources and areas sources
were prepared for the UAM(CB-II) by Engineering Science, Inc. in 1981 for the EPA
(EPA, 1982). The POS UAM used the same emission inventory used in the PLANR
UAM (see Figures 4-5 and 4-6). (The differences between the emissions inventories
for UAM(CB-II) and UAM(CB-IV) (PLANR UAM and POS UAM) are shown in Table
4-4.)
The UAM(CB-II) and POS UAM simulations provide one of the first opportunities for
a comparison of the CB-II and CB-IV versions of the UAM. However, differences in
emissions (see Table 5-4) and in the speciation of the initial and boundary VOC
concentrations complicate the comparison of the CB-II and CB-IV versions of the
UAM. Other comparisons of the CB-II and CB-IV versions of the UAMfor
applications to New York (Morris et al., 1989a) and St. Louis (Section 3 of this
report)were complicated by the fact that different meteorological and air quality
inputs were used in each model version. In the following paragaphs we discuss the
performance of the POS UAM and then compare the performance of the PLANR
UAM, POS UAM, and UAM(CB-II). The performance of the UAM(CB-II) is discussed
by Haney and Braverman (1985).
89092r2 7
91
-------
POS UAM Performance
The daily maximum ozone concentrations predicted by the POS UAM are shown in
Figure 6-1. Time series of predicted ozone, NO, NO2, and CO concentrations are
shown in Appendix G; the predicted hourly ozone concentrations are shown in Appen-
dix H. As seen in Figure 6-1, the region-wide maximum ozone concentration predic-
ted by the POS UAM (23.6 pphm) is within 15 percent of the observed peak (20.5
pphm) and occurs approximately 28 km nortneast of the location of the observed
peak. At the location of the peak observation, the POS UAM predicts the observed
peak to within 14 percent (17.7 pphm).
An examination of the time series of predicted and observed ozone concentrations in
Appendix G reveals that the POS UAM does a respectable job of replicating the
hourly observed ozone concentrations at most sites. At the three sites northwest of
downtown that exhibited the highest observed concenrations (Conshohocken, Roxy
Water, and Norristown) the model predicts the rise in the ozone observations but
underpredicts the peaks at two of the sites (Conshohocken and Roxy Water).
The fairly good ability of the POS UAM to predict the hourly ozone observations is
also reflected in the scatterplot and model performance statistics for hourly ozone
(Figure 6-2). Over all hours the model has an overprediction bias of approximately
23 percent. The absolute average (gross) error for hourly ozone concentrations was
41 percent. The high correlation coefficient (0.80) indicates that the POS UAM
reproduces the spatial and temporal variations in the hourly ozone observations quite
well.
Comparative Performance of PLANR UAM, POS UAM, and UAM(CB-II)
Model performance statistics for the PLANR UAM, POS UAM and UAM(CB-II) appli-
cations to Philadelphia on 13 3uly 1979 are given in Table 6-1. The peak observation
(20.5 pphm) is overpredicted by the UAM(CB-II) and POS UAM by 30 and 15 percent,
respectively, and underpredicted by the PLANR UAM by 15 percent. At the location
of the peak observation the peak is underpredicted by 10, 14, and 46 percent by the
UAM(CB-II), POS UAM, and PLANR UAM, respectively. The UAM(CB-II) and POS
UAM clearly exhibit considerably more skill at predicting the peak observation than
does the PLANR UAM.
The UAM(CB-II) and POS UAM also show more skill in predicting the daily maximum
ozone concentrations at the 19 ozone monitors; their bias and gross error are
approximately 3 and 0.8 times lower, respectively, than the bias and gross error for
the PLANR UAM. Although there are subtle differences in the performance sta-
tistics of the UAM(CB-II) and POS UAM, the differences are not statistically signifi-
cant.
89092r2 7
92
-------
Time : 0 - 2000 LSI
387 407 427
447
NORTH
467 487
Maximum Value = 23.59
Minimum Value = 8.77
507 527 547 567
i j i < i { i i i/ j j i i j i i ) { i t i { i
4500
C 4480
- 4460
- 444O
SOUTH
- 4420
4400
- 4380
- 4360
4340
FIGURE 6-1. Predicted maxinum daily ozone concentrations (pphin) for
the POS UAM with superimposed daily maximum observations.
93
-------
5.00 10.00 15.00
OBSERVED (pphm)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 6.22966 7.63786
STANDARD DEVIATION 4.56076 4.27899
SKEWNESS 0.64853 -0.07555
KURTOSIS -0.28793 -0.71528
OTHER MEASURES
MEDIAN 5.50000 8.34000
UPPER QUARTILE 9.30000 10.71000
LOWER QUARTILE 2.20000 4.02000
MINIMUM VALUE 0.10000 0.01000
MAXIMUM VALUE 20.50000 18.34000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.801
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.759 HIGH BOUND 0.837
RATIO OF OVER TO UNDER PREDICTIONS 2.056
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 22.727
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 8.182
FIGURE 6-2a. Scatterplot and model performance statistics for hourly
ozone concentrations in Philadelphia and POS UAM (N=330).
94
-------
0.20 -
-6.00 -3.60 -1.20
RESIDUAL IOBS-PRED1
THE BINSIZE EQUALS 1.200
1.20
3.60
6.00
RESIDUAL ANALYSIS
AVERAGE -1.40829
STANDARD DEVIATION 2.80004
SKEWNESS 0.22466
KURTOSIS -0.00769
OTHER MEASURES
MEDIAN -1.56000
UPPER QUARTILE 0.60000
LOWER OUARTILE -3.44000
MINIMUM VALUE -8.18000
MAXIMUM VALUE 7.83000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -1.9763
UPPER BOUND -0.8403
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 6.9295
UPPER BOUND 8.9555
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 3.13
THE AVERAGE ABSOLUTE ERROR IS 2.54
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7321
RESIDUAL COEFFICIENT OF VARIATION
0.4495
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.6139
FIGURE 6-2b. Residual analysis plot and model perfontance statistics
for hourly ozone concentrations in Philadelphia and POS HAM (N=330).
95
-------
TABLE 6-1. Comparison of performance statistics for the UAM(CB-II),
POS UAM, and PLANR UAM applications to Philadelphia for 13 July 1979.
UAMPOSPLANR
Performance Measure (CB-II) UAM UAM
Hourly Ozone Concentrations (matched by time and location)
Number of pairs N/A 330 330
Average observed (pphm) N/A 6.23 6.23
Average predicted (pphm) N/A 7.64 7.03
Bias (pphm) N/A -1.41 -0.80
Average percent difference N/A 23 13
Average absolute (gross) error (pphm) N/A 2.54 2.82
Gross error percent difference N/A 41 45
Correlation coefficient N/A 0.80 0.68
Daily Maximum Ozone Concentration (matched by location but not time)
Number of pairs 19 19 19
Average observed (pphm) 13.90 13.90 13.90
Average predicted (pphm) 14.90 13.21 11.75
Bias (pphm) -1.00 0.69 2.15
Average percent difference 7 5 16
Average absolute (gross) error (pphm) 2.02 2.21 3.60
Gross error percent difference 15 16 26
Correlation coefficient 0.73 0.72 -0.10
Peak Ozone Concentration
Peak observed (pphm) 20.5 20.5 20.5
Unmatched by time or location:
Predicted region-wide maximum (pphm) 26.6 23.6 18.7
Ratio of prediction to observation 1.30 1.15 0.91
Matched by location but not time:
Predicted maximum (pphm) 18.5 17.7 11.1
Ratio of prediction to observation 0.90 0.86 0.54
Hours difference in prediction +200
to observation
96
-------
Discussion
The seemingly "good" performance of the POS UAM (and UAM(CB-II)) compared to
the "poor" model performance of the PLANR UAM is due to differences in the fol-
lowing inputs:
Layer Structure; The POS UAM used four vertical layers, two below and two above
the diffusion break, whereas the PLANR UAM used five vertical layers, two below
and three above the diffusion break.
Wind Fields (WIND); The POS UAM wind field was generated using meteorological
data from 16 surface and two upper-air meteorological observations sites. The data
were extensively screened and the wind field was created after several iterations of
data manipulation, variations of interpolation techniques, and diagnostic simula-
tions. The PLANR UAM wind field was generated from just the routine data (11 sur-
face and two upper-air sites). The routine data were used in the PLANR UAM wind
preprocessor without any adjustments.
Mixing Heights (DIFFBREAK); The POS UAM used nighttime and morning mixing
heights that neglected the presence of the urban heat island. The nocturnal mixing
height was 250 m AGL over the urban areas and 100 m AGL over the rural areas.
The daily maximum mixing height used in the PLANR UAM (1775 m AGL) was
approximately 16 percent (250 m) higher than that used in the POS UAM (1525 m
AGL). At night, the PLANR UAM used the same diffusion break value as the POS
UAM in urban regions (250 m AGL), but over rural regions the PLANR UAM diffusion
break (250 m AGL) was 2.5 times the value used in POS UAM (100 m AGL).
Other Meteorological Inputs: Temperature gradients, water vapor concentrations
(METSCALERS), and surface temperatures (TEMPERATUR) were also different in
the POS and PLANR UAM.
Initial Concentrations (AIRQUAL); The initial concentrations used in the POS and
PLANR UAM are quite different. Because there were no routine VOC measure-
ments, "clean" values were used in the PLANR UAM application to Philadelphia. In
addition, since no routine air quality measurements aloft were available, "clean"
values were used for NOX, ozone, and VOC for initial concentrations for the PLANR
UAM in layers above the DIFFBREAK (250 m AGL). In contrast, considerable analy-
sis of the extensive POS data base went into the development of the initial condition
concentrations below and above the DIFFBREAK for the POS UAM. In total, the
initial VOC and ozone concentrations were approximately 4 and 2 times greater in
the POS UAM than the values used in the PLANR UAM. Initial NOX concentrations
were about the same; initial NOX was about 15 percent lower in POS UAM than in
PLANR UAM.
89092r2 7
97
-------
Boundary Conditions Aloft (TOPCONC): Boundary conditions above the region top
were much higher in POS UAM than in PLANR UAM. Assumed concentrations of
NOX, VOC, and ozone were 2 to 3 times higher in the POS UAM than in the PLANR
UAM.
The primary differences between the POS and PLANR UAM simulations can be
divided into two catgories: meteorology (winds, mixing heights, temperatures, and
metscalers) and air quality (initial and boundary conditions). It is difficult ascertain
whether the differences between the POS and PLANR UAM are primarily due to
meteorology or air quality inputs! Thus a sensitivity simulation was performed that
used the PLANR UAM meteorological inputs (WINDS, DIFFBREAK, METSCALERS,
and TEMPERATUR) and the POS UAM air quality inputs (AIRQUAL, BOUNDARY,
and TOPCONC). The results of the meteorology sensivity test are displayed in
Appendix I.
As seen in Appendix I, the sensitivity test produces results that are very similar to
the PLANR UAM; there is a very slight improvement in model performance. The
predicted region-wide maximum ozone in the sensitivity test is within 14 percent,
compared to 19 percent for the PLANR UAM. However, the predicted elevated
ozone cloud in the meteorological sensitivity test is still too far to the east. Thus
the reason for the "good1 model performance of the POS UAM and UAM(CB-II) is due
to the meteorological inputs. In order to obtain better model performance in the
PLANR UAM, the wind fields would have to be altered so that the elevated ozone
cloud is retained north of downtown.
CORRECTIONS FOR MODEL BIAS IN CALCULATIONS OF OZONE
CONCENTRATIONS IN RESPONSE TO EMISSION CONTROL STRATEGIES
As was done for the St. Louis analysis, we examined three different techniques for
accounting for model bias when using the model to demonstrate attainment of the
ozone NAAQS: the uncorrected bias approach, the decrement approach, and the per-
centage approach (see Section 4 for explanation).
The POS UAM, PLANR UAM, and UAM(CB-II) were .exercised for uniform (across the
board) VOC emission control scenarios. In the POS UAM simulations the VOC initial
and boundary conditions in excess of background (25 ppbC) were reduced the same
percentage as the VOC emissions in these control scenarios. However, because the
PLANR UAM used background initial and boundary VOC concentrations, the same
initial and boundary conditions were used for all of the PLANR UAM emission con-
trol simulations. The UAM(CB-II) VOC emission control scenarios also used the same
VOC initial and boundary conditions (Haney and Braverman, 1985).
The region-wide maximum ozone concentration calculated in the VOC emission con-
trol scenarios are given in Table 6-2. When overprediction (POS UAM and UAM(CB-
II)) or underprediction (PLANR UAM) of the observed peak is not accounted for (the
89092r2 7
98
-------
TABLE 6-2. Predicted region-wide maximum ozone
concentrations for the POS UAM, PLANR UAM, and
UAM(CB-II) for VOC emission reduction scenarios.
Percent
VOC
Emission
Reduction
0
25
50
60
75
90
Region-wide Kaximum Ozone
Concentration (ppnm)
POS UAM
23.6
N/A
N/A
13.9
N/A
10.6
PLANR UAM
18.7
N/A
N/A
12.2
N/A
10.8
UAM(CB-II)*
26.6
20.6
14.7
N/A
12.4
N/A
* From Haney and Braverman, 1985.
89092rl 3
99
-------
unconnected bias approach), it is estimated from the data in Table 6-2 that the
reductions of VOC emissions required to reduce ozone concentrations to below the
NAAQS (12 pphm) are 77 percent for POS UAM, 79 percent for UAM(CB-II), and 64
percent for PLANR UAM. This result is as expected when model bias is not correc-
ted: the higher the predicted region-wide maximum ozone concentration, the
steeper the VOC emission reductions that are needed to reduce the peak ozone con-
centration to the NAAQS.
Using the decrement approach, the POS UAM results yield a 54 percent reduction in
VOC erniicLonia, whereas the PLANR UAM results mean that over 90 percent of the
VOC emissions would have to be eliminated for ozone attainment. The UAM(CB-II)
estimates that a 36 percent reduction in VOC emissions is needed. The extra emis-
sion reduction required by the PLANR UAM, compared to the POS UAM, is not sur-
prising in view of the underprediction of the peak observations by the PLANR UAM
and overprediction by the POS UAM. Also, in the PLANR UAM more of the predic-
ted peak ozone concentration is due to the irreducible background. The difference in
VOC reductions for UAM(CB-II) and POS UAM is somewhat surprising, however,
since the modeling inputs were essentially the same. Furthermore, in the POS UAM
initial and boundary conditions were also reduced in the emission reduction scenarios,
whereas they were not in the UAM(CB-II). The only explanations are (1) the peak
observation predicted by the UAM(CB-II) (30 percent) is higher than that of POS
UAM (15 percent); (2) different chemical mechanisms were used; and (3) the
emissions were different.
Using the percentage approach, POS UAM, PLANR UAM, and UAM(CB-II) results
yield respectively 61, 87, and 46 percent reductions in VOC emissions. The differ-
ences between the POS UAM and PLANR UAM reductions are related to the over-
and underprediction of the observed peak and the fact that urban emissions have less
influence on the PLANR UAM ozone peak because of the model's poor performance
in locating the peak. Since the modeling inputs for the POS UAM and UAM(CB-II)
are very similar, the differences in the levels of VOC emission reductions for the two
models is more difficult to explain. Because of the major differences in the model
inputs (the initial and boundary conditions) for POS UAM and UAM(CB-II) VOC emis-
sion control scenarios, we would expect the POS UAM to calculate that less VOC
emission reductions are needed for attainment than the UAM(CB-II), yet the opposite
is actually true. The differences between the POS UAM and UAM(CB-II) VOC reduc-
tion requirements for ozone attainment are most likely related to the higher over-
prediction of the observed peak in the UAM(CB-II) and the fact that the CB-IV
chemical mechanism tends to estimate larger VOC emission reductions to reduce
ozone concentrations than does CB-II.
The VOC emissions reductions needed to reduce the peak ozone concentration to 12
pphm, as calculated by the POS UAM, PLANR UAM, and UAM(CB-II) and using the
uncorrected, decrement, and percentage approaches to model bias, are given in Table
6-3. When model bias is accounted for, PLANR UAM always calculates larger VOC
89092r2 7
100
-------
TABLE 6-3. Percent VOC emission reduction required to
reduce the calculated region-wide maximum ozone con-
centration to 12 pphm in Philadelphia based on the
POS UAM, PLANR UAM, and UAM(CB-II) applications and
different approaches for correcting model bias.
ApproachPOS UAMPLANR UAMUAM(CB-II)
Unconnected Bias 77 64 79
Decrement 54 >90 36
Percentage 61 87 46
89092rl 3
101
-------
emission reductions to reach attainment than does the POS UAM, whereas the POS
UAM always calculates larger emission reduction retirements than does the
UAM(CB-II). The difference in results for the PLANR UAM and POS UAM is related
to the difference in wind fields. The POS UAM wind field recirculates the urban
emissions from Phildelphia, resulting in a higher predicted ozone peak near down-
town. The PLANR UAM peak occurs further downwind of downtown, where the
emissions are more diffuse and more of the peak is due to background (initial and
boundary) conditions. In addition, the VOC emission reduction scenarios for the POS
UAM also include significant reductions in the initial VOC concentrations, approxi-
mately 35 and 52 percent reductions for the 60 and 90 percent VOC emission control
scenarios respectively. However, because "clean" concentrations were used for the
initial conditions in the PLANR UAM, the initial conditions for PLANR UAM were
not reduced in the VOC emission control scenarios. As noted above, the differences
in the VOC emission reductions for ozone attainment calculated by the POS UAM
and UAM(CB-II) are related to the higher predicted peak in the UAM(CB-II) and the
chemical mechanisms used.
89092r2 7
102
-------
7 SUMMARY AND RECOMMENDATIONS
This study evaluated a simpler approach for the application of the Urban Airshed
Model, the Practice for Low-Cost Airshed-Ppplication for Nonattainment Regions
(PLANR). In this approach only routinely available data are used and a limited num-
ber of diagnostic simulations are performed to establish a base case. The PLANR
UAM approach was used to simulate two ozone episodes: 13 July 1976 in St. Louis
and 13 July 1979 in Philadelphia. Extensive field measurement programs conducted
during these episodes provided rich data bases for evaluating the PLANR UAM per-
formancethe Regional Air Pollution Study (RAPS) in St. Louis and the Philadelphia
Oxidant Study (POS) in Philadelphia. The PLANR UAM simulations were evaluated
several different ways: (1) the predictions were compared with observations from
the rich data base; (2) the model performance was compared with that of the UAM
(CB-II) and UAM (CV-IV) in which the rich data base was used as inputs; and (3) the
calculated reductions in ozone concentrations in response to VOC emission reduc-
tions were compared with those of the UAM (CB-II) and UAM (CV-IV).
ST. LOUIS TEST
The first step in any PLANR use of the UAM is the characterization of the meteoro-
logical conditions that existed during the ozone episode. For St. Louis on 13 July
1976 these conditions included the passage of a warm front through the area and a
corresponding passage of a high-pressure system to the north that caused a wind shift
at around noon.
For the St. Louis episode of 13 July 1976 three diagnostic simulations were necessary
before the model sufficiently replicated the peak ozone observations. These diag-
nostic simulations differed in the interpretation and use of the routine meteorologi-
cal data for input. The first diagnostic simulation used the observed upper-air sound-
ing from a site west of St. Louis; although this site was farther away than the moni-
toring site to the east, it was on the same side of the warm front during the after-
noon. The second diagnostic simulation used upper-air observations from the site
east of St. Louis for the morning, and the soundings from the west site for the after-
noon. For the third diagnostic simulation, wind observations at FAA sites were used
but reduced 50 percent. During conditions typically associated with ozone episodes
(low wind speeds), hourly average wind speeds may be as low as 50 percent of those
reported at FAA stations.
89092r2 8
103
-------
The performance of the PLANR use of the UAM (inputs prepared from routine obser-
vations only) was compared with that of the conventional use of the UAM (inputs
prepared from a rich data base). In this comparison the old version of the UAM (the
UAM(CB-II) and its preprocessors) and the latest version (the UAM(CB-IV) and its
preprocessors) were subjected to a comprehensive performance evaluation using
observations from the extensive data in the RAPS data base. In addition, the UAM
was exercised using a very simplified data base (SIMPLE UAM), i.e., using meteoro-
logical conditions from a single surface site and mixing heights from the nearest
upper-air site. The results of these comparisons are as follows. In a comparison of
the peak predicted and observed ozone concentrations unmatched by time or loca-
tion, the UAM(CB-II) and the RAPS, PLANR, and SIMPLE UAM calculated the peak
observation to within, respectively, 22, 9, 12, and 47 percent of the observed peak.
Thus, the UAM(CB-IV) performed better when an extensive data base was used
(RAPS UAM) than when only routinely available data were used (PLANR UAM).
However, even though it used only routinely available data, the PLANR UAM (with
CB-IV) performed better than the UAM (CB-II), which used the extensive data base.
The use of the UAM with very simple inputs and no diagnostic simulations (SIMPLE
UAM) resulted in very poor model performance.
It should be noted that the PLANR UAM's underprediction of the peak observed
ozone by 12 percent may be due entirely to deficiencies in the 1976 emission inven-
tory used. Since that time several sources of previously uninventoried hydrocarbon
emissions (mobile source running loss emissions, temperature effects on evaporative
emissions, previously unaccounted for VOC emissions, biogenic emissions, etc.) have
been quantified that would substantially increase the predicted ozone concentrations.
The RAPS, PLANR, and SIMPLE UAM were also exercised for a series of VOC emis-
sion control scenarios to determine what effects rich versus sparse inputs have on
calculated ozone reductions in reponse to VOC emission reductions. Because the
maximum ozone concentrations predicted by the model did not exactly match the
observed peaks, several methods of correcting model bias were also studied. When
correcting for model bias, the PLANR UAM calculated 5 to 18 percent less VOC
emission reductions to reach attainment of the ozone NAAQS than did the RAPS
UAM; the SIMPLE UAM calculated that from -8 to over 22 percent more VOC emis-
sion reductions to reach attainment than did the RAPS UAM.
Three different methods of treating model bias were analyzed. For the RAPS UAM
calculations the results of the three methods varied only 5 percentage points (78 to
53 percent VOC emissions reductions needed to achieve attainment of the ozone
NAAQS). For the PLANR UAM calculations the results of the methods varied 9
points (64 to 73 percent), and for the SIMPLE UAM they varied over 26 points (74 to
> 100 percent). Clearly, the RAPS and PLANR UAM are more robust tools for calcu-
lating the emission reductions needed to attain the ozone NAAQS than is the SIMPLE
UAM. When UAM model performance is extremely poor, as in the SIMPLE UAM, the
model should not be used for emission control strategies.
89092r2 8
104
-------
PHILADELPHIA TEST
The Philadlephia ozone episode of 13 July 1979 was characterized by hot stagnant
conditions. Winds were light and appeared to advect the urban emissions south in the
morning of 13 July and then north back across downtown, resulting in the high ozone
exceedances measured to the northwest of downtown. The occurrence of high mea-
sured ozone concentrations in the morning indicated that there was a significant
reservoir of ozone aloft.
Two diagnostic simulations were performed for Philadelphia: in one the observed
data were used without adjustments, and in the other the observed surface wind
speeds at FAA sites were used but reduced by 50 percent. The results of the two
diagnostic runs were almost identical. The predicted peak ozone concentration was
lower than the observed peak and occurred too far to the east. To obtain better
model performance the wind fields would have to be subjectively altered to retain
the urban emissions in the vicinity of downtown Philadelphia. Since one of the
principal objectives of this study is to evaluate the utility of running the UAM with
objectively prepared inputs from just routine data, the alteration of the wind fields
to predict peak ozone concentrations closer to the observed peak ozone observations
would not be consistent with the PLANR use of the UAM since the sites with the
peak observations were not routine observations. Thus diagnostic run 1 was selected
as the base case because it used the routine data with no adjustments.
The performance of the PLANR UAM application to Philadelphia was compared with
the performance of the UAM when model inputs were prepared from a rich data
base, the Philadelphia Oxidant Study (POS). The applications based on the POS data
used the UAM(CB-II) and UAM(CB-IV), the latter referred to as POS UAM to dis-
tinguish it from the PLANR application of UAM(CB-IV). The region-wide maximum
ozone concentrations calculated by the POS UAM, UAM(CB-II), and PLANR UAM
were within 15, 30, and 9 percent, respectively, of the observed peak. At the loca-
tion of the observed peak the PLANR UAM underpredicted the observation by 46
percent, whereas both the POS UAM and UAM(CB-II) replicated the observed peak to
within 15 percent. In general, the performance of the POS UAM and UAM(CB-II) was
quite good, while the PLANR UAM tended to underpredict the observed daily maxi-
mum ozone concentrations at most sites.
As in the St. Louis applications, the PLANR UAM underprediction of the highest
observed ozone concentrations can be partly attributable to known sources of VOC
emissions that were missing from the 1979 inventory. However, since the PLANR
UAM predicts an elevated ozone cloud approximately W km east of the location of
the highest observations, it appears that the wind field used in the PLANR UAM does
not represent actual flow patterns on 13 July 1979. The PLANR UAM wind field
problem was confirmed by a sensitivity simulation in which the air quality inputs
(initial and boundary conditions) of the POS UAM were used as inputs; the results
very similar to the PLANR UAM base case.
89092r2 8
105
-------
The PLANR UAM, POS UAM, and UAM(CB-II) were exercised for a series of VOC
emission reduction scenarios designed to determine the level of VOC reductions
needed to reduce the peak observation to below the NAAQS. When correcting for
model bias, the PLANR UAM calculates 26 to over 36 percent more VOC emission
reductions than does the POS UAM. The UAM(CB-II) calculated 15 to 18 percent less
VOC emissions reductions than the POS UAM.
Differences in VOC emission reductions calculated by the PLANR and POS UAM can
be attributed to the fact that the POS UAM has a recirculating wind field which
resulted in urban emissions contributing most of the ozone precursors in the vicinity
of the predicted peak, whereas the PLANR UAM predicted peak occurred further
downwind where the initial and boundary conditions had a larger contribution to the
ozone precursors.
Since the POS UAM and UAM(CB-II) used almost identical inputs, the differences in
their calculations of the VOC emission reductions needed to demonstrate attainment
are more difficult to explain. It appears that the VOC emission reductions calcula-
ted by UAM(CB-II) are lower because the model predicts a higher peak in the base
case and because the CB-II chemical mechanism tends to calculate lower VOC emis-
sion reductions needed for attainment than does the CB-IV.
Because of the large amounts of transported pollutants known to exist in the Phila-
delphia region and the complex meteorological conditions (stagnation with wind
shear) that existed on 13 July 1979, this episode is most likely not an appropriate
choice for evaluating the PLANR use of the UAM. Extensive data bases or regional
modeling studies are needed to characterize transport in the northeastern U.S.
Fortunately the EPA is currently conducting such a modeling study, the Regional
Oxidant Modeling for Northeast Transport (ROMNET) program, in which a regional
oxidant model is used to define boundary conditions for the UAM in cities in the
northeast.
RECOMMENDATIONS
The evaluation of the PLANR use of the UAM to simulate two historical episodes in
St. Louis and Philadelphia was inconclusive because of known deficiences in the
emission inventories used and because the model applications deviated slightly from
the PLANR use of the UAM (i.e., nonroutine data had to be used for model evalua-
tion). In the PLANR UAM application to Philadelphia routine meteorological
observations were used as is for the base case, and the model exhibited rather poor
performance. Because there were no routine ozone observations in St. Louis during
the modeled episode, ozone observations from a special study were used (for evalua-
tion only) to develop a PLANR UAM base case, which exhibited good model perform-
ance.
89092r2 8
106
-------
The PLANR use of the UAM needs to be tested with additional cities and ozone epi-
sodes. So far, the PLANR use of the UAM has been tested for St. Louis, Philadel-
phia, Atlanta, and Dallas-Fort Worth. However, St. Louis and Philadelphia were the
only areas with the data needed to compare the PLANR UAM with a conventional
UAM application (that is, one based on an extensive data base). The PLANR UAM
also needs to be applied to more recent episodes so that more recent, higher-quality
emission inventories can be used. For the PLANR approach, inputs should be
developed using only routine meteorological and air quality data, and the diagnostic
simulations performed to arrive at a base case should be evaluated using only the
routine air quality data. Once the PLANR UAM base case is developed, special study
air quality data can be used to evaluate the PLANR results. In addition, side-by-side
simulations of an episode using both PLANR and conventional applications of the
UAM would be desirable for evaluating PLANR results.
Of highest priority in the use of the UAM to demonstrate ozone attainment is the
development of high-quality emission inventories for air quality modeling. In many
past UAM modeling studies meteorological inputs were altered to get model predic-
tions to match the observations. Because past emission inventories are known to
understate the amount of VOC emissions, these alterations may have been unwarran-
ted. Although much progress has been made over the last 10 years in the develop-
ment of higher-quality emission inventories, further progress is needed before a
comprehensive emission control strategy to eliminate the ozone problem can be
developed.
89092r2 8
107
-------
References
Allard, D., M. Chan, C. Marlia, and E. Stephens. 1981. Philadelphia Oxidant Data
Enhancement StudyAnalysis and Interpretation of Measured Data. U.S.
Environmental Protection Agency (EPA-450/4-81-011).
Ames, J., T. C. Myers, L. E. Reid, D. C. Whitney, S. H. Golding, S. R. Hayes, and
S. D. Reynolds. 1985a. SAI Airshed Model Operations Manuals. Volume I
User's Manual. U.S. Environmental Protection Agency (EPA-600/8-85-007a).
Ames, 3., S. R. Hayes, T. C. Myers, and D. C. Whitney. 1985b. SAI Airshed Model
Operations Manuals. Volume IlSystem's Manual. U.S. Environmental Protec-
tion Agency (EPA-600/8-85-007b).
Bass, A. M., L. C. Glasgow, C. Miller, 3. P. 3esson, and D. L. Filken. 1980. Planet
Space Sci., 28:675.
Boris, 3. P., and D. L. Book. 1973. Flux-corrected transport: I. SHASTA, a fluid
transport algorithm that works. 3. Comp. Phys., 11:38-69.
Burton, C. S. 1988. Comments on "Ozone Air Quality Models." 3. Air Pollut. Con-
trol Assoc., 38(9): 1119-1128.
Chameides, W. L., R. W. Lindsay, J. Richardson, and C. S. Kiang. 1988. The role of
biogenic hydrocarbons in urban photochemical smog: Atlanta as a case study.
Science, 241:1473-1475.
Chock, D. P., and A. M. Dunker. 1983. A comparison of numerical methods for solv-
ing the advection equation. Atmos. Environ., 17:11-24.
Chock, D. P. 1985. A comparison of numerical methods for solving the advection
equationII. Atmos. Environ., 19:571-586.
Cole, H. S., D. E. Layland, G. K. Moss, and C. F. Newberry. 1983. The St. Louis
Ozone Modeling Project. U.S. Environmental Protection Agency (EPA-450/4-
83-019).
Douglas, S., and R. Kessler. 1988. "User's Guide to the Diagnostic Wind Model. Ver-
sion 1.0." Systems Applications, Inc., San Rafael, California.
89092rl i»
ins
-------
EPA. 1982. Emissions Inventories for Urban Airshed Model Application in the Phila-
delphia AQCR. U.S. Environmental Protection Agency (EPA-450/4-82-005).
EPA. 1986. Guideline on Air Quality Models (Revised). U.S. Environmental Protec-
tion Agency (EPA-450/2-78-027R).
Gery, M. W., G. Z. Whitten, and 3. P. Killus. 1988. "Development and Testing of the
CBM-IV for Urban and Regional Modeling." Systems Applications, Inc., San
Rafael, California (SYSAPP-88/002).
Godden, D., and F. Lurmann. 1983. "Development of the PLMSTAR Model and Its
Application to Ozone Episode Conditions in the South Coast Air Basin."
Environmental Research & Technology, Inc., Westlake Village, California (ERT
P-A702-200).
Haney, 3. L., and T. N. Braverman. 1985. Evaluation and Application of the Urban
Airshed Model in the Philadelphia Air Quality Control Region. U.S. Environ-
mental Protection Agency (EPA-450/4-85-993).
Haney, 3. L., and C. S. Burton. 1988. "UAM-CMB-IV-Calculated Effects of Emission
Reductions on Ozone Levels in the Philadelphia Metropolitan Area." Systems
Applications, Inc., San Rafael, California (SYSAPP-88/117).
Haney, 3. L., D. R. Souten, T. W. Tesche, L. R. Chinkin, H. Hogo, and M. C. Dudik.
1986. "Evaluation and Applications of the PARIS Photochemical Model in the
South Central Coast Air Basin." Volume I. Systems Applications, Inc., San
Rafael, California (SYSAPP-86/065).
Hogo, H., L. A. Mahoney, and M. A. Yocke. 1988. "Draft Air Quality Management
Plan 1988 Revision. Draft Appendix V-R. Urban Airshed Model Performance
Evaluation for 5-7 3une 1985." Systems Applications, Inc., San Rafael, Califor-
nia (SYSAPP-88/138).
Layland, D. E., and H. S. Cole. 1983. A Review of Recent Applications of the SAI
Urban Airshed Model. U.S. Environmental Protection Agency (EPA-450/4-84-
004).
Morris, R. E., M. W. Gery, M. K. Liu, G. E. Moore, C. Daly, and S. M. Greenfield.
1989d. "Sensitivity of a Regional Oxidant Model to Variations in Climate
Parameters. Volume I: Results." Systems Applications, Inc., San Rafael, Cali-
fornia (SYSAPP-89/014a).
89092rl if
109
-------
Morris, R. E., R. C. Kessler, S. G. Douglas, K. R. Styles, and G. E. Moore. 1988.
"Rocky Mountain Acid Deposition Model Assessment Acid Rain Mountain Meso-
scale Model (ARM3)." Systems Applications, Inc., San Rafael, California
(SYSAPP-88/152).
Morris, R. E., T. C. Myers, E. L. Carr, and M. C. Causley. 1989c. "Urban Airshed
Model Study of Five Cities. Demonstration of Low-Cost Application of the
Model to the City of Atlanta and the Dallas-Fort Worth Metroplex Regions."
Systems Applications, Inc., San Rafael, California (SYSAPP-89/122).
Morris, R. E., T. C. Myers, M. C. Causley, and L. Gardner. 1989b. "Low-Cost Appli-
cation of the Urban Airshed Model to Atlanta and Evaluation of the Effects of
Biogenic Emissions on Emission Control Strategies." Systems Applications,
Inc., San Rafael, California.
Morris, R. E., T. C. Myers, H. Hogo, L. R. Chinkin, L. A. Gardner, and R. G.
Johnson. 1989a. "A Low-Cost Application of the Urban Airshed Model to the
New York Metropolitan Area and the City of St. Louis." Systems Applications,
Inc., San Rafael, California (SYSAPP-89/038).
O'Brien, J. 3. 1970. A note on the vertical structure of the eddy exchange coef-
ficient in the planetary boundary layer. J. Atmos. Sci., 27:1213-1215.
OTA. 1988a. "Urban Ozone and the Clean Air Act: Problems and Proposals for
Change." Office of Technology Assessment, Washington, D.C.
OTA. 1988b. "Ozone and the Clean Air Act: Summary of OTA Workshop with State
and Local Air Pollution Control Agency Officials." Office of Technology
Assessment, Washington, D.C.
OTA. 1988c. "Ozone and the Clean Air Act: A Summary of OTA Workshops on
Congressional Options to Address Nonattainment of the Ozone Standard."
Office of Technology Assessment, Washington, D.C.
Rao, S. T. 1987. Application of the Urban Airshed Model to the New York Metro-
politan Area. U.S. Environmental Protection Agency (EPA-450/4-87-011).
Ross, D. G., and I. Smith. 1986. "Diagnostic Wind Field Modeling for Complex Ter-
rainTesting and Evaluation." Centre for Applied Mathematical Modeling,
Chisholm Institute of Technology (CAMM Report No. 5/86).
SAL 1989. "User's Manual for Preparing Emission Files for Use in the Urban Airshed
Model." Systems Applications, Inc., San Rafael, California (SYSAPP-89/114).
Schere, K. L. 1983. An evaluation of several numerical advection schemes. Atmos.
Environ., 17:1897-1907.
89092rl if
nn
-------
Schere, K. L., and K. L. Demerjian. 1977. Calculation of Selected Photolytic Rate
Constants over a Diurnal Range. A Computer Algorithm. U.S. Environmental
Protection Agency (EPA-600/4-77-015).
Schere, K. L., and 3. H. Shreffler. 1982. Final Evaluation of Urban-Scale Photo-
chemical Air Quality Simulation Models. U.S. Environmental Protection
Agency (EPA-600/3-82-094).
Science. 1988. Rural and urban ozone. Editorial in Science, 241(4873); 1569.
Seinfeld, 3. H. 1988a. Ozone air quality models. A critical review. 3. Air Pollut.
Control Assoc., 38(5):616.
Seinfeld, 3. H. 1988b. Closing remarks. 3. Air Pollut. Control Assoc., 38(8);1136-
1137.
Smolarkiewicz, P. K. 1983. A simple positive definite advection scheme with small
implicit diffusion. Monthly Weather Review, 111:479-486.
Whitten, G. Z., 3. P. Killus, and H. Hogo. 1980. Modeling of Simulated Photochemi-
cal Smog with Kinetic Mechanisms. Volume 1. U.S. Environmental Protection
Agency (EPA-600/3-80-028a).
Whitten, G. Z., T. C. Meyers, C. Daly, L. R. Chinkin, S. D. Reynolds, N. M. Yonkow,
and B. Austin. 1985. "Application of the Urban Airshed Model to Kern
County." Systems Applications, Inc., San Rafael, California (SYSAPP-85/200).
Yamartino, R. 3., and 3. S. Scire. 1984. "ADOM/TADAP Model Development Pro-
gram. Volume 3. The Transport and Diffusion Modules." Environmental
Research & Technology, Inc., Concord, Massachusetts (P-B980-210).
Yocke, M. A., R. E. Morris, H. Hogo, L. R. Chinkin, and L. A. Mahoney. 1985.
"Analysis of the Air Quality Impacts of the San Miguel Project. Volume I."
Systems Applications, Inc., San Rafael, California (SYSAPP-85/127).
Zalesak, S. T. 1979. Fully multi-dimensional flux-corrected transport algorithms for
fluids. 3. Comput. Phys., 31:335-362.
89092rl i»
111
-------
Appendix A
MODEL PERFORMANCE STATISTICS FOR HOURLY AND DAILY
MAXIMUM OZONE CONCENTRATIONS FOR THE RAPS, PLANR, AND
SIMPLE UAM APPLICATIONS TO ST. LOUIS
89092rl
-------
Appendix
Scatterplots, residual analysis plots, and model performance statistics for:
A-l RAPS UAM hourly ozone concentrations
A-2 RAPS UAM daily maximum ozone concentrations
A-3 PLANR UAM hourly ozone concentrations
A-4 PLANR UAM daily maximum ozone concentration
A-5 SIMPLE UAM hourly ozone concentrations
A-6 SIMPLE UAM daily maximum ozone concentrations
89092 1
-------
VARIABLE...OZONE
BEGINNING DATE. . .8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYS I S...A IRS DATA
THE PREDICTION ALGORITHM...UAM(CB-I V) OZONE
AVERAGING TIME. . .ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE...265
20.00
L5.00
c.
Q.
Q.
10. 00
LJLJ
IX
CL.
5.00
1 I I I I I I I I I I I I I | I I I I \ I I I F
x
r
I I I I I I I I
I I I I I I I I I
5.00
10.00
OBSERVED
15.00
20.00
(pphm)
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 6.77843 7.51675
STANDARD DEVIATION 4.85895 4.99583
SKEWNESS 0.68694 0.69536
ICURTOSIS -0.47257 -0.01638
OTHER MEASURES
MEDIAN 4.90000 6.79000
UPPER QUARTILE 10.60000 10.57000
LOWER QUARTILE 2.90000 3.42000
MINIMUM VALUE 0.20000 0.11000
MAXIMUM VALUE 22.20000 23.37000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.914
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.891 HIGH BOUND 0.932
RATIO OF OVER TO UNDER PREDICTIONS 2.442
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 6.792
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 4.906
FIGURE A-l. RAPS UAM hourly ozone concentrations.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGOR ITHM...UAM(CB-I V) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE...265
0.40 -
-6.50 -3.90 -1.30
RESIDUAL (OBS-PRED:
THE BINS1ZE EQUALS 1.300
1.30
3.90
6.50
RESIDUAL ANALYSIS
AVERAGE -0.73835
STANDARD DEVIATION 2.05125
SKEWMESS 0.40933
KURTOSIS 2.00005
OTHER MEASURES
MEDIAN -0.88000
UPPER QUARTILE 0.25000
LOWER QUARTILE -1.92000
MINIMUM VALUE -8.62000
MAXIMUM VALUE 7.36000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -1.4451
UPPER BOUND -0.0316
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 3.6681
UPPER BOUND 4.8841
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 2.18
THE AVERAGE ABSOLUTE ERROR IS 1.67
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7168
RESIDUAL COEFFICIENT OF VARIATION
0.3026
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.4222
FIGURE A-l. Concluded.
89092
-------
VARIABLE...OZONE
BEGINNING DATE. . .8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE ... OZONE
OBSERVATION ANALYS I S . ..A I RS DATA
THE PREDICTION ALGOR ITHM...UAM(CB-I V) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE. . . 14
20.00
15.00
Q.
Q.
10.00
Q
LU
CC
5.00
I I I i I I I I I I I I I I I I I I I I I I I I
5.00 10.00 15.00
OBSERVED (pphm)
I I I I I I I I I I I I I I
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 15.01427 15.96999
STANDARD DEVIATION 2.65243 4.18404
SKEWNESS 1.10826 0.43840
KURTOSIS 1.55511 -1.30929
OTHER MEASURES
MEDIAN 14.50000 15.06000
UPPER QUARTILE 15.30000 17.39000
LOWER OUARTILE 13.70000 12.02000
MINIMUM VALUE 10.60000 10.87000
MAXIMUM VALUE 22.20000 23.37000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.676
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.227 HIGH BOUND 0.888
RATIO OF OVER TO UNDER PREDICTIONS 1.000
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 0.000
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 0.000
FIGURE A-2. RAPS UAM daily naxLnum ozone concentrations.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE. . .8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGORITHM..,UAM(CB-I V) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE... 14
0.40
0.30
0.20
CO
OL
E o. 10
o
I
-6.00 -3.60 -1.20
RESIDUAL (OBS-PRED)
THE BINSIZE EQUALS 1.200
1.20
3.60
6.00
RESIDUAL ANALYSIS
AVERAGE -0.95571
STANDARD DEVIATION 3.08663
SKEWNESS -0.71570
KURTOSIS -0.38413
OTHER MEASURES
MEDIAN -1.42000
UPPER QUARTILE 0.26000
LOWER QUARTILE -3.43000
MINIMUM VALUE -7.92000
MAXIMUM VALUE 3.08000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -3.2883
UPPER BOUND 1.3769
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 5.6337
UPPER BOUND 20.3216
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 3.12
THE AVERAGE ABSOLUTE ERROR IS 2.36
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.1767
RESIDUAL COEFFICIENT OF VARIATION
0.2056
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
1.1637
FIGURE A-2. Concluded.
-------
VARIABLE...OZONE
BEGINNING DATE.. .8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE...OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGORITHM...PLANR UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA SITES
SAMPLE SIZE...265
20.00 -
15.00 -
e
C.
a.
a.
o
LU
<-> 10.00 -
ce
a.
5.00 -
i I i I I I I i I I I I I I I I I I I I I
5.00 10.00 15.00 20.00
OBSERVED (pphm)
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 6.77843 5.35212
STANDARD DEVIATION 4.85895 4.62851
SKEWNESS 0.68694 0.80435
KURTOSIS -0.47257 -0.04013
OTHER MEASURES
MEDIAN 4.90000 4.30000
UPPER QUARTILE 10.60000 8.30000
LOWER QUARTILE 2.90000 1.32000
MINIMUM VALUE 0.20000 0.01000
MAXIMUM VALUE 22.20000 18.87000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.895
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.868 HIGH BOUND 0.916
RATIO OF OVER TO UNDER PREDICTIONS 0.221
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 2.642
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 24.906
FIGURE A-3. PLANR UAM hourly ozone concentrations.
-------
VARIABLE. ..OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE ... OZONE
OBSERVATION ANALYSIS...A IRS DATA
THE PREDICTION ALGOR I THM...PLANR UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VARI ABLE...RAPS DATA SITES
SAMPLE SIZE...265
0.40 -
-6.00 -3.60 -1.20
RESIDUAL (OBS-PRED)
THE BINSIZE EQUALS 1.200
1.20
3.60
6.00
RESIDUAL ANALYSIS
AVERAGE 1.42629
STANDARD DEVIATION 2.18869
SKEWNESS -0.47534
KURTOSIS 1.75804
OTHER MEASURES
MEDIAN 1.37000
UPPER QUARTILE 2.69000
LOWER QUARTILE 0.20000
MINIMUM VALUE -6.86000
MAXIMUM VALUE 7.80000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 0.7457
UPPER BOUND 2. 1069
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 4.2761
UPPER BOUND 5.5605
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 2.61
THE AVERAGE ABSOLUTE ERROR IS 2.02
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7168
RESIDUAL COEFFICIENT OF VARIATION
0.3229
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.4504
FIGURE A-3. Concluded.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYSIS...A IRS DATA
THE PREDICTION ALGOR ITHM...PLANR UAM(CB-IV)
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA SITES
SAMPLE SIZE... 14
OZONE
20.00
15.00
c.
Q.
Q.
10. 00
UJ
X.
Q_
5.00
I 1 T I
T I
1 [ I I [ I | I I I I
I I I I I I I I I I I I I I I I I I I I I I I I
5.00 10.00 15.00
OBSERVED (pphra)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 15.01427 12.26499
STANDARD DEVIATION 2.65243 3.75908
SKEWNESS 1.10826 0.30977
KURTDSIS 1.55511 -1.27994
OTHER MEASURES
MEDIAN 14.50000 11.69000
UPPER QUARTILE 15.30000 14.15000
LOWER QUARTILE 13.70000 9.14000
MINIMUM VALUE 10.60000 6.93000
MAXIMUM VALUE 22.20000 18.87000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.550
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.027 HIGH BOUND 0.837
RATIO OF OVER TO UNDER PREDICTIONS 0.167
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 0.000
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 0.000
FIGURE A-4. PLANR UAM daily iraxLirum ozone concentrations.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE ... OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGOR I THM...PLANR UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA SITES
SAMPLE SIZE...14
0.40
0.30 -
0.20
CD
cc
. 10
I
-5.00 -3.00 -1.00
RESIDUAL (OBS-PRED)
THE BINSIZE EQUALS 1.000
1.00
3.00
5.00
RESIDUAL ANALYSIS
AVERAGE 2.74929
STANDARD DEVIATION 3.19348
SKEWNESS -1.13820
KURTOSIS 0.34709
OTHER MEASURES
MEDIAN 3.53000
UPPER QUARTILE 3.75000
LOWER QUARTILE 0.72000
MINIMUM VALUE -4.97000
MAXIMUM VALUE 6.64000
FIGURE A-4. Concluded.
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 0.5831
UPPER BOUND 4.9155
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 6.0305
UPPER BOUND 21.7528
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 4.13
THE AVERAGE ABSOLUTE ERROR IS 3.85
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.1767
RESIDUAL COEFFICIENT OF VARIATION
0.2127
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
1.2040
89092
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYS I S...A I RS DATA
THE PREDICTION ALGOR I THM...S IMPLE UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VARI ABLE...RAPS DATA SITES
SAMPLE SIZE...265
20.00 -
5.00 10.00 15.00
OBSERVED (pphra)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 6.77843 4.32029
STANDARD DEVIATION 4.85895 2.23995
SKEWNESS 0.68694 0.33689
KURTOSIS -0.47257 0.12427
OTHER MEASURES
MEDIAN 4.90000 4.20000
UPPER QUARTILE 10.60000 5.45000
LOWER QUARTILE 2.90000 2.95000
MINIMUM VALUE 0.20000 0.01000
MAXIMUM VALUE 22.20000 10.56000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.793
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.743 HIGH BOUND 0.833
RATIO OF OVER TO UNDER PREDICTIONS 0.318
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 6.038
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 27.170
FIGURE A-5. SIMPLE UAM hourly ozone concentrations.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYS IS . . . A IRS DATA
THE PREDICTION ALGOR I THM.. . S IMPLE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA
SAMPLE SIZE...265
UAM(CB-IV) OZONE
SITES
0.40 -
o
-10.50 -6.30 -2.10
RESIDUAL (OBS-PRED1
THE BINSIZE EQUALS 2.100
2. 10
6.30
10.50
RESIDUAL ANALYSIS
AVERAGE 2.45812
STANDARD DEVIATION 3.37279
SKEWNESS 1.01240
KURTOSIS 0.26181
OTHER MEASURES
MEDIAN 1.10000
UPPER QUARTILE 4.35000
LOWER QUARTILE 0.01000
MINIMUM VALUE -2.31000
MAXIMUM VALUE 14.57000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 1.9155
UPPER BOUND 3.0008
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 9.9170
UPPER BOUND 13.2046
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 4.17
THE AVERAGE ABSOLUTE ERROR IS 2.85
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7168
RESIDUAL COEFFICIENT OF VARIATION
0.4976
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.6941
FIGURE A-5. Concluded.
-------
VARIABLE...OZONE
BEGINNING DATE...8/8/80
ENDING DATE...8/8/80
OBSERVATION SOURCE ... OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGOR I THM...S IMPLE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA
SAMPLE SIZE... 14
UAM(CB-IV) OZONE
SITES
20.00
15.00
Q.
a.
10.00
o
LJJ
5.00
1 I I I
i i r
i i i
x
\.
I I I I I I I
I I I I I I
I I I I
5.00 10.00 15.00
OBSERVED Ipphm)
20.00
MOMENTS OF THE PROBABILITY DENSITY FUNCTION
OBSERVED PREDICTED
AVERAGE 15.01427 8.17999
STANDARD DEVIATION 2.65243 1.58534
SKEWNESS 1.10826 0.20928
KURTOS1S 1.55511 -1.54830
OTHER MEASURES
MEDIAN 14.50000 7.63000
UPPER QUART1LE 15.30000 9.19000
LOWER QUARTILE 13.70000 6.71000
MINIMUM VALUE 10.60000 5.85000
MAXIMUM VALUE 22.20000 10.56000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.399
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND -0.371 HIGH BOUND 0.660
RATIO OF OVER TO UNDER PREDICTIONS 0.000
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 0.000
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 42.857
FIGURE A-6. SIMPLE UAM daily naxiittun ozone concentrations.
-------
VARIABLE. . .OZONE
BEGINNING DATE...8/8/80
ENDING DATE. . .8/8/80
OBSERVATION SOURCE... OZONE
OBSERVATION ANALYSIS...A IRS DATA
THE PREDICTION ALGORITHM...S IMPLE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...RAPS DATA
SAMPLE SIZE. . . 14
UAM(CB-IV) OZONE
SITES
0.40
0.30
0.20
CD
cr
0. 10
-9.00 -5.40 -1.80
RESIDUAL (08S-PRED)
THE BINSIZE EQUALS 1.800
1.80
5.40
9.00
RESIDUAL ANALYSIS
AVERAGE 6.83428
STANDARD DEVIATION 2.80639
SKEWNESS -0.04242
KURTOSIS -1.02145
OTHER MEASURES
MEDIAN 7.08000
UPPER QUARTILE 8.59000
LOWER QUARTILE 3.34000
MINIMUM VALUE 2.34000
MAXIMUM VALUE 12.25000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 5.3793
UPPER BOUND 8.2893
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 4.6571
UPPER BOUND 16.7990
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 7.35
THE AVERAGE ABSOLUTE ERROR IS 6.83
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.1767
RESIDUAL COEFFICIENT OF VARIATION
0.1869
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
1.0580
FIGURE A-6. Concluded.
-------
Appendix B
WIND FIELDS USED IN DIAGNOSTIC RUN 1
APPLICATION TO PHILADELPHIA
89092rl
-------
387
437
487
Wind Speed (m/s)
0 5 10
H
537
4490
4440
-4390
10
20
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 7 on 84194
-------
387
437
487
Wind Speed (m/s)
0 5 10
liii.lii.il
537
*'«-«- ^rr» ' V" "'
»-»-»-^-«-»-»"W*V^»
4490
4440
4390
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 8 on 84194
-------
Wind Speed (m/s)
0 5 10
387
437
487
537
4490
r 4440
- 4390
10
20
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 9 on 84194
-------
Wind Speed (m/s)
0 5 10
487
537
4490
4440
- 4390
30
4340
Philadelphia PLANE Winds Eval # 1
Layer 1 at hour 10 on 84194
-------
387
437
487
Wind Speed (m/s)
0 5 10
537
4490
4440
H4390
10
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 11 on 84194
-------
Wind Speed (m/s)
0 5 10
387
487
537
\\\
'ft t f-
4490
4440
4390
0
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 12 on 84194
-------
Wind Speed (m/s)
0 5 10
38(7
r 4490
4440
4390
10
20
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 13 on 84194
-------
Wind Speed (m/s>
0 5 10
I 4490
1* 4440
- 4390
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 14 on 84194
-------
Wind Speed (m/s)
0 5 10
4490
4440
- 4390
30
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 15 on 84194
-------
Wind Speed (m/s)
0 5 10
4490
4440
- 4390
0
10
30
4340
Philadelphia PLANE Winds Eval # 1
Layer 1 at hour 16 on 84194
-------
Wind Speed (m/s)
7 7 7
m
v/M
- 4490
" 4440
' # 4390
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 17 on 84194
-------
Wind Speed (m/s)
0 5 10
387
4490
4440
4390
4340
Philadelphia PLANR Winds Eval # 1
Layer 1 at hour 18 on 84194
-------
Appendix C
HOURLY OZONE CONCENTRATIONS (pphm) IN PHILADELPHIA
PREDICTED FROM DIAGNOSTIC RUN I
89092rl
-------
1SV3
I s
If
S :§
2 5
in
O
O
o
O
O
o
in
_i j I I i
o
o
03
t
O
O
*
to
I IJTI I If j Til
I t { I I I
O
rO
O
r-J
O
(ft
1 i I I I i I 1. I 1 I I { I
I I I I
o
to
o
(N
1S3M
$.8
t E
li
o c
2 2
o o o o o o o
m
_i j t i i | f l l | t i 4 j i i i j i f i j { i { | i l f j l i i_
o
&
Of.
o
0)
.E
o
rO
s.
O
o '
'l } ( I 1 I t I l I i 1 I I I I I I I i I I t I i l
! 1 i { 1 i !
O
rO
O
(N
05
r^
01
00
c
o
"(/)
w
"E
(D
(D
W
O
-O
O)
r^
CT)
a
a
(D
o
N
O
1S3M
-------
00
O
O
II
O
O
o
o
rO
o
o
IT)
CM
IT)
o
n
ce
O O
03
O
o
irt
o
33
O
*
Tf
1S3M
1.SJ3
o
tsl
o
o
o
03
o
(O
o
*
10
m JT
CN
irt
(/I
O
O
8
o
10
8
O
cn
r-,
en
3
~5
CO
c
_o
.w
0)
(t)
w
o
X)
O)
CD
a.
CL
3)
o
N
O
03
*f t t ! t i i i t > i i i I i i > i i't i; i ') I i i i:: i i t i } i*
o
rO
O
CM
1S3M
-------
I
x
o
2
SE
*
o
o
ID
o o
O 03
tO *
1SV3
o
eg
t
o
00
O O
(D *
IO 10
_l J I I I | I I I j 1 I _L 1 I i I j ! I I I I i i I f It I I I f_
O
CM
o
1S3M
§
03
0)
J3
o
£
3
E
'c
2
in
ce
o
10
^
o
o
in
o
o
O O O
O 00 tO
55$
1SV3
o o
o
o
o
00
o
(O
o
*f
IO
_i j i i i I i i i I j i ± i i i i j i i i i f i i i i i i j i i r^
'i i i t i i i i i i i i i i i
o o
IO CS
i i i i
i i i i i i i
o
en
(35
O
'(/)
w
'£
w
o
X!
e
x:
a
D.
0)
o
N
O
1S3M
-------
r- o
od ^1
11 ii
* «
$:±
2 2
ao
I-
10
O
o
o
o
iTt
o
CO
t
tr
o
o
to
1SV3
o o
O
O
t
o
00
5
I I I I J. { I i I { I ! I I i I I I i I I I
O
(N
o
K)
o
ro
O
1S3M
I S
II
81
2 2
8
o
o
o
03
O
to
rt-
XSV3
O
IN
>*-
o
o
o
oo
o
-------
1SV3
q
IN
I 5
o
If
J I
s:=
2 2
oo
o
o
in
o
03
o
CM
o o
o m
* f>
* *
O
to
o
*
tO
>t
I j I I I j I I ) I I i J. j I I I I I f I I i I I I 1 I I j I
o:
o
0>
T=
?
I I I j I 1 I 1 1 I 1 I I I I I I t I I I i I I I t 1 I 1 l'
p
X
O
(M
1S3M
1SV3
Ol
» r-
i-
-------
1SV3
o
o
in
o
03
O
to
o
CM
O
o
o
CO
O
t
ro
j i i i i | i i t | i t J. i i
Q)
3
2 5
r-
CO
I *
en
i-
(O
*
o r-
o *
CM
-------
o
O
*~ CN
" II
1 5
I!
s:i
2 2
o o
00
i i i i i
i { i i i t i
i i I t i r i i > i i
ro
O
(N
1S3M
°) 00
CM
<£>
10
o
o
n
p
r~
CM
r^
s
I f I I I I I I i I i ) i I I J I t l_
- §
O
ro
O
CM
O
to
r-
03
f>
'l I 1 1 i I 1 I I I I I I i I 1 t i I i j I t I i I } I I I I I
O O O
rO (M r-
1S3M
05
r^
en
M
C
O
"w
w
"E
tu
r--
cn
QL
Q.
C
O
N
O
-------
1SV3
tO
I!
10
_J
O
O
to
£
P
o
o
in
o
03
o
CM
O
O
*
o
00
5
o
(O
to
o
^-
to
_l j t 1 1 I I 1 I I I I J. I 1 I I I I I I I I I I I I I I { I I l_
00
I *-
o:
o
2 r-
§
I
I
'l I i I 1 I i 1 1 I t t I i I I I
I i I t I 1 I i I
O
CM
J.S3M
1SV3
> I "I ..I i 1 I I It t i I t I i t I '{ l'
O
CM
CD
r-v
O)
3
>
n
c
o
'w
to
'E
tu
D
JD
CD
E
.c
Q.
CL
0)
O
N
O
1S3M
-------
J 5
II
to
o
o
o
o
IT)
o
CD
1SV3
o o
o o
O 03
^- rO
* *
K)
_l j t I i j ! 1 I | J I _L j I I I | I » I | ( I J | i 1 I
r-
It O
E *
p
2
CO
o
f
to
I I I I t I 1 I 1 I I I I i 1 I I I I t I I I I } I 1 I t I I i i
1S3M
(O
» m
03 r-
I s
it
X =
o .£
S S
o
O
o
03
o
to
o
O
o
03
O
(O
X
o:
o
O
8
ID
E
l l
I I 1 I t I t I I t _L I I I I I I t ! I t ! i I t I i I 1
1-
o
o
CO
i < i ) i i i i I i l i i : l l l I l I t
I ; I i I
£
I
O
eg
CJ)
rv
CD
M
O
E
i:
a
a
c
o
N
O
1S3M
-------
Ji
x =
o .E
2 2
cc
o
O
O
CM
8
0)
£
f=
o
o
o
03
O
r*
to
in
O
O
o
00
r»
<»
m
r~
CM
m
r-
o
in
r--
CO
rv
CM
*
r~~
o
r~
oo
ro
^*^?t-
tt^/.EE^-
%&' ~*3
iu ^.
ex
d
II I
I I t I I I t I I I I t I i I 1 t I 1 I 1 1 I I ! 1 I I I
O
CO
o
CM
1S3M
1SV3
03
in 03
*~ <*
II II
8 o
J -.
Ji
SI
r~-
ID
m
en
o
O
O
O
CO
o
o
m
o
00
o
<£>
O
o
CM
O
O
o
OD
O
(O
IO
^1 J t I J | I I I | t I ^ | I ? I | I J t ] 5 I J | t I I j I I l__
r^
*
m
i^~
CM
m
r~-
o
in
t^
03
1^-
O
03
ro
'l 1 I I I I I I 1 I t 1 I 1 I i I i 1 1 i I 1 I j I I 1 I I I
o
CM
r-
cn
n
C
O
w
f
w
o
CD
P--
CD
Q.
a
c
o
N
O
1S3M
-------
Appendix D
TIME SERIES OF PREDICTED AND OBSERVED HOURLY NO, NO2, AND CO
CONCENTRATIONS IN PHILADELPHIA FOR DIAGNOSTIC RUN 1
89092rl
-------
20
- N
a 10
o
z
0
3
1
-
-
-
-/
3
*>'
I10
O
0
)
1
-
1
3
6 12 18 2
I i i i | i l l i i | i i i > i | i l i i i
AMS LAB m ~
OBSERVED [D -
PREDICTED
-
/t
/E EH
6 12 18 2
420 20
S
I
10 £ 10
o
0 0
4 (
3 6 12 18 2
1 1 1 1 1 | 1 1 1 1 1 | 1 1 ! 1 1 | 1 1 I 1 1
' ANCORA m ~
OBSERVED E -
PREDICTED
-
-
-
", , , , ,-Lt . .-L. ,l,,,,,'
3 6 12 18 2
420
10
0
4
TIME (HOURS) TIME (HOURS)
6 12 18 2
1 I l l | l l I l l | I I I I l | l 1 l l I
BRIGANTINE
OBSERVED E -
PREDICTED
-
1 1 1 1 1 1 1 1 1 ! 1 1 1 1 1 1 1 1 1 1 1 1
6 12 18 2
4 (
20 20
10 £ 10
0
0 0
4 {
3 6 12 18 2
l i i l l | l l i l i | i l i l l | l l l l l
" BRISTOL
OBSERVED [JJ -
PREDICTED
-
-
3 6 12 18 2
4
20
10
0
4
TIME (HOURS) TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - PLANR RUN
SYSTEMS APPLICATIONS. INC.
-------
20
12
18
24
1 I I I r
CAMDEN
OBSERVED CD
PREDICTED
8- 10
20 20
12 18
24
I
10 9- 10
i I I I I | r
' CHESTER
OBSERVED CO -
PREDICTED
20
10
6 12 18
TIME (HOURS)
6 12 18
TIME (HOURS)
24
20
12
18
I I T I T | I I I
CLAY MO NT
2420
10
o
OBSERVED [JJ -
PREDICTED
i i i i I i i i I i
12 18
TIME (HOURS)
20
12
18 24
10 10
o
z
24°
CONSHOHOCKEN
OBSERVED UJ
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - NO - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
12
18 24
zu
^
I 10
O
z
n
I I I I I | I I I I I | I I I I I | I I I I I
I DEFENSE SUPP OBSERVED d I
PREDICTED
H
20 20
12
18
24
12
TIME (HOURS)
18
X
10 9- 10
I I I I I | I I I I I I I I I I I | I I I I I
' DOWNINGTON
OBSERVED 03 -
PREDICTED
24
m
12
TIME (HOURS)
mmn IQ
20
10
18 24
20
12 18
10
O
z
I I I \ \ I I I I I | I I I I I | I I
FRANKLIN INS
24
m
OBSERVED CD -
PREDICTED -
20 20
12 18
24
10 g; 10
ISLAND RD Al
OBSERVED [JJ
PREDICTED
12
TIME (HOURS)
18
24
20
10
12
TIME (HOURS)
18
24
PHILADELPHIA - 7/13/79 - NO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
12 18
24
^u
I10
O
C
0
i i i i i I i i i i i I i i i i i I i i i i i
" LUMBERTON _ ~
OBSERVED CO -
PREDICTED
-
-
acOmCococo coco
i l"H i i i I t ~~\ 1 1 II 1 IH ti i ii t ii t ii MI 1 11 1 IT t IF nrY ir\ iJTlrf i
3 6 12 18 2
^
1
0
4
20
12
18 24
TIME (HOURS)
I
10 g: 10
O
NORRISTOWN A
OBSERVED CO
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
20
12
18 24
fc 10
ROBB.NSVILLE
PREDICTED
20 20
12 18
24
5.
i
ROXY WATER P
TIME (HOURS)
OBSERVED CO
PREDICTED
20
10
I I I I I I I I I I I I Q
6 12 18 24
TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
20
12
18 24
I 10
SE SEWAGE PL
OBSERVED CD
PREDICTED
20 20
12
18 24
10 g; 10
l I I I | I I I I I | I I I I I | I I I I i
SOUTH BROAD
OBSERVED [JJ -
PREDICTED
6 12 18
TIME (HOURS)
24
20
10
6 12
TIME (HOURS)
18 24
20
12
18 24
I 10
SUMMIT BRIDG
OBSERVED CD
PREDICTED
12 18
TIME (HOURS)
20 20
12 18
24
io 9- io
o
24
o o
SW CORNER
OBSERVED CD
PREDICTED
i
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - NO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
20
12
18
24 0
10
TRENTON
OBSERVED |U
PREDICTED
20 20
6
12
18 24
I
10 9- 10
I I I I | I I I I I | I I I I I | I I I I i
VANHISEV,LL ^^
PREDICTED
6 12 18
TIME (HOURS)
24
CD
mm
TIME (HOURS)
20
10
24
20
12
18
24
10
o
I I I I | i I I I I | I I T I I | I I I I
VINELAND m
OBSERVED CO
PREDICTED
Ql I I __[ I I I I I I I I I I I I I I I I I I I I I Q
24
20
10
6 12 18
TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
20
12 18
24
i
OBSERVED
PREDICTED
- QQQQQ
CD
20 20
2
X
10 g: 10
CM
O
6 12 18
~iiiii|iiiii|iiiii|iiir
' ANCORA m
OBSERVED CD
PREDICTED
6 12 18
TIME (HOURS)
Ol i i i l l I i l i T~^0 i i | | , | M | |
20
10
6 12 18
TIME (HOURS)
24
20
12
18
I 10
i i i i i i i i i i i i i i i i i ] i i i i r
BRIGANTINE
OBSERVED [JJ -
PREDICTED
24 0
20 20
12 18
24
10 g: 10
i
I I I I | I I I I I I I I I I I | I I I I I
BRISTOL
OBSERVED CD -
PREDICTED
20
10
PHILADELPHIA - 7/13/79 - N02 - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
20
12 18
24
i I I I I | I I I I I | I I I I I | I I I I I
CAMDEN m
OBSERVED CO -
PREDICTED
"TDQ
20 20
2
I
10 g- 10
6 12
~~\II1|IIIII|T
CHESTER
d
18
24
12
TIME (HOURS)
18
24
OBSERVED El -
PREDICTED -
12 18
TIME (HOURS)
20
24
20
12
18 24.
CLAYMONT
OBSERVED
PREDICTED
I
CL
CL
d
BBB-
20 20
12 18
24
10 Q! 10
CM
O
CONSHOHOCKEN
6 12 18
TIME (HOURS)
OBSERVED
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
1
-------
20
12
18 24
£ 1°
I
i i i i | i i r i i | i r i I
DEFENSE SUPP
PREDICTED
20 20
12
18 24
X
10 & 10
(NJ
o
i i i i i i i i i i i i i i i i | i i i i r
DOWNINGTON
OBSERVED [JJ
PREDICTED
6 12 18
TIME (HOURS)
24
20
10
12
TIME (HOURS)
18
24
20
12
18
10
CN
O
24 0
20 20
12
18
24,
FRANKLIN INS
OBSERVED [JJ
PREDICTED
dJ ma m
m
m
mm
m
m
immm-
I
10 £ 10
(NJ
O
6 12 18
TIME (HOURS)
24
0 0
n i i i i | i i i i i | i i i i i i i i i i r
ISLAND RD Al . m
OBSERVED m '
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - PLANR RUN
SYSTEMS APPLICATIONS. INC.
-------
12
18
24
6
12
18
24
£ io
O
I I I I I I I I I I I I
" LUMBERTON
CD
^B^H? JSL
l I l l l I l l L.T^rtin
i i i i I i i i i i
OBSERVED CD -
PREDICTED
-
_=m_Wfflannn°:
18
TIME (HOURS)
X
10 n- 10
(M
O
n | | | | | | | ! | | | p
NORRISTOWN A
OBSERVED CD -
PREDICTED
m
24
10
TIME (HOURS)
20
12
18 24
10
CN
O
ROBBINSVILLE
OBSERVED CD
PREDICTED
20 20
12
18 24
2
X
10 g: 10
O
Z
0 0
ROXY WATER P
OBSERVED CD
PREDICTED
u-/,
20
10
6 12 18 24 0
TIME (HOURS)
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
20
12
18
24
a 10
I I I I I I i I I I I I I I I I I I
SE SEWAGE PL
OBSERVED CD
PREDICTED
I I I I I I I I I I ! I I I I I I I I I I I I
20 20
12
18
24
I
10 Q- 10
TIME (HOURS)
SOUTH BROAD
m
a
OBSERVED Q
PREDICTED
mna
i I I I I I I I L..L_L. I I L__L.l I I. I I I
20
10
6 12 18
TIME (HOURS)
24
20
12
18
24
£lO
I
SUMMIT BRIDG
OBSERVED UJ
PREDICTED
12
TIME (HOURS)
18
20 20
12
18
24
10 fc 10
SW CORNER
24
OBSERVED [JJ
PREDICTED
12
TIME (HOURS)
18
20
10
24
PHILADELPHIA - 7/13/79 - N02 - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
20
12
I I I I I I I I I I I I I I I I I I I I I I I
TRENTON
OBSERVED UJ
PREDICTED
2420 20°
12
18
24
10
o
2
2
X
10 CL 10
(M
O
12 18
TIME (HOURS)
24°
I I 1 I I I I I I I I I I I I I I I I I ! I I
VAN HISEVILL ^
PREDICTED
12
TIME (HOURS)
18
20
10
24
20
12 18
24
VINELAND
OBSERVED
PREDICTED
I 10
C1L
m
.mm
mL
20
10
6 12 18 24
TIME (HOURS)
0
PHILADELPHIA - 7/13/79 - NO2 - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
80
12
18
24 0
i r r i i | i i r i i | i i i i i I i i i i i
u AMS LAB m
OBSERVED 0]
PREDICTED
o
(J
m
c iimmmmmm in
mmm -
I I I I I I I I I I
80 80
12
18
24
i I i I I I I I I r I | i I I I I | I I I r
- ANCORA
OBSERVED m
PREDICTED
40
o
o
80
6 12
TIME (HOURS)
18
24
m
m
cr
i i i i i i i i i i i i i i
6 12 18
TIME (HOURS)
24
80
12
18
i r i i i | i i i i i i i i r i r | i i i i i
- BRIGANTINE
OBSERVED D
PREDICTED
40
O
o
2480 80°
12
18
I I I I I I I I I I
i i i i |i i i i i i i i i i i | r r \ r
h BRISTOL
OBSERVED m
PREDICTED
40 a. 40
O
o
24
80
12
TIME (HOURS)
18
24
6 12
TIME (HOURS)
18
40
0
24
PHILADELPHIA - 7/13/79 - CO - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
12
18
24
12
18
o
o
ou
40
[
0
! 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1 | I 1 1 1 1
- CAMDEN
OBSERVED UJ
PREDICTED -
~
nniTJITJIir
5 6 12 18 2
ou ou
"S
40 £ 40
O
(J
0 0
4
1 I 1 I I | I 1 I 1 [ | I I I I I | I I i I !
- CHESTER
OBSERVED UJ
PREDICTED -
~ ~
1 1 1 i 1 1 1 i i i i 1 i i i i i 1 i i i i i
3 6 12 18 2
40
0
4
TIME (HOURS)
TIME (HOURS)
80
0
12
18
24
i i i i | i i i i i | i i r T i \ \ i i i r
OBSERVED [JJ
PREDICTED
I 40
8
12 18
TIME (HOURS)
80 80
12
18
24
40 a! 40
T T i i i | i i i i i [ r i i i i
- CONSHOHOCKEN
24
PREDICTED
80
40
6 12 13
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - CO - PLANR RUN #1
SYSTEMS APPLICATIONS, INC.
-------
80
12
18
24
I I I I | I I I I I \ V 1 1 I I I 1 r I I I
h DEFENSE SUPP OBSERVED -
PREDICTED '
8: 40
o
o
80
12
18
24_
O
o
6 12 18
TIME (HOURS)
ou
40
°c
1 1 1 1 1 | 1 1 1 I ! | ! 1 1 1 1 j 1 1 ! 1 1
^ DOWNINGTON
OBSERVED CO
PREDICTED -
) 6 12 18 2
ou
0
4
TIME (HOURS)
80
12
18
I I I I | I I i I I |
- FRANKLIN INS
fc 40
o
o
24
80
OBSERVED CD
PREDICTED
80
12
18
24
40 a! 40
I I I I I | I I I I
- ISLAND RD Al
_
-
I i i i i i | i i i i i
OBSERVED CO
PREDICTED -
L_ I I 1 1 1 I I | | | | |
12
TIME (HOURS)
18
24
12
TIME (HOURS)
18
80
40
24
PHILADELPHIA - 7/13/79 - CO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
80
12
18
24
T I I I I | I I I I I | I I I I I | I I I I T
- LUMBERTON ^^ -
PREDICTED
40
O
o
CIDEEf
80
12
18
24
a.
a.
o
o
12
TIME (HOURS)
18
24
ou
40
0
I 1 I I I | I I I 1 I | I I I I I | I I I I I
- NORRISTOWN A m -
OBSERVED OQ
PREDICTED -
i i i i i 1 i i i i i 1 i i i i i 1 i i i i i
D 6 12 18 2
ow
40
0
4
TIME (HOURS)
80
12
18
24
I I I I T | I I I I I | I I I I I | I I I I I
r- ROBBINSVILLE ^^ -
PREDICTED '
S: 40
O
o
_L I I I i i I i I i i
12
TIME (HOURS)
18
80 80
12
18
24
40 a; 40
8
24°
i i i i i i i i i i | i i i i i | i i i r
- ROXY WATER P m
OBSERVED [0
PREDICTED
80
40
12
TIME (HOURS)
18
24
PHILADELPHIA - 7/13/79 - CO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
80
12
18
24
1 r i I [ | I I I I I | I I i I
h SE SEWAGE PL
40
O
o
I I i I i
PREDICTED
80 80
6 12
T I I I I | I I I I I| T
- SOUTH BROAD
40 g;
o
o
12
TIME (HOURS)
18
n n
nan
18
1 1 I | I I I T
OBSERVED [JJ
PREDICTED
ma-
m n
I I I I I I I I I I I I I I 10
12 18 24
80
TIME (HOURS)
12
18
24
12
18
24
ou
40
0
I I I I i | I I I I I | I I I I I | I I I 1 I
- SUMMIT BRIDG
OBSERVED [JJ
PREDICTED -
_
-
5 6 12 18 2
ou ou
f
40 £ *0
8
0 0
4 (
I I ! I I | I I I I I | I I I 1 I | I I I I I
- SW CORNER
OBSERVED CD
PREDICTED -
-
-
5 6 12 18 2
ou
40
0
4
TIME (HOURS)
TIME (HOURS)
PHILADELPHIA - 7/13/79 - CO - PLANR RUN
STSTEMS APPLICATIONS. INC.
-------
80
12
18
24
40
o
o
1 1 1 1 1 1 1 1
- TRENTON
-
! 1 1 1 1 1 1 1
1 1 | 1 1 1 1 1 | 1 1 1 1 1
OBSERVED UJ
PREDICTED -
i i 1 i i i i i 1 i i 1 i 1
12
TIME (HOURS)
18
80 80
12
18
24
1 I I I I I I I I I I | I I I i 1 I I I 1 I T
- VAN HISEVILL m
OBSERVED D
PREDICTED
40 §: 40
8
24
12 18
TIME (HOURS)
80
4-0
0
24
80
i i i i i i i i i i i i i i i i i i i r
- VINELAND
OBSERVED
PREDICTED
2
40
O
O
12
18
24
80
40
12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - CO - PLANR RUN #1
SYSTEMS APPLICATIONS. INC.
-------
Appendix E
WIND FIELDS USED IN DIAGNOSTIC RUN 2
APPLICATION TO PHILADELPHIA
89092rl
-------
387
437
487
Wind Speed (m/s)
0 5 10
I i i i i I i i i i I
537
4490
4440
r 4390
0
20
30
4340
Philadelphia PLANE Winds Eval # 2
Layer 1 at hour 7 on 84194
-------
Wind Speed (m/s)
0 5 10
387
437
487
537
4490
4440
4390
0
10
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 8 on 84194
-------
387
30-
437
487
Wind Speed (m/s)
0 5 10
537
*./ 449°
4440
- 4390
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 9 on 84194
-------
387
30^
437
487
Wind Speed (m/s)
0 5 10
H-i i i I i i i i I
537
4490
4440
- 4390
4340
Philadelphia PLANE Winds Eval # 2
Layer 1 at hour 10 on 84194
-------
387
30 r
437
487
Wind Speed (m/s)
0 5 10
I i i i i I i i i i I
537
4490
4440
.- 4390
0
10
30
4340
Philadelphia PLANR Winds Eval #
Layer 1 at hour 11 on 84194
-------
387
437
487
Wind Speed (m/s)
0 5 10
I i i i i I i i i i I
537
4490
4440
- 4390
10
20
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 12 on 84194
-------
Wind Speed (m/s)
0 5 10
;- 4490
4440
4390
20
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 13 on 84194
-------
Wind Speed (m/s)
0 5 10
+
ml
H
ni
- 4490
4440
4390
0
10
20
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 14 on 84194
-------
Wind Speed (m/s)
0 5 10
)- 4490
4440
j- 4390
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 15 on 84194
-------
Wind Speed (m/s)
0 5 10
I l i i i
387
mnmii
4490
4440
r 4390
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 16 on 84194
-------
Wind Speed (m/s)
0 5 10
387 tt t f f f
t t t
f t f
t f\f f f f
f t\ t M t-
f f f t t Kt t f
f f t t t At t t
f f t t t t \ t t
t t t\t t ^
t t f-
tt'tv't't't'rr'f
'''''
iw,mi'i
m
- 4490
4440
£4390
30
4340
Philadelphia PLANR Winds Eval # 2
Layer 1 at hour 17 on 84194
-------
Wind Speed (m/s)
0 5 10
387
1111
,,1111
20-f f f f f
ft f f f
t f f t /
NNNXNXNN
NNNNNNNN
4490
4440
r 4390
20
30
4340
Philadelphia PLANE Winds Eval # 2
Layer 1 at hour 18 on 84194
-------
Appendix F
HOURLY OZONE CONCENTRATIONS IN PHILADELPHIA PREDICTED
FROM DIAGNOSTIC RUN 2
89092rl
-------
IS
.0 -5
Ii
I
CM
i
8
«
o
o
m
o
oo
o
«. t i i'
§
a
0
'35
8
o
.Q
0)
E
.c
a
a.
x^-
9)
O
N
O
1S3M
-------
o
o
o
to
"
§ §
^ K}
* *
o
« -
I f.« « « f J t i j i i ]
C ,;.. s?
X \: » « .
o
CN
C
"5.
v_x
O)
c
o
'(75
10
8
o
CT)
CT1
a
a.
3)
O
N
O
"t. t r t t t i ir t- f i i \- r t t i
i- t t. r ft- f *
1S3M
-------
6
u
I
E
E
=
.5
2
O
o
r- «
» t I * » t * f± I ' ''» I
1
r>
in
rv
CM
in
(
o
in
a:
(^
oo
rx
to
rv
CN
m
I I j I I (
O
CM
i i t i i i r
i t } t t r= l.. i. i i i; i I i r i t t .1. i
1S3M
CN
u
O
1SV3
§8
o
o
o
o
o
J l l I-. I I- l,.l t
» J. l I T^
<0
(ft
r^
N
O
CM
CD
-5
C
O
'«
I/}
E
IV
a
a
v^^
v
o
N
O
00
't r i t t t 'i t t. t t t. t t r t r <^ t. t >- i- t-1 »( J_L_I_LI I
o
-------
II
i!
o o o
O 93 <0
10 * *
****
1SV3
o o o o o
f
i
I
00
03
i i \\ i i \ T r i | i i T_
i t i f i r i i i i 4. j. t
X
§
'i r i \ r t \ t i i i i t f r; i i li i tit r i. i t \ it i f r*
o
(N
1S3M
CN
k_
O
II
o
(O
1SV3
o
(M
O
O
O
03
O
> in
i! -
J£ E es
II -
(^
o
in
p>
o
x *
25
2 5
I *
I
^ i-
\, ,J i 4. j it. j. i; ^ j r t. | r t i: | i. i. r | i t [^
o
cs
"l t i: M' t t T ! t r t > t' f V 1 ..t- t 1 I 1= ! t' t' t t' t' t-. t Li L"
o
CN
O)
in
o
£
9)
o
.Q
O)
r^
O)
a
a
o
N
O
XS3M
-------
o
o
in
o
03
o
o
O
o
(O
o
*
«r
i- (O
in
1
in
r-
03
I *
ce.
is
o «*
8 *
CM
IB O
i *
eo
I I I | I I i f i i ± j t i i J. I I I | i I- I I I I I I I
I/)
"i- i- i I i i t t i i \ i i } r r I i t. i. i i i I' i i. f i t i t i i
o
CN
t
O
1S3M
CM
i_
D
1SV3
O3
oo ^:
I 5
I!
ii
in
8
m
o
SJ
o
(O
o
(M
8
o
oo
o
-------
(O
3 i
<- ID
in
f» *
1SV3
So o o o o
f CM O 03 fr « * * * *
II
j i
x =
11
a:
O
§
CM
I
§
5=
OQ
I » [ I I J. | k I I J I 1 I | I I I | I I
o
r»5
o
CM
i: t i I t i i t i t t t i i f i t t r i i t r r t t t r t.. i i t i
o
CM
1S3AA
CM
i_
0
1SV3
to
to
o
"'. r*-
i- IO
in
II
5
E
3
O
O
in
o
03
o
to
O O
CM O
* *
g §
^ *
x -E m
oo
to
o r>-
o *
; i. - i i.
rjr i i r | i i .
< ^ s -<,-* ~ J^ »^_-J<
o
n
o
CM
't: I. t I t I' I' t t t I 1 E- I f I t."l f: 1 t. t. t f' 1. f; I -t ft t I
Q O O
o
CM
O)
o
'55
%
O
O3
.
a.
a.
o
N
o
1S3M
-------
^ in
1*5 tn
» n
II
E E
J I
x =
J.E
2
o o
O 00
O
O
o
tO
*
*
1SV3
o o
O
00
i i i | i i i i i i A j i i i { > i i j t i i | t i i | i
o
o
o
CM
i-i t i i i i { i r i i r 'i r r t { \ i t i i. t i i. i'
1S3M
CM
^.
D
in
*>
_[ j i i t j. i i i ).. i i^i t i i' 1.1 t j i' i .1 v i r i i j i i i_
§
I
8
!=
o
n
o
(N
"t I t I" t t I i t-.t.'f. i i- t: t t t \ i't t t »_ f *" t Li it t I- I ~
o
(N
05
rv
CD
c
o
'35
w
1
V
8
0
O)
Oi
a
a.
0
N
O
1S3M
-------
o
1
i 1
s:§
2 2
p^
to
m
p»
2
p»
CN
m
§
m
o
so
J.SV3
o
(M
§
o
03
o
-------
to
*;
03
°! r-
4- t I i t J. J I I I ) I I I ( I I t [; I t t |
5
§
I"
tf)
't i t f i i t 't. i i >.. i t I r r i i i r i r t I i t t t i t t i
o o o
o
(N
1S3AA
1SV3
oo
I S
J!
ii
§000000
oo } r.-t i i i i; t: r i- I i
o
r*)
CN
L.
c
"o.
^~s
o
'35
n
1
0)
o
.0
OJ
r--.
a
a.
N^<
0)
o
N
o
1S3M
-------
00
«
m
*~
"
10
m
z *
z£
1SV3
§000000
oo to * CM o ao
in <***** o
» * * ^ ***
o
o
o
8
o
i t M t i. t > t. r f i r { M i i t i i i i"
1S3AA
.
_O
Q.
1SV3
CT)
f KI t r ft t-'t t f- r i t i i i
-5
o
E
a>
CJ)
a
a
o
N
O
1S3M
-------
Appendix G
TIME SERIES OF PREDICTED AND OBSERVED HOURLY OZONE, NO,
AND CO CONCENTRATIONS IN PHILADELPHIA FOR THE
UAM(CB-IY) USING A RICH DATA BASE (POS 'JAM)
89092rl
-------
I I I I I I I I I
- ANCORA
I I I I I I I I I I I I I ! I I
12
TIME (HOURS)
18
6 12 18
TIME (HOURS)
24
30
12
18
r T i i r | i r i i i p i i i i i | r i i i r
- BRIGANTINE
OBSERVED n
PREDICTED
2
7-
0.
M
o
2430 30°
6
12 18
24.
\ i \\ i | i i i i i i i i i i i | i i i i r
- BRISTOL
OBSERVED UJ
PREDICTED
15 g- 15
12 18
TIME (HOURS)
24
0 0
I m i i i i I i i i t i I i i i [Dm"
15
6 12 18
TIME (HOURS)
PHILADELPHIA - 7/13/79 - O2 - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
30
6
12
18
i r i i i | i i i i I
- CAMDEN
24 0
30 30
12
18
OBSERVED Q ~
PREDICTED =-
CESSSHB
15 9- 15 -
Nl
O
12
TIME (HOURS)
18
24
i i i i i i i i i i i [ i r i
= CHESTER
- 15
6 12 18
TIME (HOURS)
24
24 0
30 30
_ i i i i i I i i i i i i i i i i i i i i i
- CLAYMONT
_ I I I I I | I I I I I | I I I 4 I | I I I I i
- CONSHOHOCKEN
OBSERVED
PREDICTED
OBSERVED
PREDICTED
12
TIME (HOURS)
12
TIME (HOURS)
PHILADELPHIA - 7/13/79 - OZ - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
30
12
18
i r i r i | i i i i i | i i i i i
DEFENSE SUPP
15
M
O
2430 30°
6
12
18
24.
PREDICTED
E -
6 12 18
TIME (HOURS)
I i i i i | i i i i i i i i i i i i i i i i r
- DOWNINGTON m
OBSERVED G3
PREDICTED -=-
15 15
O
15
24
6 12
TIME (HOURS)
18
30
6
12
18
I I I I | i I I I I | i I I I I | l I I I I
- FRANKLIN INS m
OBSERVED [JJ
PREDICTED
15
2430 30°
6
12
18
c
12
TIME (HOURS)
18
i i i i i | i i i i i f i r r i i | i r i i r
\- ISLAND RD Al m
OBSERVED m
PREDICTED
15 15
M
O
30
24
15
6 12 18
TIME (HOURS)
PHILADELPHIA - 7/13/79 - OZ - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
30
6
12
18
I I I I \ I I I I I I I I I I I I I I I I
: LUMBERTON
PREDICTED
30
6
12
18
24
__ I IIII|IIIII|IIIII|IIII
- NORRISTOWN A m
OBSERVED CD
PREDICTED -=-
15 9- 15
M
o
12
TIME (HOURS)
18
24
30
12
TIME (HOURS)
18
24
6
12
18
M
O
i i i i i | i i ii i ( i i r i \ i i i i i r
- ROB8INSVILLE m
OBSERVED [JJ
PREDICTED -=-
24 0
30 30
6
12
18
24
15 15
M
O
_lIIII|IIIIII\\IIiII III!
- ROXY WATER P
OBSERVED C3 -
PREDICTED -=-
30
15
6 12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - OZ - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
30
12
18
24.
I i I I I ) I I I 1 I I I I I I I ( I I I I 1
: SE SEWAGE PL
2
fr 15
M
O
PREDICTED
12
TIME (HOURS)
13
i i i i i I i i i i i I i i i i i I i i i i i
: SOUTH BROAD OBSERVED
PREDICTED
15 £ 15
6 12
TIME (HOURS)
30
6
12
18
" i i i i | i i r T r | i i i \ i j i i i i r
- SUMMIT BRIDG
OBSERVED CD -
PREDICTED
15
ISI
O
2430 30°
6
12
18
24.
15 g- 15
i I I I I | I i I i i | i I I I i [ i i i I r
- SW CORNER
r OBSERVED (TJ -
PREDICTED
12
TIME (HOURS)
18
24
i i i i I i i i i i I i LJ
,a
12 18
TIME (HOURS)
15
24°
PHILADELPHIA - 7/13/79 - OZ - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
30
6
12
18
a.
a.
N
O
i i r i i | i i i i i | i i i i i | i i
- TRENTON
OBSERVED
PREDICTED
2430 30°
6
12
18
24
12 18
TIME (HOURS)
2
15 9- 15
rsi
O
I I I I I | I I I I I I I I I I I I I 1 I 1 i
- VAN HISEVILL m
OBSERVED CO
PREDICTED -=-
12
TIME (HOURS)
30
15
18
24
30
2
I
6
12
18
I I I I I I I I I I I I I I I I 1 I I I I I I
- VINELAND
OBSERVED 13
PREDICTED
m
30
15
0 6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - OZ - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
20
6
12
18
i i i r | i i i i i j i i i i i | i i i i i
AMS
OBSERVED
PREDICTED
2420 20°
12
18
10
o
10 10
O
12
TIME (HOURS)
18
0 0
1 1 1 1 1 I 1 1 I
ANCORA
-
I | i i i I r i i
1 j 1 1 1 1 1 j 1 1 1 1 1
OBSERVED CO -
PREDICTED .
i 1 i i i i i 1 i i i i i
12 18
TIME (HOURS)
20
10
24
20
6
12
18
i i r r | i i r i T^ | i r i i i | r i r i r
BRIGANTINE
OBSERVED [JJ -
PREDICTED
24 0
20 20
6
12 18
24
5
CL
Q.
O
10 9- 10
O
12 18
TIME (HOURS)
24
0 0
i i i ri| i i r r T | i i i i i | i r i\
OBSERVED UJ -
PREDICTED
I 1 I I I j I I I I 1
1. I 1 ! 11(11!
20
10
12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
12
18
12
TIME (HOURS)
18
zu
10
0
i i i i i | i i i i i | i i i i I | i I i i i
" CHESTER
OBSERVED Q -
PREDICTED
^^ ;
3 6 12 18 2
^u
10
0
4
TIME (HOURS)
12
18
2
T
10
O
i i : i i | i i i I i |
" CLAYMONT
'\ i i i i 1 i i i i>~J
i i i i 1 i i i i i
OBSERVED 03 -
PREDICTED
2420 20°
12
18
24
12 18
TIME (HOURS)
2
10 10
o
I I I I I I I I I I I I I I I I I I I I i I
CONSHOHOCKEN
24
PREDICTED
20
10
6 12
TIME (HOURS)
18
24
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
2o'
< \
f 10
*» '
O
0
5
I
~
_
/
3
20'
1
£ 10
o
C
0
)
I
-
-
-
J
**
3
6 12 18 2
i i i i I i i i i i | > i i i i | i i i i '
DEFENSE SUPP ^^ =
PREDICTED
-
-
t 1 t I 1 t 1 1 1 1 ^^^"t~~ 1 I 1 I 1 I 1 I 1 1
6 12 18 2
420 20{
10 {£ 10
8
0 (£
4 (
5 6 12 18 2
1 1 I 1 1 | t 1 I 1 1 | 1 1 1 1 1 | ! 1 1 1 1
' DOWNINGTON
- OBSERVED CD -
PREDICTED -=-
* ^
_
'- &b°m '-
1UJITJITJITJ i H i i '"CD^CCICDCDlZllTJtlHTJ^ ' CDCDCD
5 6 12 18 2
420
10
0
4
TIME (HOURS) TIME (HOURS)
6 12 18 2
1 ! 1 1 | 1 1 1 1 1 | 1 1 ! 1 1 | I 1 1 1 1
FRANKLIN INS m "
OBSERVED CD -
PREDICTED _
-
m°m -
A ^
**ll 1 i 1 ! 1 1 1 1 | 1**^ I 1 I 1 1 1 1 1 1
6 12 18 2
20 2o'
S
X
10 ^ 10
O
0 0
4 (
5 6 12 18 2
i i i I i | i i i i i | i i i i i | i i i i i
: ISLAND RD Al ^^ =
PREDICTED -=-
-
-
-
:^/~^\^ :
5 6 12 18 2
4
20
10
0
4
TIME (HOURS) TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
«TT">TFM<; APPI irATinw
-------
20C
^-s
£ 10
-~s
0
C
)
I
-
3C
i
°0
20(
S
I 1°
0
3
I
_i
0
6 12 18 2
i i i i | i i i i i I i i i i i | i i i i i
LUMBERTON ^^ =
PREDICTED
«
-
^M 1 J^ 1 1 f^""t^l I ll Ml ^1 i 11 HI 1 11 HI 1 II ! ll 1 II Ml 1 11 Ml I {1 I I
420 20<
^"n
10 {£ 10
g
) 6 12 18 2
i i i i i I i i i i i I i i i i i | i i i i i
~ NORRISTOWN A "
I PREDICTED -=- I
-
-
1 1 1 -LXI 1 I I l-»< 1 1 1 1 1 _L_ L 1 1 I 1 I 1
420
10
rt
6 12 18 24" "0 6 12 18 24"
TIME (HOURS) TIME (HOURS)
6 12 18 2
i i ( i | i i i i i | i i i i i | i i i i i
ROBBINSVILLE J
PREDICTED -=- I
i i i _u \- i -j i i i 1 i i i i i 1 i i i i i
6 12 18 :
420 20
X V
I
10 i 10
0 0
.4
) 6 12 18 2
I 1 1 1 1 | ! 1 i 1 1 | 1 1 ! 1 1 | 1 1 1 1 l
' ROXY WATER P
OBSERVED CD -
PREDICTED
\ \ J \ \ \ \ r--i i i 1 i i i 1 i | i i i
0 6 12 18 ;
420
10
0
.4
TIME (HOURS) TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
20
6
12
18
2
fr 10
o
SE SEWAGE PL
OBSERVED
PREDICTED
6 12 18
TIME (HOURS)
J-lO
20 20
6
12 18
1 I I I I | I I I I I I I I I I 1 I 1 I I T
' SOUTH BROAD m
OBSERVED CD
PREDICTED
2
10 9- 10
24
12
TIME (HOURS)
18
20
10
24
20
6
12
18
24
2
I
O
1 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1 | 1 1 I 1 !
SUMMIT BRIDG
OBSERVED CD
PREDICTED -=-
III!
6 12
TIME (HOURS)
18
12
18
24
I
10 10
O
24°
l l l l l | l I i l l | l l i l l | l l i I l
SW CORNER
OBSERVED CD -
PREDICTED
12
TIME (HOURS)
18
20
10
2*
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
20
12
24
I 10
TRENTON
OBSERVED
PREDICTED
I I I I I
20 20
12
18
24
3
10 9- 10
o
6 12 18
TIME (HOURS)
24
20
VAN HISEVLL
PREDICTED
TIME (HOURS)
10
20
12
18
fr 10
o
i i i r \ f i i r i i [ T i i i i | i i r
VINELAND
OBSERVED [
PREDICTED -
QI I I I I I I I I I I ; I I I I I I I I I I I i i _
0 6 12 18 24
20
10
TIME (HOURS)
PHILADELPHIA - 7/13/79 - NO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
2420 20°
o
i i i i i I i i i i i i i i i i i i i i i i i
~ AMS LAB
OBSERVED UJ -
PREDICTED _
6
12
18
I 1 1 I I | T I I I F | I I I I I | I I I I
' ANCORA
OBSERVED
PREDICTED
2O
12
TIME (HOURS)
18
12
TIME (HOURS)
10
18 24
20
12
18
2
8: 10
s^
i
i i i i i [ \ r i i i i i i i i \ | i \ r i i
BRIGANTINE
OBSERVED D
PREDICTED
2420 20°
6
2
X
10 i 10
12
TIME (HOURS)
18
24
0 0
12
18
i i i i i | ii i r i | r i i i i | i i i i
' BRISTOL
OBSERVED UJ
PREDICTED
20
10
j i
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - NO2 - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
20
6
12
18
i i i i i I i i
" CAMDEN
2420 20°
6
12 18
24
OBSERVED CO -
PREDICTED -=-
CO
I I I I I I I I I I I I I ! I I I I I I I I
X
10 §: 10
CM
i I r i i | i i i r r | i i i i r i IT i i r
CHKTER OBSCRVEO n H
PREDICTED
CO
12
TIME (HOURS)
18
24
12
TIME (HOURS)
16
20
10
24
20
6
12
18
2
10
i
i i i I i | i I i \\ | i i i r i | i r i i r
CLAYMONT m
OBSERVED Q ~
PREDICTED
CJ CD
CDCDCD-
20 20
6
12 18
24.
2
10 i 10
(M
o
6 12 18
TIME (HOURS)
24
0 0
i i i i I i I i r i | i i i i i | i i r
CONSHOHOCKEN
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
20
12
i i i i i i i i i i r ] i i i I i i r i i i i
DEFENSE SUPP
18 2420 20°
12
18
PREDICTED
10
i
2
I
10 J 10
PsJ
o
n I I I I ( I I I I i | I I I I I | I i I I r
; DOWNINGTON OBS£RVED ;
PREDICTED .
20
10
6 12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
20
12
18
ni i i i | i i i i i | i i i i i i i i i i r
' FRANKLIN INS
OBSERVED 03 -
PREDICTED
10
20 20
0 .
12
18
24
I
10 it 10
s
6 12
TIME (HOURS)
18
24
0 0
i i t t r|Tir
' ISLAND RD Al
20
OBSERVED
PREDICTED
12 18
TIME (HOURS)
10
24
PHILADELPHIA - 7/13/79 - N02 - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
i i i i i I i i i i i I i i i i i
" NORRISTOWN A
I I I I I I I I I I I I I I I I I I I I I I I
LUMBERTON
OBSERVED
PREDICTED
TIME (HOURS)
6 12 18
TIME (HOURS)
20
6
12
18
i i i i I i i i i i I i i i i i i i i i r r
R088INSV,UE
24 0
20 20
6
12 18
PREDICTED
O U
10 g; 10
6 12
TIME (HOURS)
24
0 0
i i i i i i i i i i ) i i r i r i i i i r T
- ROXY'WATER p
OBSERVED [Tj
PREDICTED
20
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
20
12
18
24
T i i r i i i i i i i j i i i i i | i i i i i
SOUTH BROAD OBStRVED ;
PREDICTED
20
CM
O
24 0
20 20
I I I I I I I I I I I I I I I I I
i i i i i I i i i i i _l__L_i_. .1 i. i I i i i i i
TIME (HOURS)
6 12 18
TIME (HOURS)
20
6
12
18
24.
T I I 1 I I I I I I I | I I I i 1 I I 1 I
' SUMMIT BRIDG
OBSERVED [
PREDICTED -
10
i
6 12
TIME (HOURS)
18
20 20
12
18
24
2
X
10 £ 10
i
24°
20
SW CORNER
OBSERVED (JJ
PREDICTED
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
20
i i i i i | i i i i i
' TRENTON
I 10
O
12 18 24_
I I I i | i I I i I
OBSERVED CD
PREDICTED -=- .
20
I
10 £ 10
i
6 12 18
TIME (HOURS)
24
12
18
24
1 1 1 1 I
n
PREDICTED
JCDUffCDUlCD"
20
10
6 12
TIME (HOURS)
18
24
20
6
12
18
S 1°
(M
O
z
i i i i i I i i i i i | i
VINELAND
2420
OBSERVED CD -
PREDICTED
°
1 1 1 1 1
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - N02 - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
o
o
80
4O
c
r\
} 6 12 18 2
i i i i i I i i i i i I i i < i i I i i i i i
- AMS UB OBSERVED D '
PREDICTED ~
-
-
- m mm -
3mmmmmm m mm mmm -
4 (
80 80
^
40 ^ 4O
O
O
r
L
o o
) 6 12 18 2
1 1 1 1 1 | ! 1 1 1 1 | 1 1 1 1 1 | 1 1 1 1 1
- ANC°RA OBSERVED D *
PREDICTED -
*
-
- -
-
ma nan -
i i i i i i i i i i i i i i i i i i i i i i i
4
80
AO
n
12
TIME (HOURS)
18
24
12
TIME (HOURS)
18
24
80
6
12
I I i i I I I I I I I I I r
- BRlGANTINE
40
o
o
18
\ i i i r
2*80 80°
6
12
18
OBSERVED CD
PREDICTED
i i i i i | i i r r T^ i i i i i | i i i i r
OBSERVED [JJ
PREDICTED
t
40 g: 40
8
0 0
6 12 18 24 0
TIME (HOURS)
6 12 18
TIME (HOURS)
80
40
24
PHILADELPHIA - 7/13/79 - CO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
80
6
12
18
T I I I I | I I I I I | I I I I I | I I
I- CAMOEN
8
2480 80°
6
12
13
CD
ci c
_mcDaa
PREDICTED
i i i i I i i i i i i i i | i i i i i
- CHESTER
OBSERVED CD
PREDICTED -=-
40
o
O
6 12 18
TIME (HOURS)
24°
2480
12
TIME (HOURS)
18
40
24°
12
18
24
12
18
40
I 1 I 1 1 | l 1 I I I | ! I I I I | I 1 I I !
- CLAYMONT
OBSERVED CD
PREDICTED -=- ~
-
' '''! l i i i i 1 i i i i i 1 * i i i i
0 6 12 18 2
a\j ou
f
4O £ 4O
O
cj
0 0
4
1 1 1 1 ! | ! 1 1 1 1 | 1 1 I 1 1 | 1 1 1 ! 1
- CONSHOHOCKEN m -
OBSERVED CD
PREDICTED -=- -
-
-
3 6 12 18 2
ou
40
4
TIME (HOURS)
TIME (HOURS)
PHILADELPHIA - 7/13/79 - CO - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
80
12
18
24
1 I I i i | i i i i i | i i i
h DEFENSE SUPP
40
O
O
OBSERVED
PREDICTED
i i i i I i i i i i
80 80
6
12
18
24
1 I I I 1 I I I I I i I I I i i i I I I I I r
- DOWNINGTON
OBSERVED CD
PREDICTED -=-
40 i 40
O
o
12 18
TIME (HOURS)
24°
80
4.Q
12
TIME (HOURS)
18
24
an
40
0
) 6 .12 18 2
- FRANKLIN INS
OBSERVED CD
PREDICTED -=- -
-
- -
-
5 6 12 18 2
an an
"£
4O i 4O
O
O
0 0
4 t
) 6 12 18 2
1 1 1 1 1 | 1 1 1 1 1 | 1 1 ! I 1 | l 1 ! 1 1
- ISLAND RD Al
OBSERVED CD
PREDICTED -
-
-
- i i i i i I . 1 1 ! I 1 1 t 1 1 1
3 6 12 18 2
4
an
*0
0
TIME (HOURS)
TIME (HOURS)
PHILADELPHIA - 7/13/79 - CO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
80
12
18
24_
g: 40
O
o
I 1 1 1 I 1 1 1 1
- LUMBERTON
-
Jddddddnn
t
1 1 1 1 1 1 1 1 1 1 1 i 1 1
-
PREDICTED -
^^1^1 Tim
12
TIME (HOURS)
18
80
12
18
24
40 8: *o
o
u
24°
1 1 1 1 1 1 1 1 1 1
- NORRISTOWN A
-
-
i | i i i i i | i i i i i
OBSERVED UJ
PREDICTED -=- -
i I i i i i i | , i i i i
12
TIME (HOURS)
18
40
24
80
iiiii|iiiii|iiir i | i i i i i
- R03BINSVILLE
OBSERVED d
PREDICTED
3
40 -
O
o
6
12
18
24
6 12
TIME (HOURS)
18
80
40 £40
8
24°
6
12
18
24
1 1 1 1 1 | 1 1 1 1 I | 1 1 1 1 1 | ! 1 1 1
- ROXY WATER P
PREDICTED
12
TIME (HOURS)
18
30
40
24°
PHILADELPHIA - 7/13/79 - CO - EVALUATION RUN
SYSTEMS APPLICATIONS. INC.
-------
80
12
18
8: 40
O
-------
80
40-
8
0 6 12 18 2
i i i i i I i i i i i | i i i i i | i l i i i
* TRENT°N OBSERVED D '
PREDICTED -=- -
-
i i i i i 1 i i i i i 1 i i i i > l ' j i i i
4 C
30 80
t
40 £ 40
O
U
n n
) 6 12 18 2
i i i i i I i i i i i I i i i i i I i i i i i
- VAN HISEVILL -
PREDICTED ~
-
I 1 1 > I 1 I | 1 1 1 1 1 1 ! 1 ! 1 I 1 1 I I
SO
6 12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
24
80
6
12
18
24
i i I l I i I I I I | I l I I I | I I I T
- VINELAND
OBSERVED m
PREDICTED -=-
£40
8
80
40
Ql I I I I I I I I I I I I I I I I I I I I I I I I Q
6 12 18 24
TIME (HOURS)
PHILADELPHIA - 7/13/79 - CO - EVALUATION RUN
SYSTEMS APPLICATIONS, INC.
-------
Appendix H
HOURLY OZONE CONCENTRATIONS (pphm) IN PHILADELPHIA
PREDICTED BY THE UAM(CB-IV) USING A
RICH DATA BASE (POS UAM)
89092rl
-------
«8
I!
o :§
2 2
o
o
in
1 t ^ | 1 i 1 j t 1 1 \. 1 t J.}. I I I J
O
fl
O
CM
'i i i lii I: t 't ft i- i i t i' i j f r i r i t i
t i t L i > i i i
o
CM
n
c
o
'55
VI
£
0)
J2
O)
a>
£
j:
a
Q.
o
N
O
1S3M
-------
1SV3
o
o>
'£
2
o
o
o
as
o
(O
o
CM
8 8
i=
o
to
FO
m
I I i I t [ i I * [ i I J, ] I I ' I ' ' ' I < ' ' I '
I « i * « >_
t i t r t. t i t.
i t: i= t;: t: r I t t t t t t f
o
CS
1S3AA
o
l»)
o
CM
u
o
a.
1SV3
ii
=
o
O
2 i
o ^
> «
c ^
E CM
1 ^
r>«
o
m
^i' I i i' i I i i i l\i i £ t i r i i. i t-. i. \. r i i. j t t
m
iv
as
I «
|8
8
CM
o
03
o
CO
o
(M
o
o
o
93
§
o
*
g
O
"ft i t it i t t. r= t t r \ t t i 1-r t..i:. I. f I t F-i ^ t t t t-
O)
n
en
c
o
9)
CO
O
a>
r^
01
E
JZ
a
CL
CD
O
N
O
1S3M
-------
££
§0
OD
1SV3
000
00
I E
J I
X =
ii
r» »
(O,
m
X?
(fl
I
8
m
t j f r r r j f i j j i i ij
t i i i i i i I r t
t 1. 1 r i I i ttii i rt
1S3M
u
o
a
U
If
II
tfl
m
i
03
in
o
QD
*
o
CO
1SV3
o
CM
o
o
o
00
o
^
to
i i I ' '
1*5
"t. t r \. t i I t i t i: i t t r i t i t t t t » E } i... i i ill i
o o o
i«5 P>4 «-
O)
r-.
O3
O
0)
8
o
J3
05
r^.
en
?
£.
a
o
N
O
-------
1SV3
*. r-
i- *
r-
oo
I «
I,
2
ts
8 *
o
E *
oo
o o o o o o o
00 (D ^^ ^4 ^3 flD (O
i I i j I I 4 | I ( I {. I I i f f t ( ( i t I j i i i_
o
£N
'r i. i t t. t r t.: i i t i i I t ii r i r i. i t. t t > i t r t i i
o o o
l*> CN ^
1S3M
u
o
Q.
1SV3
O)
n. r*-
o
-------
CO
> > in
i E
Jl 5
S 'c ">
2 2
oo
I *
1SV3
o oo
oo
ii
?
E
3
x
O
Li i ' l '
o
m
2 5
8 *
S §
93
1 I j I t t j I
i i \ t . i i r i i t » i 1 i i i i-
1 I _LLJ_LJ_L » t ! I f 1 I
§
C7)
(/I
C
O
'ft
in
1
0
.0
0)
r^-
CD
Q.
Q.
£>
O
N
O
1S3M
-------
!« * 1
o
CD
1SV3
8
o
o
o
o
o
I f; J I I j J I ^. j 11 t | 1. I » J i I * I 1 I I j 11 I_
I
00
o
o
O
rO
't- f i \ i t i' i i it t t { > t ii t t- i i- t fit i i i' i i
o
CN ^
P
§
O
CM
01
c
o
"w
o
-Q
O5
r^
05
Q.
a
o
N
o
1S3AA
-------
o *
N *
" n
I
X
o
2
o
o
3}
o
o
m
o
00
» J I t I
o o
* i i ' i i * « _
'*'
i i \. i i i r r i i i i I i i i \ ! t t f i \ \ r t i t i i i t'
O
oo
o
(N
1S3M
u
o
a
at
t o
o
o
o o
oo
'ft i ft t t r i i it i; t . i: r i i t i > f t J t
t l t J I til
o
CN ,-
£
O
00
o
-------
1*1
II
5
1^
CO
in
8
m
o
ao
1SV3
o o
§
o
00
f>
o
(O
o
I
E
:§
2
f»-
in
r*
CM
in
r»
o
in
r»
CD
a;
O
en
_j
o
o
i i t r
1S3M
u
o
a
o
CM
it
?
s
2
O
O
O
03
o
to
1SW3
§ g
05
§ §
O
CO
to ,_,_ ,
«.'rT
r^
in
r^
CN
\n
r-
o
in
r-
ao
(T
O
I/I
m
i
r»>
*
r«
CM
PV
5
O
(0
O
(D
O
JD
a>
a.
a
o
N
O
00
1*1
't t-.l I r t i t'. t t i
f i r f i" i t t ₯ i I t i r r t i t \ r'
o
ts
1S3M
-------
<0
^ CM
CM aq i*.
CM
11 II
|l J
> > m
I i -
.£ E CM
8 "5 *>
3 2
o
O
o
ao
o
«M
o
o
1*5
O
(O
cc.
is
*
O f>
o *
03 *
1 r^
8 5!
2 3
_J j I I I I I t I f I I ± j f I I J I I I [ I I I I I I I I I I
o
CM
O
I I i I I t I t
\ i L } i
i i i i i I
t ! i t i i
o
(N
1S3M
u
o
a
i*i to
N r--
11 II
II
I!
ii
o
ao
o
10
o
«
»
*
o
CM
O
oo
O
-------
1SV3
§0 O O O O O O
00 I I I I I I t f | I I I | I
'i i i I i i i i i i t i i < r i t t i i i t i f i i r i i i i
ooo
i*> CM »-
1S3M
U
o
a
00
^
CN
II
s
E
j
"x
o
f.
in
8
m
33
i;
i- *
in
_i
o r>
o -4-
01 *
o o
oo
-------
Appendix I
RESULTS OF THE UAM SENSITIVITY TEST FOR PHILADELPHIA
USING METEOROLOGICAL INPUTS DEVELOPED FROM A SPARSE
DATA BASE (PLANR UAM) AND AIR QUALITY INPUTS DEVELOPED
FROM A RICH DATA BASE (POS UAM)
89092rl
-------
30
6
12
18
I I I I T | I I I 1 I | I I I I I | I I I
h AMS LAB Q8SERVED
PREDICTED -
20
o.
0.
10
12
TIME (HOURS)
18
24 0
30 30
6
12
18
f I I I I | I I I I i I I I I I I I I I I I r
- ANGORA
OBSERVED
PREDICTED
20 20
Q.
CL
10
30
24
6 12 18
TIME (HOURS)
30
6
12
18
i i i i i I i i i i i I ii i i i | i i i i r
- BRIGANTINE m
OBSERVED CD
PREDICTED
20
2
0.
CL
10
Cl
2430 30°
m
20 20 -
a.
0.
s
10 10 -
6 12 18
TIME (HOURS)
24
0 0
i i i i i | i i i i i | i i i i i | i i i I
- BRISTOL
OBSERVED CO
PREDICTED
- 10
6 12 18
TIME (HOURS)
PHILADELPHIA - 7/13/79 - OZ - PLANR RUN #3
SYSTEMS APPLICATIONS. INC.
-------
I
- CAMDEN
i i i i i i i i i i i i i i
2430 30°
6
12
24,
20
a.
a.
I I I I I I I I I I I I I I I I I I I I i I I
- CHESTER
OBSERVED a
PREDICTED -=-
12
TIME (HOURS)
18
20
10
a-
12
TIME (HOURS)
18
24
30
6
12
18
iriii|rrrii| i i r r r | i i i r i
- CLAYMONT
OBSERVED CO
PREDICTED
20
3
a.
a.
n
o
10
24 0
30 30
I i i i i i
20
10
12
TIME (HOURS)
18
24
I i I I I I I I I I I I I I I -I I I I I ' I
- CONSHOHOCKEN
0 0
12
TIME (HOURS)
PHILADELPHIA - 7/13/79 - OZ - PLANR RUN #3
SYSTEMS APPLICATIONS. INC.
-------
30
6
12
18
i i i i r | i i i i i | i i i i i | i i i i
I- DEFENSE SUPP ^^
PREDICTED -=-
20
2
0.
0.
10
2430 30°
6
12
18
24.
mmnnc] -
20 20
2
I
o.
0.
10 10
1 I I I I I I I 1 1 I I 1 Ti I 1 I I 1 I I I
h DOWN.NGTON ^
PREDICTED
30
20
6 12 18
TIME (HOURS)
24
10
6 12 18
TIME (HOURS)
24
30
6
12
18
20
2
I
0.
0.
10
^1iiiiiiiiiiiiiiiiiiiir
- FRANKLIN INS
OBSERVED CD
PREDICTED -=-
2430 30°
12
18
i i i r i i i r r i i [ i i i i i | i i i i r
- ISLAND RD Al
OBSERVED C]
PREDICTED
20 20
a,
a.
10 10
20
12 18
TIME (HOURS)
24
10
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - OZ - PLANR RUN #3
SYSTEMS APPLICATIONS, INC.
-------
30
6
i i i i I ' '
- LUMBERTON
12
18
24.
PREDICTED
30 30
6
12
18
24
T I I I I | I I I I I | I I I I I I I I I I l
r NORRISTOWN A m
OBSERVED D
PREDICTED
20 20
a.
a.
10 10
12
TIME (HOURS)
18
JQ o
30
20
10
12 18
TIME (HOURS)
24
30
6
12
18
i i i i i | i i i i i | i i i i i | i i i i
- ROBBINSVILLE
OBSERVED [0
PREDICTED
20
a.
a.
10
2430 30°
12
18
24.
i t i i r | r i ( i r i r i i r i | i i r i r
- ROXY WATER P m
OBSERVED CO
PREDICTED
20 20
a.
a.
10 10
30
20
C3C3
J I
10
6 12 18
TIME (HOURS)
24
6 12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - OZ - PLANR RUN #3
SYSTEMS APPLICATIONS. INC.
-------
30
12
18
i r i i i ] i i i i i i i i i i i | i i i i i
- SE SEWAGE PL ^^
PREDICTED
20
o.
a.
rO
O
10
2430 30°
12 18
i i i
i i i i i I i i i i i I i i i i i I i i i i i
h SOUTH BROAD ^^
PREDICTED -=-
20 20
a.
a.
10 10
2430
20
6 12 18
TIME (HOURS)
24
10
6 12 18
TIME (HOURS)
24°
12
18
I I I 1 I I I I I i I 1 I 1 1 1 I I I I T
- SUMMIT BRIDG
OBSERVED CD
PREDICTED
20
2
a.
0-
2430 30°
6
12
18
T I I I I | I I I I I | I T
h SW CORNER
20 20
3
a.
a.
s
10 10
30
OBSERVED
PREDICTED
20
12
TIME (HOURS)
18
24
QIX I I
10
12 18
TIME (HOURS)
24
PHILADELPHIA - 7/13/79 - 02 - PLANR RUN #3
SYSTEMS APPLICATIONS. INC.
-------
30
12
18
20
Q.
-------
o
o
m
J.SV3
03 tO ^ CM O 03
o t
> «
_rji r i [ i i. i i {. »i j i i f { I i. ij»>i|.iii|iii_<-i
X
o
8
oo
I*
is
CM
8 *
r-
« o
E *
03
t i \ i i- i i i i > r i t t t t t t r i f t T r r t t i i i i
o o o
i*5 C4 «-
1S3M
o
o
m
§
o j ( i i. « i « i.* i.' « * I** * i' * * f * * * i * * '_
cr
I l I I- t V I I I I I- 'I I { I I t \ i fi fill f if r-
0
(N
C
3
CD
05
3
_Q.
3)
E
Q.
O
N
O
1S3M
-------
1SV3
3 2
n
§
10
m
|I 5
> > m
1 i ^
.= C CM
£ ~c >n
r-
03
I *
a
s£
3 ;
o *
I r«
(N
« o
*
oo
o o o o o o o
33 !O T CM O 00 (D
j i r i j
o
V
V
o
n
O
V)
trt t r i t- i i i t i i t t r i i r t t r-1 f i t- i r- t i t t i"
1S3M
1SV3
58
I 5
.1 1
II
in
o
oo
o
o
o
(O
03
2
8
8
CM
t f pi I !' t I i'£ | .r i I ] L-I »' I 1 T: I I i-I
'' "
I I l
X
I
C
3
u
o:
z
OJ
r-
(35
O
".£
'J.
0.
I
a
o
N
o
"t.rt.l'i r'i :i t:-r
-------
1 t i i i i i i i i f i f t t i i t
00
I S
'x -at
ii
§
in
oo
1S3M
1SV3
o o o o
(Q ^> fs4 Q
o
O
Z
.0.
Q.
Q.
O
N
O
't t t \ f i f f i f it i } i.. i i i- f t i i t f- r r t L r t t i
1S3M
-------
33
15
I E
i I
SI
2 2
o o
O 03
m «
(-« «
to
in
I
I
10
00
_ t { ». i \
o
§000000
00
-------
o>
" 0>
o> ^;
>< o
II
If
ii
o
o
\n
§ g
o o o
o oo 10
*«*«
01
O
o *
oo
t I f f I I I I I J. I I I I I ; t i i ( I t I t I I I I I- I
o
CN
r i t. i I i it i I t r r i ri. t t.t .1 i r i ^ it i i
o
Csl
" n
1 S
I i
< ?
ll
o
03
»- to
in
s
«fr
iSVG
o
(N
o
00
o
i
_i } i t r | i i r | i-"i'i {. ) i t'.Ji'f < r i i t i i i"i i | i f i__
o
CM
'r t t f r i i i i i t i i {' f 'i r t r r i i i [ i t t i L L L i i~
o
(M
01
l^.
O)
3
-J
_g
^
_g
r~
^
a
a
O
N
O
1S3M
-------
9 _
o
o
o
oo
o
-------
in
CM
(^ f*-
CM
_t j.-i i i f i i i i i i ± \ i I.»
I 5
11
I i
& :=
2 2
N
n
r-»
CN
in
fv
O
in
|
O
o
r»
*
(V
5
1 H «i J
^
o
r*>
o
CN
I
E
(V
S
t \ Ft
i i . t t t i 1 t i r t i i i f i f t t t t i i t. i r"
P-
in to
I §
S 1
2 2
r«.
(O
m
O
o
in
o
00
o
§
CM
fv
CM
rv
o
3
L.
a:
a>
C^v
O)
-3
O
CD
T3
£
a
o
N
O
t { \ t t t T t i 1. 1 i I i r i i r t i f
i f I F f I I ! I
1*5
O
-------
to
*
00
I
I
1
"x
O
§
10
5
(O
I
p
o
00
o
to
o
CM
§
0
so
"r. t. i t' t. t i t. i » t. t > I i ri i t i 1= i i f
O
CM
1SV3
I
i
w: r*-
CM
1 S
c f
E CM
;§ «
i^
S
o
O
lO
o
CO
0
to
0
CM
O
83
O
to
iv
X?
is
i- *
tn
o P»
o *
in t
I ( t. | i I- ] [ i. i 4 [ I I- I {; M I I | i'.l. I [ .1 f i | I t" i_
I*--
-£
m
O --:;.;;"" ':
" . s;.!: V.: : - : -.
f.i^-r:-::.:
>..,;.
x
§
8*
w' «:;-=>:=
t t i
t t \ ft: t i t-t i r i t'.ft. r f It
\ t i
o
CM
Qi
O)
O
r
T3
0
a
a
0)
o
N
O
1S3M
-------
CM
T iv
S3 *-
II
I!
S 1
2 Z
o
o
m
o
oo
*
o
i r t t I ri lit i t i t'
I
O
o
o
ao
(O
in
I 5
S!
J E
S =
2 2
71
I
§
(O
-------
l»)
00
8
1
E
'£
2
in
_i
§
o
CM
I
8
O)
o
oo
1SV3
o o
* CM
o o
O 00
o
'I t- 'I I I:
II I- I 1 f t 1 i t t t
o
fM
1S3M
^x
a>
o
N
O
1S3M
-------
VARIABLE...OZONE
BEGINNING DATE...7/13/79
ENDING DATE...7/13/79
OBSERVATION SOURCE...OZONE
OBSERVATION ANALYSIS ... A IRS DATA
THE PREDICTION ALGORITHM...PLNR3 UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE...330
20.00 -
15.
Q.
Q.
X.
o.
5.00
10.00 -
5.00 10.00 15.00
OBSERVED (pphm)
20.00
MOMENTS OF THE PROBABILITY DENS!
OBSERVED
AVERAGE 6.22966
STANDARD DEVIATION 4.56076
SKEWNESS 0.64853
KURTOSIS -0.28793
OTHER MEASURES
MEDIAN 5.50000
UPPER QUARTILE 9.30000
LOWER QUARTILE 2.20000
MINIMUM VALUE 0.10000
MAXIMUM VALUE 20.50000
TY FUNCTION
PREDICTED
7.45774
4.02787
-0.43465
-0.62439
8.42000
10.37000
4.47000
0,01000
16.95000
SKILL OF PREDICTION PARAMETERS
CORRELATION COEFFICIENT OF PREDICTED
VERSUS OBSERVED 0.721
THE BOUNDS OF THE CORRELATION AT THE
CONFIDENCE LEVEL OF 0.050 ARE
LOW BOUND 0.665 HIGH BOUND 0.769
RATIO OF OVER TO UNDER PREDICTIONS 1.750
PERCENT OF OVER PREDICTIONS
GREATER THAN 200 PERCENT OF THE
OBSERVED 22.121
PERCENT OF UNDER PREDICTIONS
LESS THAN 50 PERCENT OF THE
OBSERVED 10.303
-------
VARIABLE...OZONE
BEGINNING DATE...7/13/79
ENDING DATE...7/13/79
OBSERVATION SOURCE...OZONE
OBSERVATION ANALYS IS...A IRS DATA
THE PREDICTION ALGORITHM...PLNR3 UAM(CB-IV) OZONE
AVERAGING TIME...ALL HOURS
STRATITFYING VAR I ABLE...A IRS DATA SITES
SAMPLE SIZE...330
0.20 -
-8.00 -4.80 -1.60
RESIDUAL (OBS-PRED)
THE 8INSIZE EQUALS 1.600
1.60
4.80
8.00
RESIDUAL ANALYSIS
AVERAGE -1.22816
STANDARD DEVIATION 3.24537
SKEWNESS 0.37073
KURTOSIS 0.26506
OTHER MEASURES
MEDIAN -1.18000
UPPER QUARTILE 0.99000
LOWER QUARTILE -3.53000
MINIMUM VALUE -8.19000
MAXIMUM VALUE 10.44000
BIAS CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND -1.7808
UPPER BOUND -0.6755
STD RESIDUAL CONFIDENCE INTERVAL
AT THE 0.0500 LEVEL
LOWER BOUND 9.3089
UPPER BOUND 12.0307
THE MEASURES OF GROSS ERROR
THE ROOT MEAN SQUARE ERROR IS 3.47
THE AVERAGE ABSOLUTE ERROR IS 2.79
VARIOUS MEASURES OF RELATIVE VARIABILITY
OBSERVATION COEFFICIENT OF VARIATION
0.7321
RESIDUAL COEFFICIENT OF VARIATION
0.5210
RATIO OF RESIDUAL TO OBSERVED ST. DEV.
0.7116
-------
Appendix 3
COMPARISON OF INSTANTANEOUS FAA/NWS WIND VELOCITY
OBSERVATIONS WITH HOURLY AVERAGED WIND SPEEDS
S9092r2 9
-------
-------
Appendix J
COMPARISON OF INSTANTANEOUS FAA/NWS WIND VELOCITY OBSERVATIONS,
WITH HOURLY AVERAGED WIND SPEEDS
The UAM requires hourly averaged wind speeds and wind directions, whereas Federal
Aviation Administration (FAA) and National Weather Service (NWS) meteorological
sites report instantaneous wind observations. When wind varies throughout the hour,
the vector averaging of the hourly average wind speeds results in a lower value than
the one reported at an FAA/NWS site, which is essentially an instantaneous value
taken at the end of an hour. The FAA/NWS wind observations are used primarily to
advise pilots of adverse wind conditions at the airport. Thus, during slow wind
conditions, which are typical during an ozone episode, the FAA observer may report
higher wind gusts that will affect operation. The fact that the hourly average wind
speed is calm is of no interest to the pilot if gusts exist that may affect flight
operations. This bias of FAA wind speeds was first noticed while developing UAM
inputs for the California South Coast Air Basin (SOCAB), where several hourly inte-
grated and FAA wind observation sites are located near each other. A brief analysis
of these sites is reported next.
COMPARISON OF HOURLY AVERAGED AND
FAA WIND SPEED OBSERVATIONS
During the development of the California South Coast Air Basin (SOCAB) Air Quality
Management Plan study four three-day episodes of high ozone days were extensively
studied. A systematic bias was seen between the one-hour vector averaged observa-
tions at SCAQMD (South Coast Air Quality Management District) surface wind'speed
monitors and nearby observations of one-minute averaged surface wind speed at
FAA/NWS sites. To estimate the extent of the bias, seven NWS/FAA stations were
compared with seven SCAQMD wind monitors that were located close to the NWS/FAA
stations (i.e., within approximately one UAM grid cell). Table J-l shows the
station pairs and the distance between the FAA/NWS and hourly average stations.
The study periods were the high ozone episodes of 5-7 June, 12-14 August, 21-23
August, and 26-28 August 1985. A total of 1920 collocated data points were collec-
ted. This set was reduced by 801 data points because of either a missing station pair
or a station pair that was below the speed of 1 knot, the lowest value that FAA/NWS
wind monitoring stations can measure.
89092r2 9
-------
TABLE J-1. Surface meteorological observation sites and distance between collo-
cated observations pairs.
Distance
between
NWS/FAA Location UTM (Zone 10) SCAQMD Location UTM (Zone 10) Collocated
Site Name UTMX UTM^
Burbank 356.48 3768.00
Airport
Los Angeles 352.48 3744.40
International
Airport
Long Beach 370.88 3733.76
Airport
El Toro 401.60 3720.16
Airport
Ontario 411.68 3754.16
Airport
Norton 438.80 3756.96
Air Force
Base
Compton 364.40 3740.24
Airport
Site Name UTMX UTM^
Burbank 359.60 3766.40
Lennox 354.40 3744.00
Long 368.00 3734.40
Beach
El Toro 404.8 3716.72
Upland 408.00 3758.48
San Bern- 434.40 3759.20
ardino
Lynwood 366.40 3743.20
Stations
3.52 km
1.96 km
2.95 km
4.70 km
5.67 km
4.94 km
3.57 km
89092r2 9
-------
The mean value for the remaining 1119 data points for the FAA/NWS stations was
3.49 m/s, while the mean for the SCAQMD stations was 1.93 m/s, suggesting that the
FAA/NWS stations were biased by approximately 45 percent (i.e., the hourly average
wind speeds were a little over half of the FAA instantaneous observations). The dif-
ferences in the median values were also very similar, with values of 3.10 m/s for
FAA/NWS stations and 1.80 m/s for SCAQMD stations. Each day was examined to
see if any particular day showed an extreme bias; the bias ranged from 36 percent on
13 August to 51 percent on 26 August.
Similarly, all seven stations were examined to see if any particular station may have
been the cause of the bias. Five of the seven stations showed a similar bias, ranging
from 45 to 56 percent. However, two pairs showed a significantly smaller bias; the
Upland site and Ontario Airport pairs showed a bias of only 28.8 percent, and the San
Bernadino site and Norton Air Force Base pair showed a bias of only 2.4 percent. In
an attempt to see if the agreement was caused by low nighttime wind speeds at
Norton Air Force Base, the wind observations during the daytime period were
analyzed and almost no bias occurred for the San Bernadino-Norton Air Force pair.
The possibility exists that the data may have been incorrectly processed or reported;
for example, incorrect conversion from knots to m/s could account for the small
bias.
We believe that the one-minute average wind speeds reported from FAA/NWS sta-
tions are correct. As a result of the SCAQMD study, which may be true for other
locations also, it was concluded that FAA/NWS surface wind observations may have a
positive bias of as much as a factor of 2.
This preliminary analysis is by no means complete or statistically roboust. However,
it does have important implications for the development of wind fields for air quality
simulations models, such as the UAM. It also helps explain some of the UAM's
tendency to underpredict peak ozone concentrations at some locations in the past,
e.g., St. Louis and Philadelphia. Further analysis of the SCAQMD wind data base and
data from other locations is necessary to determine the extent and frequency of this
bias. For example, hourly average wind speeds are calculated by vector averaging of
a series of lower frequency wind observations. A comparison of one-minute wind
speeds with the hourly average wind speed may give some indication of the extent of
the bias.
8,9092r2 9
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA 450/4-90-006C
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
URBAN AIRSHED MODEL STUDY OF FIVE CITIES - Evaluation o1
Base Case Model Performance for the Cities of St. Louis
and Philadelphia Using Rich and Sparse Meteorological Ir
5. REPORT DATE
April 1990
6. PERFpRMING ORGANIZATION CODE
puts
7. AUTHOR(S)
Ralph E. Morris, Thomas C. Myers, Edward L. Carr
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, CA 94903
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This document presents Urban Airshed Modeling results for St. Louis and Philadelphia.
Two sets of meteorological inputs, representing rich and sparse observed data fields,
were developed for each city. Comparison simulations based on the different input
approaches are presented.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Ozone
Urban Airshed Model
Photochemistry
18. DISTRIBUTION STATEMENT
EPA Form 2220-1 (Rex. 4-77) PREVIOUS EDITION is OBSOLETE
19. SECURITY CLASS (This Report!
21. NO. OP PAGES
226
20. SECURITY CLASS (This page/
22. PRICE
-------
INSTRUCTIONS
1. REPORT NUMBER
Insert the EPA report number as it appears on the cover of the publication.
2. LEAVE BLANK
3. RECIPIENTS ACCESSION NUMBER
Reserved for use by each report recipient.
4. TITLE AND SUBTITLE
Title should indicate clearly and briefly the subject coverage of the report, and be displayed prominently. Sot subtitle, if used, in smaller
type or otherwise subordinate it to main title. When a report is prepared in more than one volume, repeat the primary title, mid volume
number and include subtitle for the specific title.
5. REPORT DATE
Each report shall carry a date indicating at least month and year. Indicate the basis on which it was selected (e.g.. date oj issue, date of
approval, date of preparation, etc.).
6. PERFORMING ORGANIZATION CODE
Leave blank.
7. AUTHOR(S)
Give name(s) in conventional order (John R. Doe. J. Robert Doe. etc..). List author's affiliation if it differs from the performing organi-
zation.
8. PERFORMING ORGANIZATION REPORT NUMBER
Insert if performing organization wishes to assign this number.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Give name, street, city, state, and ZIP code. List no more than two levels of an organizational hiiearchy.
10. PROGRAM ELEMENT NUMBER
Use the progiam element number under which the report was prepared. Subordinate numbers may be included in parentheses.
11. CONTRACT/GRANT NUMBER
Insert contract or grant number under which report was prepared.
12. SPONSORING AGENCY NAME AND ADDRESS
Include ZIP code.
13. TYPE OF REPORT AND PERIOD COVERED
Indicate interim final, etc., and if applicable, dates covered.
14. SPONSORING AGENCY CODE
Insert appropriate code.
15. SUPPLEMENTARY NOTES
Enter information not included elsewhere but useful, such as: Prepared in cooperation with. Translation of, Presented at conference «l.
To be published in, Supersedes, Supplements, etc.
16. ABSTRACT
Include a brief (200 words or less) factual summary of the most significant information contained in the report. It' the report contains a
significant bibliography or literature survey, mention it here.
17. KEY WORDS AND DOCUMENT ANALYSIS
(a) DESCRIPTORS Select from the Thesaurus of Engineering and Scientific Terms the proper authon/ed terms that identify the major
concept of the research and are sufficiently specific and precise to be used as index entries for cataloging.
(b) IDENTIFIERS AND OPEN-ENDED TERMS - Use identifiers for project names, code names, equipment designators, etc. Use open-
ended terms written in descriptor form for those subjects for which no descriptor exists.
(c) COSATI HELD GROUP - Field and group assignments are to be taken from the 1965 COSATI Subject Category List. Since the ma-
jority of documents are multidisciplinaiy in nature, the Primary Field/Group assign men t(s) will be specific discipline, area of human
endeavor, or type of physical object. The application(s) will be cross-referenced with secondary Held/Group assignments that will follow
the primary posting(s).
18. DISTRIBUTION STATEMENT
Denote releasability to the public or limitation for reasons other than security for example "Release Unlimited." Cite any availability to
the public, with address and price.
19. & 20. SECURITY CLASSIFICATION
DO NOT submit classified reports to the National Technical Information service.
21. NUMBER OF PAGES
Insert the total number of pages, including this one and unnumbered pages, but exclude distribution list, if any.
22. PRICE
Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
EPA Form 2220-I (Rev. 4-77) (Reverse)
------- |