EPA-450/4-82-008
Evaluation of the EPA PLUVUE Model
and the ERT Visibility Model
Based on the 1979 VISTTA Data Base
By
Christian Seigneur, A. Belle Hudischewskyj,
and Robert W. Bergstrom
Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, California 94903
EPA Contract No. 68-02-3225
68-02-3582
Prepared for
U.S ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise, and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
June 1982
-------
This report has been reviewed by the Monitoring Data cind Analysis Division of the Office of Air Quality Planning and
Standards, EPA, and approved for publication. Mention of trade names or commercial products is not intended to
constitute endorsement or recommendation for use. Copies of this report are available through the Library Services
Office (MD-35), U .S. Environmental Protection Agency, Ftesearch Triangle Park, North Carolina 27711; or, for a fee, from
the National Technical Information Services, 5285 Port Royal Road, Springfield, Virginia 22161.
Publication No. EPA-450/4-82-008
-------
PREFACE
This report presents the results of the evaluation of the EPA PLUVUE
model and the ERT Visibility Model based on the 1979 EPA VISTTA data base,
which was obtained during the June-July and December, 1979, field programs
conducted at the Navajo Generating Station. This study was performed
under Contract 68-02-3225 with the EPA-ESRL Office of Regional Studies
under the direction of Dr. William E. Wilson. This report was produced
under Contract 68-02-3582 with the U.S. Environmental Protection Agency
(EPA). Thanks are due to Mr. James Dicke of the EPA Office of Air Quality
Planning and Standards for his continuing interest and helpful suggestions
during the course of this study. Thanks are also due to Dr. L. Willard
Richards of Meteorology Research, Inc., for providing useful information
and comments regarding the VISTTA field measurements. Production of this
report was due largely to the dedicated efforts of Sandra Golding and
Carol Lawson, technical editors, as well as several members of the
Publications Center.
111
-------
CONTENTS
PREFACE i i i
LIST OF TABLES vi
1 OVERVIEW AND SUMMARY ]
BACKGROUND..- 1
CONCLUSIONS 3
ORGANIZATION OF THE REPORT 4
2 OVERVIEW OF THE DATA BASE 5
3 PERFORMANCE MEASURES 10
4 OVERALL EVALUATION OF PLUVUE 16
5 OVERALL EVALUATION OF THE ERT VISIBILITY MODEL 26
APPENDIX A: EVALUATION OF THE MODULES OF PLUVUE 32
APPENDIX B: EVALUATION OF THE MODULES OF
THE ERT VISIBILITY MODEL 46
APPENDIX C: SENSITIVITY/UNCERTAINTY ANALYSIS OF PLUVUE 55
APPENDIX D: MODIFICATIONS TO THE ERT VISIBILITY MODEL
COMPUTER CODE 61
REFERENCES 64
-------
LIST OF TABLES
1 Input Data for PLUVUE and the ERF Plume
Visibility Model (1979) 9
2 Comparison of Measured and Predicted Plume/Sky
Intensity Ratios for the Overall Evaluation of PLUVUE 17
3 PLUVUE Performance Analysis of 20 Case Studies
under Clear-Sky Background Conditions..... , 22
4 PLUVUE Performance Analysis of 2 Case Studies under
Dark-Mountain Background Conditions—Normal ized
Residual Statistics 23
5 PLUVUE Visibility Model Performance Analysis of
5 Case Studies under Hazy-Sky Background Conditions--
Normalized Residual Statistics 23
6 Comparison of Measured and Predicted Plume/Sky
Intensity Ratios for the Overall Evaluation of the
ERT Visibility Model 27
7 ERT Visibility Model Performance Analysis of 20 Case
Studies under Clear-Sky Background Conditions.... 28
8 ERT Visibility Model Performance Analysis of 2 Case
Studies under Dark-Mountain Background Conditions--
Normalized Residual Statistics 29
9 ERT Visibility Model Performance Analysis of 5 Case
Studies under Hazy-Sky Background Conditions--
Normalized Residual Statistics 29
-------
SECTION 1
OVERVIEW AND SUMMARY
BACKGROUND
To evaluate the impact of plumes from emissions sources on atmo-
spheric visibility in class I areas, the U.S. Environmental Protection
Agency (EPA) has had to develop both monitoring and modeling capabili-
ties. Consequently, Systems Applications, Inc. developed a plume visi-
bility model for the EPA (Latimer et al., 1978; Johnson et al., 1980) as
part of its obligation to supply technical guidance to states and federal
land managers who enforce the newly promulgated visibility regulations.
Because the performance of a plume visibility model must be carefully
evaluated before it can be used in a predictive mode, one goal of the EPA-
sponsored VISTTA (Visibility Impairment Due to Sulfur Transformation and
Transport in the Atmosphere) field programs was the provision of a data
base needed to evaluate visibility models. The June-July and December,
1979, EPA VISTTA field programs consisted of the recording of a set of
coordinated measurements for characterizing the impact of a specific power
plant—the Navajo Generating Station--!ocated in the vicinity of mandatory
class I areas. Measurements of the emission rates from the stacks and of
the attendant meteorological conditions were carried out. Telephotometers
were used to measure the visual effects of the plume, and plume concentra-
tion measurements were performed from an instrumented airplane. The
aircraft made continuous measurements of S02» NOX, and 03 concentrations
and aerosol-scattering coefficients. To the extent possible, the flight
paths were aligned with the telephotometer sight paths so that concentra-
tion measurements throughout the plume were made at the plume downwind
-------
distance corresponding to the telephotometer measurements. This report
presents the evaluation of two plume visibility modelS--PLUVUE, and the
ERT Visibility Model, version 3--based on the 1979 VISTTA data base.
PLUVUE was developed by Systems Applications, Inc. for the U.S. Environ-
mental Protection Agency. The model., which has been described by Latimer
et al. (1978), and a comprehensive user's guide developed by Johnson et
al. (1980) are available from the EPA. The ERT Visibility Model was
developed by Environmental Research <»nd Technology, Inc. (ERT). A
technical description and user's guide are available from ERT
(Environmental Research and Technology, Inc., 1980).
The overall evaluations of PLUVUE and the ERT Visibility Model were
performed for three different types of background conditions:
> Clear-sky background (20 case studies using data from
13 July, 7 December, and 15 December, 1979).
> Dark-mountain background (2 case studies using data from
28 June 1979).
> Hazy-sky background (5 case studies using data from
4 December 1979).
Simulations were carried out on the basis of emission, meteorological, and
background air quality data—the inputs listed in the 1979 VISTTA data
base (Seigneur et al., 1981).
Evaluation of the dispersion and chemistry modules was conducted from
case studies in which the aircraft had flown through the approximate plume
center!ine. The evaluation of the optics module was conducted for 10
cases from 7 December 1979, for which there was good alignment between the
telephotometer sight path and the aircraft flight path. Thus, the
aircraft measurements could be used as direct inputs to the optics
modules.
-------
CONCLUSIONS
Results of the evaluation of PLUVUE and the ERT Visibility Model
indicate the following conclusions for the conditions considered:
> PLUVUE tends to overpredict the plume visual effects; the
average absolute relative errors in plume contrast are 69,
83, 69, and 154 percent at 405, 450, 550, and 630 nm,
respectively.
> The ERT Visibility Model tends to underpredict the plume
visual effects; the average absolute relative errors in
plume contrast are 42, 37, 30, and 262 percent at 405,
450, 550, and 630 nm (the wavelengths used for the
telephotometer measurements), respectively.
> The dispersion modules of both PLUVUE and the ERT Visi-
bility Model appear to contain the primary source of
uncertainty in model predictions: plume dispersion is
generally underestimated by the modules.
> The chemistry modules of both PLUVUE and the ERT Visi-
bility Model showed reasonable agreement between predic-
tions and measurements of the N02/NOX ratio, with con-
sideration given to the uncertainties introduced by the
dispersion module because of the dependence of plume
chemistry upon plume dispersion.
> The PLUVUE optics module compares satisfactorily with the
case studies. Average absolute relative errors in plume
contrast are 47, 45, 39, and 164 percent at 405, 450, 550,
and 630 nm, respectively. The evaluation of the PLUVUE
optics module showed that the plume visual effects were
overestimated in 9 out of the 10 case studies.
-------
> The ERT optics module performance compares satisfactorily
with the case studies. Average absolute relative errors
in plume contrast are 46, 34, 21, and 51 percent at 405,
450, 550, and 630 nm, respectively. The ERT optics module
underestimated plume visual effects in all 10 cases.
ORGANIZATION OF THE REPORT
First, an overview of the data base of the 1979 VISTTA field program
used here for model evaluation and discussed briefly in terms of data
uncertainties is presented in section 2. The statistical measures used to
evaluate model performance are introduced in section 3. They include
measures suggested by the American Meteorological Society (AMS) Workshop
on Dispersion Model Performance (Fox, 1981) and some additional measures
used by Bergstrom et al. (1981) in their evaluation of PLUVUE. Sections 4
and 5 present the results of the overall evaluations of PLUVUE and the ERT
Visibility Model. The evaluation of the dispersion,, chemistry, and optics
modules of PLUVUE is presented in Appendix A. The evaluation of the
dispersion, chemistry, and optics modules of the ERT Visibility Model is
presented in Appendix B. A discussion of the sensitivity/uncertainty
analysis of PLUVUE is presented in Appendix C. Modifications made to the
ERT Visibility Model computer code for the purpose of carrying out the
evaluation study are described in Appendix D.
-------
SECTION 2
OVERVIEW OF THE DATA BASE
During the 1979 VISTTA field programs, an important data base was
developed for evaluating plume visibility models. To evaluate a mathe-
matical model such as PLUVUE or the ERT Visibility Model, however,
consistent and reliable data that involve stable meteorological conditions
and a fairly homogeneous background (e.g., clear sky, homogeneous cloud
layer, mountain) must be selected.
For this evaluation, we considered five days during which plume and
visibility measurements were obtained at downwind distances ranging from
4 to 31 km. Although data from other days could also have been analyzed,
the five days chosen--28 June, 13 July, and 4, 7, and 15 December, 1979--
appeared to offer the best data sets for plume visibility modeling. These
case studies covered a wide range of conditions:
> On 28 June, good telephotometer measurements of the visual
effects against a dark background--Navajo Mountain—were
obtained.
> During the summer program, the best day for obtaining
plume data with a clear sky was 13 July. Although the
activities on that day were not designed for visibility
evaluation, a flight path of the aircraft corresponded to
two telephotometer plume targets.
-------
> A stable plume was observed on 4 December, with a cloud
layer in the background, that could be approximated in the
model by a homogeneous light gray background.
> The data obtained on 7 December represent almost one-third
of all the data available for optics module evaluation
because meteorological conditions, which included a clear
sky, were favorable.
> On 15 December, the darkest plume was observed, with a
clear sky in the background.
The overall evaluations of the visibility models were carried out
using those cases for which acceptable telephotometer measurements were
made at the time of a flight:
> Flights 5 and 7 on 28 June.
> Flight 1 on 13 July (two telephotometer sites).
> Flights 2, 3, 4, 5, and 12 on 4 December.
> Flights 2, 3, 4, 7, 8, 9, 10, and 11 on 7 December
(morning).
> Flights 1, 2, 3, 4, 5, and 7 on 7 December (afternoon).
> Flights 2, 3, 7, and 9 on 15 December.
Both PLUVUE and the ERT Visibility Model offer the possibility of
calculating the visual effects of
-------
determine the reflectance of the background (cloud layer on Navajo
Mountain). These photographs were compared with the photographs of the
MacBeth colorchecker chart that corresponded to the same sets of measure-
ments (Hudischewskyj et al., 1981). The reflectances of the gray from the
colorchecker charts are available, and the following values were used:
> 28 June, Navajo Mountain background: Reflectance = 0.20
> 4 December, cloud layer background: Reflectance = 0.60.
From data concerning the observer location and sight path azimuth,
the distance between the observer and Navajo Mountain was estimated to be
about 63 km. Some uncertainty is associated with the distance between the
observer and the cloud layer. A distance of 100 km was chosen, but the
model did not appear to be very sensitive to that value unless it
approached the visual range.
The evaluations of the diffusion modules and the chemistry and
aerosol modules were carried out for the same cases as those used for the
overall model evaluations. This technique allowed parallel evaluation of
the dispersion modules and the chemistry and aerosol modules and, thus,
the relative uncertainties introduced by the separate components of the
visibility models could be estimated. In some cases (15 December—cases
3, 7, and 9), the telephotometer sight path was not perfectly aligned with
the aircraft flight path; therefore, evaluations of the overall model and
the optics modules were performed at different downwind distances from
those for the diffusion modules and the chemistry and aerosol modules.
Plume widths were obtained from aircraft measurements by considering
the variation in S02 concentrations. For this study, measured S02^
concentrations above 20 ppb defined the plume. Corrections were made for
variations in the altitude of the aircraft and for the angle between the
plume center-line and the flight path. It should be noted, however, that
-------
the use of SC^ concentration measurements can lead to some uncertainty in
the measured plume width because of the slow time response of the S02
monitor (Seigneur et al., 1981).
In addition to plume widths and plume centerline concentrations,
concentrations integrated across the plume were also of interest because
they provide an indirect measure of plume vertical dispersion. Moreover,
the integrated plume concentration:; of N0£ and particles are of special
interest because they are directly input to the optics modules. Thus, NOp
concentrations and particle-scattering coefficient values integrated in
the plume along the telephotometer sight path were also analyzed.
The optics modules were evaluated on the basis of the data obtained
on 7 December. Because the input data required for these modules are
plume width and plume concentrations integrated along the sight path
through the plume center, it was necessary to choose aircraft flight paths
that coincided with the telephotometer sight paths. For some flight
paths, plume widths and concentrations measured by the aircraft were
inconsistent with the bulk of the data. Such paths probably represent
cases in which the aircraft flew through the top or bottom of the plume
rather than through the plume center and these cases were not considered
in the optics module evaluations. It should be noted that for the 10
cases selected, some uncertainties still exist because we assumed that the
aircraft flight path and telephotometer sight path were exactly aligned
and that the aircraft flight path moved through the plume centerline. An
overview of the input data is shown in table 1.
8
-------
01
O
>-
ac
oo
UJ
UJ
:r
(—
a
•a
a:
o
CU
o
c_
OJ
^3
(11
O
0)
ro
0)
00
CM
o
•(->
ID
•o
t_
eu
Q.
c
0
OJ
-i->
ID
l_
C
o
•^
t/)
•r?
<~~.
•
CO
CM
i — I
•
r--
.— i
O
•
ro
CM
cn
i — i
CM
cn
CM
CM
o
r*»
•
CM
*3-
•
cn
r-^
•
CM
cn
CO
o
r—<
cn
00
o
i — i
X
o
z
o
•
CO
1 — 1
oo
r-l
•
«— 1
O
•
CM
CM
O
•
CM
CM
cu
10
3
O
4->
C_
Q.
>>
(D
.^
Q.
rological data
O
CU
4->
IO
•
CM
«d-
•
CM
C
•
CM
IT)
•
CM
1
•
.— 1
O
CO
•
CO
CO
•
1— 1
1— 1
1
o
O)
in
•
^
•o
cu
CU
Q.
in
C
•^
3
O
CM
CM
! 1
O
•
CM
i — I
1
O
•
r— 1
•
CO
CM
rH
O
•
f— 4
oo
o
•
CM
O
cn
CM
i
CM
O
C
o
•r-
4->
O
0)
t_
•r*
•o
T3
3
CO
en
CM
i
co
r~.
CM
r-~
•
CO
CM
1
CM
r~
•
CO
CM
c— 1
•
oo
cn
CM
CM
•
en
CM
^
OJ
(_
3
+J
L.
UJ
CL
O)
C
0)
J3
^
UJ
UJ
UJ
UJ
1/1
in
"o
>>
•r-
•r—
JD
ID
to
CM
CO
CM
0
•
1
o
VO
CO
0
•
LO
0
•
10
1 — 1
o
•
LT5
•5
'i
CD
>
r—
cu
ce
CO
o
•
o
r*** f~*>
CO •
o o
C CM
r^ c^
CO •
C CM
• cn
O CM
,-H CM
CO •
0 •*
O CM
«3" CM
CO •
0 CO
• CM
O CM
O CO
• CO
C i— i
^~ 1 — 1
CO •
0 r-
• f^*,
O *-•
1
a.
cu H
C -^
o —
Nl
O O)
a>
T3 C
C ID
3 (~
O
t_ r-
O^ ID
o
(O •!-
CO >
co <£>
• PO
i—l CM
CO <— •
1 1
• CM
< — I •—*
r-< i— 1
m o
CM U">
CM r-4
1 1
LO oo
CM O
«* 0
rH
oo ^J-
• CM
*^~ 1— 1
CM 1
CO CO
• cr>
en o
r^ ^O
• 0
ec P-
CM O
o LO
00 CO
CM O
o 10
• CM
CM f-»
^-l O
1 £
2:
in — *
cu
o eu
c =
4-> 4->
•o c
CU
•a E
c cu
•i- C-
S 3
C «/)
5 ID
O CU
/^^ ^
QJ
-------
SECTION 3
PERFORMANCE MEASURES
After the independent evaluations of each of the individual modules
and the overall evaluations of PLUVUE and of the ERT Visibility Model were
conducted, several model outputs v»ere considered in the performance
evaluation procedure:
> Plume/sky intensity ratios, plume contrasts, and blue/red
ratios for the overall model evaluation and for the optics
module evaluation:
- The plume/sky intensity ratios is defined as Ip(x)/Is(x),
where X is the wavelength, Ip(x) is the light intensity
at the plume centerline, and IS(X) is the light
intensity of the background intensity at the same
elevation angle as if the plume were not present. As
previously mentioned, the wavelengths used in the
telephotometer measurements of light intensity were
405, 450, 550, and 630 nm; PLUVUE model predictions are
performed at 39 wavelengths ranging from 370 to 750 nm
and model predictions can therefore be obtained for the
four telephotometer wavelengths. The wavelengths used
in the ERT Visibility Model are 400, 450, 550, and 650
nm.
- Plume contrast is related to the plume/sky intensity
ratio and is defined as; follows:
C(X) =^X) " I$(X) . (1)
ls(x)
10
-------
The blue/red ratio is a measure of plume discoloration
and is defined from the plume/sky intensity ratio in
the blue and red wavelengths:
Blue/Red Ratio - IP(450 nm) • Is(630 or 650 nm) ^ (2)
Is(450 nm) • Ip(630 or 650 nm)
A ratio of less than one characterizes a reddish plume,
whereas a ratio greater than one characterizes a bluish
plume.
> Plume width for the dispersion module evaluation.
> The concentration of NOX at the plume center!ine for the
dispersion module evaluation.
> The ratio of N02 to NOX concentration at the plume
center!ine for the chemistry module evaluation. (This
ratio is also affected by plume dispersion.)
> Integrated NC^ concentrations and bsp (550 nm) values
along the line of sight, where bsp is the scattering
coefficient of particulate matter at 550 nm. This
formulation can be used to evaluate dispersion module
performance.
The statistical performance measures recommended for air quality
model evaluation by the American Meteorological Society Workshop (Fox,
1981) were used to evaluate the performance of PLUVUE and the ERT
Visibility Model. These measures, which include the Pearson correlation
coefficient, systematic bias and variance of differences, gross error and
confidence interval, and root-mean-square error, are discussed in the
11
-------
remainder of this section. Their appropriateness for air quality model
evaluation has b-on discussed elsewhere by Fox (1981).
The correlai ion coefficient r is defined as follows:
r =
- rj '
\
J 'I
N
2 (x -
1=1 i ,m
(X. - X )2
i ,m m
X ) (X.
m i
N
2
1=1
•p " P)
(X. - X )2~
l,p p
1/2
(3)
where X^ m and X., „ are the measured and predicted values of some model
output (e.g., pl'jme contrast) for case study i, x" and Y are the average
m p
measured and predicted values, and M is the total number of case studies
considered in the- model performance evaluation.
Bias is defined as the average value of the differences between
measured and predicted values:
N
d =
(X. - X. )
i ,m i,p
(4)
1 = 1
The confidence interval for the bias can be constructed from
Student's t distribution:
(5)
where tv is Student's t-statistic with v degrees of freedom, s£ is the
estimated variance of the observations, and S^ is the estimated variance
of the predictions.
12'
-------
The variance of the distribution of differences is defined as
fol lows:
=
(6)
The confidence interval for the true variance, o£j, is given by
c<-
v S,
v S
xv,
where x is the *th percentile of the chi-square distribution for
v, x
v degrees of freedom.
The gross error is defined as the average value of the absolute
values difference between measured and predicted values:
(7)
N
GE = -n
.
xi, m ~ xi
(8)
1=1
The root-mean-square error is defined as the square root average
value of the square of the differences between measured and predicted
values:
RMSE =
i , m
1/2
(9)
In addition to the statistical performance measures recommended by
the AMS workshop, several additional statistical performance measures were
computed that are of interest for this analysis. A linear regression
analysis of the measured and predicted values that were considered yields,
13
-------
in addition to the correlation coefficient, the equation of the line of
regression between these measured and predicted values. This line of
regression is characterized by its slope and its intercept on the ordinate
axis. A perfect regression has a slope of one and an intercept value of
zero.
To quantify model performance in predicting plume contrast, plume/sky
intensity ratios, and blue/red ratio, four additional performance measures
were also defined: The average absolute relative error of the plume/sky
intensity ratios, E^, and the average signed relative error of the
plume/sky intensity ratios, S^, are given in equations (10) and (11),
respectively:
1 = 1
i, Measured
i, Predicted
i, Measured
(10)
1 = 1
i , Measured
i, Predicted
(11)
i, Measured
where I^/I^ is the plume/sky intensity ratio, which depends on the
wavelength A, and N is the number of comparisons. The absolute and signed
average relative errors of the contrast are defined by equations (12) and
(13), respectively:
E2(X) =1
N
1=1
c - c
i, Measured i, Predicted
r
i , Measured
(12)
14
-------
- C \
, . _ i x -'i, Measured i, Predicted)
'* ' ~ "N / f
f?. i, Measured
where C^ is the contrast, defined as
These measures were calculated for the cases in which the measured
contrast was greater in absolute value than 0.02, which corresponds to the
accepted value for the contrast threshold (Friedlander, 1977).
It should be noted that the performance measures S^(x) and $2(x) have
the opposite sign of those presented in the performance evaluation of
PLUVUE by Bergstrom et al. (1981). This occurs because the definitions of
S}(\) and $2(A) were chosen to be consistent with the definition of the
bias recommended by the AMS Workshop (Fox, 1981).
Plume/sky intensity ratio is a characteristic of the telephotometer
measurements, whereas contrast is perceived by the human eye. The
performance measures Ei and Si are therefore representative of how well
the model prediction compares with measurements, whereas the performance
measures ^2 and $2 represent the ability of the model to predict a visual
effect such as contrast. The measures E^(x) and E2(x) can be defined as
gross errors when the difference between measured and predicted values are
normalized with respect to the measured values, whereas the measures Si(A)
and S£(A) can be defined as bias when the differences are normalized with
respect to the measured values.
15
-------
SECTION 4
OVERALL EVALUATION OF PLUVUE
The data base used in this model evaluation includes cases exhibiting
clear-sky, dark-mountain, and hazy-sky backgrounds. PLUVUE simulations
can be conducted with or without an object in the background. Blue-sky
case studies were simulated using tie model option of sky background
(i.e., no background object). The dark-mountain background was simulated
by including a dark object having a reflectance of 20 percent. The hazy-
sky background was simulated by including a white object having a reflec-
tance of 60 percent.
Results of the model evaluation are presented in table 2. The data
listed in the table include the date and time of measurement, the downwind
distance of the plume target from the source, the plume-observer distance,
the background conditions, and the plume/sky intensity ratio predicted by
the model and measured by the telephotometer. Model predictions are
conducted at 39 wavelengths, with increments of 10 nrn. Results are shown
at the four telephotometer wavelengths mentioned earlier: 405, 450, 550,
and 630 nm. Model predictions at 405 nm were interpolated from predic-
tions performed at 400 and 410 nm. The measured and predicted
plume/background intensity ratios of typical case studies—one for each
set of measurements taken over a day or a half-day--are shown in figure
1. In general, the model overpredicts the reduction in light intensity
caused by the plume.
For the case studies of 28 June 1979, when the plume is viewed
against a dark background, the model underpredicts the plume/background
intensity ratio in all cases. However, large uncertainties in the plume
16
-------
y
§
_j
0.
u.
o
g
»— <
^
§
^ ">
> C "E
LJ i: c
_l -*'-•'
a!
ce
LJ
>
o
LJ
X
*—
oc.
o
u.
l/>
0
1— 1
&~
<
Q;
>-
t (
in
z
LJ
Z
»— <
>-
K
\
2
_i
Q_
a
^—
o
1 *
£ s i
Q fi ,=
-y 03 *^
i go.
O
a
ce
S "°
;£ c
* 1
^
° 1
Z
to
I-H
cc
Q
C
CO
•4J
CO
•H
o
i
•H
J->
ft
•H
i— 1
Lu
>.
a
O
•a
CO
t-i
O-
•
CO
<$
2E
•
•D
0)
(*
D.
B
CO
v
£
^
XI
CO
u
a.
co
«
£
•D
CO
t-«
O.
CO
«
z.
CD
§
• H
JJ
• H
•D
c
cS
x-v
J
Jt
%^x
J
^ —
*—
5
u
CO
§
'ON'
r-
ON
NO
O
,_!
r^
O
r—
OD
•»
ON
O
NO
(^
CD
r^
CM
ON
CO
O
rA
NO
ON
O
lA
O
ON
O
O
lA
ON
a
O
'*-)
a
>
CO
z
CM
NO
CM
IA
^1.
"
NO
CM
r-
i— • i
lA
CO
c
>-)
NO
o
ON
C3
O
lA
^
-a-
ON
00
O
^~
v
•-H
3
n
^
O N^CNOCOVO NC\OOOCACM'^'K\
vO OOOOOO f^KNCMKNCMr^CMON
Cfv COCDCOOON OONC^ONCT^ONO^CO
o ooooo oooooooo
lA CNKSvOtfSOoDOON
ON cococoOr^ vOp^-^'coOco''^^
Q\ Cr^Cf^CrsO^CN C^CNC^iC7NOCT\t>'CjN
o ooooo oooo*-ooo
r^- *-r*coo^- r^f^coco^t^-^O
CM NO*~K>CDON QNCNf^-COCOCTNC>VA
ON r*-cocococo cccococococococo
o ooooo oooooooo
lA ONCM<3-CMlACMf^-O
CN ^"lAO^r^- ^^f^^-KN^CM\o
0^ (T* O*' O^ O^ CT^ ^ CN C^ CT* ^ CN CJ^ C30
o ooooo oooooooo
o r^r*-cN*^^' cov-r-Nou^r^ooN
\O ^^OO^"<±\O r~cNCOOO«— I/>K\
CD vO^Or^COCD CDr^r--CDCOOOCOCD
o ooooo oooooooo
vA ^iA^CMOp<*(Oe>
fA CD'*OCDI*DON KNcM^oCf^r^-o*—
o cococococo oc^coococoo.co
o ooooo oooooooo
>s Ql
>* N >* 5 >s
J* 8 .* ^H ^
01 X CO CD CO
iA
r** OOOOCM oNCNONoo^^-^-
CJ CMf^ ON ON CO
oooooo oooo
"3" fN CO «— ^O ^— vOiA^^tA
lA ^ T— ON ^~ O f*- *— O o
CDCOCDr^COCO XOCOCOCO
oooooo oooo
•^ocoovO^ ^r^r-^co
«d" *— ^ ^— •& f*+ OCO'^CM
co co co co o^ r^- CD co co co
oooooo oooo
ONCO1^^^^ CMK\r^O
vOlACMeM1^^ O^-lAv^
co co co co co CD r^ co co co
oooooo oooo
OO^COOKN CMCOCNvO
^f*-viAs
^H ^ «-H ^
CD en CD co
lA\OOvOCOf^\ 1-CMSOCM
\o\o^
•^•^•^•^frlAiA ^-r-CMeM
**cM^
-------
1.10 -
1.05 -
1.00 -
0.95 -
0.90 -
O.t
450
500
550 600
Wavelength (nm)
650
700
750
(a) 28 June, Flight 5
1.1
i.o
0.9
0.8
0.7
"I ' T
KcASURED
•PREDICTED
400 450 500 550 600
Wavelength (nm)
650
700
750
(b) 13 July, Flight 1 (Observer-Plume Distance = 15.1 km)
Figure 1. Comparison of measured and predicted plume/sky intensity
ratios for the overall evaluation of PLUVUE--1979.
18
-------
1.05
—1
MEASURED
• PREDICTED
1.00
0.95
0.90
0.85
0.80
350 400 450 500 550 600
Wevelength (ran)
650
700
750
i.oo -
0.95 -
0.90 -
(c) 4 December, Flight 5
400
450
500 550
Mavelenqth (nm)
600
650
(d) 7 December, Morning Flight 3
Figure 1 (Continued).
700
750
19
-------
1.05
1.00
0.95
0.90
0.85
0.80
350
MEASURED
•PREDICTED
400 450 500 550 600
Wavelength (r»n)
650
700
750
(e) 7 December, Flight 2
1.05
i.oo -
0.95 -
0.90 -
0.85 -
0.80
350
400
450
500
550 600
Wavelength (nut)
650
700
(f) 15 December, Flight 9
Figure 1 (Concluded).
20
-------
NOo and particulate integrated concentrations make a detailed evaluation
of the model difficult. It should be noted that the negative contrast of
the plume at short wavelengths (400 and 450 nm) and the positive contrast
at higher wavelengths (e.g., 650 nm) are well predicted by the model for
case 5 [see figure l(a)].
The case studies of 13 July 1979 show an overprediction of the plume
visual effects, especially at shorter wavelengths. This results mainly
from the overestimation of the formation rate of N02 for this plume
simulation.
The case studies of 4 December 1979 correspond to plume intensity
measurements taken against a cloudy background. The model overpredicts
the plume visual effects in all cases. The approximation of the cloud
layer by a homogeneous light gray background in the model seems to be the
main source of uncertainty, because the model components—the diffusion
and chemistry modules—perform as well as for the other days.
For the case studies of the morning of 7 December 1979, the model
shows a general trend toward overestimating the plume visual effects;
model performance, however, is better than for the cases of 4 December.
Excellent agreement is obtained for the cases of the afternoon of 7
December 1979 [see figure l(e)]. This result is consistent with the good
performance of the optics module and of the chemistry module (see Appen-
dix A).
The model predictions also agree well with the measurements for the
four case studies of 15 December 1979. This day corresponds to a stable
dark plume against a clear sky, and it appears that the plume visibility
model is able to predict plume visual effects well under such conditions.
Performance measures for the PLUVUE overall evaluation are presented
in tables 3, 4, and 5 for cases under clear-sky, dark-mountain, and hazy-
sky background conditions, respectively. Statistical performance measures
for the 2 case studies under dark-mountain background conditions (table 4)
21
-------
00
LU
i—i
Q
^
I—
oo
00 00
< z
C_5 O
t—t
O I-
CM i—i
O
u- z
o o
o
oo
1-1 Q
00 Z
>- =3
_J O
•=C OC
Z 13
< i^
O
UJ UJ
rD Q
CO
UJ
00
*£.
T3
CO O
Q£ -i-
~-» -t->
cu ro
3f^f
cc~
CD
O
OO
* LO
O1
•!->
01
ro
t_
4->
C
o
o
cu
E
3 0
i LO
0. -— - LO
•" C
in i
O 1
•i- JE
4-> 4-i
ro CT>
OS C
eu
>•> •—
-4-> CU
•1- >
01 ro o
C 3 LO
cu — • «a-
•4-j
c
»— (
>>
-i^
oo
••^
eu
E
3
D. LO
O
*±
0)
(_
3
01
ro
cu
S
eu
o
c
ro
E
t_
O
14-
t_
cu
o.
CM LO
cn VD cn. LO o
CO OO OO OO CM
o o o o t— i
i i
<£> .-H «fr
CO OO VO LO LO
cn cn co CM r^
o o o o o
i
cn o cn
oo o <— i lO O
cn cn vo co 10
00000
co cn o
r-. o LO cn co
oo oo < — i « — ' o
o o o o .-H
i i
CO O O
VO CM IO «*• «3"
OO OO CVJ CO r-l
O O O O i-i
1 1
4->
c
cu
CU -i-
CU 3 O
3 i — -r~
i— ro <*-
re > >*-
c > cu
O CO
•r- c re o
01 ro eu
m cu E c
eu E o
C_ T3 ••- +J
01 -a co -t-s ex
cu cu -•-> ro eu
a; t- o >— o
3 -i- CU CU f-
«_ 01 T3 t- CX CU
ro ro cu t- o -iJ
eu cu t- o i— c
c z: o- o oo i— i
«^-
_i
VO
^—1
o
•
LO O CO O CM
CM oo o i cn
o o o LO o
* • • t"i •
o o o o o
-H •
O
IO
o
0
CO •
•— i •=*• o o cn
LO i— 1 O 1 LO
o o o LO o
• • • CO •
0 0 O 0 0
+ -H 0
•
0
^~
o
o
•
un r~ CM o i —
00 CM O 1 LO
O O O r-l O
• • • C3 •
o o o o o
+ -H •
o
en
o
o
•
cn LO LO o LO
i£> CM o i en
O o o oo o
O 0 O 0 0
+ -H •
o
.— i
,— t
o
o
cn vo us • ^*-
CO CM O O OO
O O O 1 O
CD 0 0 0 CD
+ -H O
•
O
f—
ro
>
t_
cu
4_>
c
•r-
i — CU
ro o
> c t_
t- cu o
eu T3 t-
+J -r- (_
c <»- cu
in -i— c
u o cu
•t- CU O t-
-i_> o re
01 C CU 3
•r- eu o cr
4-> T3 C 01
ro •!— ro i
^J >»- -r- C
oo c eo «- ro
O o re cu
i— 0 C > E
re ro i
3 01 m ••- cu -4->
•D ro ro i_ 3 o
•r- -f— -r- re t- O
01 co co => I— on
cu
Of
^- oo
OO CM
0 0
••II
O O 1 1
LO CM cn LO
LO LO O") CM
O O LO OO
• * II •
O O t-H i— 1
+ 1
«* oo i— i cn
LO LO cn oo
o o vo ID
o o o o
+ i
00 O co «d-
oo r^* oo co
O O co r-.
O O O O
+ 1
f> i— i o co
oo «* cn CM
O O IO LO
o o o o
+ i
01
o
•r—
-P
O1
•^~
4->
ro
+J
oo
r—
re
3
TJ
•r—
01
eu
o:
T3
eu
NJ
• r-
i—
ro •»-•»-
E I-H •— i ^r\j CM
C. UJ OO UJ OO
0
z
cu
g-
4_>
• •• t_
01 CU •«
O <*- C
•r- CU O
+-> t- -r-
ro 4->
t- CM re
00 r-
>1 3
+-> -o o
•»~ C i —
01 re re
c o
CU CM •
4-> UJ CU >j
c: c- , —
•r- -O 4-> CU
C >
>, re E ••-
-^ O +J
O1 « t- O
^ 01 IK- cu
cu o ex
E -i- T3 01
3 +j cu eu
i — re T3 £-
CX t- C-
re «
eu >, o E
_C 4_> O1 C
+-> -1— T—
00 T3 O
£- C OO
O CO CU «- 4-> C-
C CU T3
•D -i- S C
cu ro
4-> >, CM
01 -i£ O •>
•r- O1 • O
r — •>». O LO
cu LO
CU E C
i_ 3 ro •>
ro i— j= o
ex 4-> LO
O1 ^3"
O O 01
•r- +J O1 •>
+i cu LO
01 T3 i — O
• *— *— • t4-
•t~ t- ^J
4-> O lt-
ro ex o 4->
^01 re
01 cu •+->
t- 01 CM
F— t_ ro <— i
re o t-
3 O -I-J "O
•o c c
•r— »-H o re
01 OO U
CU
t_ -o cu f~-
c E '-i
T3 ro 3
CU i— •
IM •— 1 CX O
•i- UJ CM
r— CU
re 01 -i-> •>
E O 3O
t_ .r- , — CM
O 4-> O
C 01 01 CO
C •-- JO L
o +J re co
c re S
-(-> "D
T3 01 CU 01
C £- CO
re i — • 3 01
re 01 in re
c 3 -t-> ro o
O TD 01 CU
•r- -I- ro E t-
01 01 i- O
01 cu -t-> cn
cu t_ c c i_
<_ O •!- CO
O1 T3 O 5 J3
cu eu o E
«_ IM eu jc 3
•t- E 01 c
!- r- 3
ro re i — 01 i —
cu E ex co ro
C t_ 01 +J
•t- o o ro o
—I C 4J 0 -!->
* +-
22
-------
TABLE 4. PLUVUE PERFORMANCE ANALYSIS OF 2 CASE STUDIES UNDER
DARK-MOUNTAIN BACKGROUND CONDITIONS—NORMALIZED
RESIDUAL STATISTICS
Plume/Sky Intensity Ratios;
PIume Contrasts*
(Wavelength--nm)
Performance Measure
Measured mean value
Predicted mean value
El
Sl
E2
s?
405
1.075
0.932
0.124
+0.124
1.053
+0.153
450
1.112
0.904
0.174
+0.174
1.621
-0.298
550
1.243
0.921
0.243
+0.243
7.135
7.135
630
1.301
0.983
0.218
+0.218
0.803
0.803
Blue/Red
Ratio
0.856
0.926
0.120
-0.084
--
__
TABLE 5. PLUVUE VISIBILITY MODEL PERFORMANCE ANALYSIS OF 5 CASE
STUDIES UNDER HAZY-SKY BACKGROUND CONDITIONS-
NORMALIZED RESIDUAL STATISTICS
Plume/Sky Intensity Ratios;
PI ume Contrasts
(Wavelength--nm)
Performance Measure
Measured mean value
Predicted mean value
El
sl
E2
s?
405
0.874
0.761
0.129
+0.129
0.902
-0.902
450
0.862
0.736
0.147
0.147
0.927
-0.927
550
0.944
0.837
0.113
0.113
2.179
-2.179
630
0.974
0.874
0.103
0.103
1.717
-1.717
Blue/Red
Ratio
0.89
0.84
0.0588
-0.0544
--
__
Ej and S} correspond to plume/sky intensity ratios, and E2 and S2 refer
to plume contrasts.
23
-------
and the 5 case studies under hazy-sky background conditions (table 5) were
limited to normalized residual statistics because of the relatively small
number of case studies conducted. The statistical performance measures
for the case studies under clear-sky background conditions are of the
greatest value because they involve a larger number of cases--20 studies--
and the simulations did not involve any background object.
The correlation coefficient for measured and predicted values for the
overall model evaluation under clear-sky background conditions is negative
in the blue light range-- -0.263 and -0.153 at 405 and 450 nm, respec-
tively; average (0.619) at 550 nm, and low (0.366) at 630 nm. A negative
correlation coefficient (-0.392) is calculated for the blue/red ratio.
This poor overall performance of PLUVUE can, at this point, be attributed
to any of the modules (dispersion, or chemistry, or optics). Uncertain-
ties in the model input data could also affect model performance; however,
it is unlikely that this is the main cause of poor model performance. The
sensitivity of PLUVUE to uncertainties in the input data is analyzed in
Appendix C.
The average absolute normalized error values in plume/sky intensity
ratios are 8.3, 8.8, 5.4, and 5.6 percent at 405, 450, 550, and 630 nm,
respectively. The average absolute normalized error values in plume
contrast are 69, 83, 69, and 159 percent at the four wavelengths. From
the values of the signed deviations in plume/sky intensity ratio and plume
contrast, it appears that PLUVUE overestimates the plume visual effects at
all wavelengths, on the average.
Only two cases under dark-mountain background conditions were
available, an insufficient number for calculating sound statistical
measures. Thus, only E^, Sp E2> and $2 were calculated. From these
results (shown in table 4), model performance appears to be poor. From
the data presented in table 2, it appears that for case 5, the plume was
darker in the blue light range and brighter in the green and red, whereas
for case 7, a bright plume was observed at all wavelengths. PLUVUE
predicts a brighter plume for case 5 at 630 nm only.
24
-------
Five case studies under hazy background conditions were available for
model evaluation and, therefore, only E}, S^, £2, and $2 were calculated
(table 5). Model performance is poor, perhaps partly because of the
approximation of the haze by a light gray reflector.
25
-------
SECTION 5
OVERALL EVALUATION OF THE ERT VISIBILITY MODEL
The data base for this model evaluation includes 20 cases under
clear-sky, 2 cases under dark-mountain, and 5 cases under hazy-sky
background conditions. The cases having clear-sky backgrounds were
simulated by placing the background object beyond the visual range. The
dark-mountain background was simulated by placing a dark object having a
reflectance of 20 percent at an inclination of 60°. The hazy-sky back-
ground was approximated by a white object having a reflectance of 60
percent placed vertically at 100 km from the observer.
The results of the 27 simulations are presented in table 6. The date
and time of the measurement, the downwind distance of the plume target
from the source, the piume-observar distance, and the background condi-
tions are listed, along with the Dlume/sky intensity ratio predicted by
the model and measured by multiwa/elength telephometers at four wave-
lengths. The ERT Visibility Model predictions were carried out at 400,
450, 550, and 650 run. Model performance is described in tables 7, 8, and
9 for cases under clear-sky, dark-mountain, and hazy-sky backgrounds
conditions, respectively. Statistical performance measures for the 2 case
studies involving dark-moutain backgrounds (table 8) and the 5 case
studies involving hazy-sky backgrounds (table 9) were limited to normal-
ized residual statistics, again because of the relatively small number of
case studies conducted. The case studies having clear-sky backgrounds are
of the greatest interest because "hey do not involve background objects
and, therefore, they can more easily be simulated by a plume visibility
model.
26
-------
CO
P—<
in
an
LJ
Z3
_l
«r
>
LJ
>
o
Oi
o
LO
O
LO
0- C
O
PA
vO
0
LA
LA
R
LA
o
-a-
. CM CM
T> ON 00
0> VO r-
£ ,- t-
fA
CO O O
i) r- LA
<- *-'
• Ov vO
0> PA ON
0- <-^ O
• vO
CO r- <-
jj O •*
>- «-"
PA 00
0) ON r-
o_ o_ o
• fA
CO vo vo
& ON CM
O «-
• T- O
T3 CO CNI
0) CD r-
o_ d d
CO LA O
QJ ON CM
o «-
1 — fA
CM O
O ^A
fA ON
r- O
co r-
<- o
0 LA
OV CN
o o
r- *•"
ON CT^
0 0
LA U"\
CN <*
CD ON
O O
00 CN
O 0
LA LA
CO CN
0 0
CO-^MCO^~ COr— ^•iA-3'COCMfA CN *A ^ ^ CO -^ CO CN CO CN
CN CN (N CO f-- CN O -^ O O »— *— fA CN O *~ *~- CN O rA r-. xO \O
u"i iA \O 'O "^ CNOOOOOOO CfNOOOO^O r^-^O^iT^
OOOOO OT-*-T-t-t-«-^- O^-^-r-Dr- t-OOO
CNfA^OlAOCOCNON CN tA «— K\ O VO r^^CMON
COCOCO^P^ ^Or-»^cOOGOf^lA \Of--COCOOf1-- KNCO^O'vO
OOOOO OOOOt-OOO OOOOr-O OOOO
f^^-^-iAtA r^CMvOf^-vO^iAxO T- CO «— rA ^O rA CN -U
co c
> D
2 f
CM CM
vO vo
CM "3
LA O
«- CO
r- CM
VO LA
CM •*
t>. CO
LA 1—
CO
1
CO
»- •It
ir\ ON
0 0
r- r^
CO CO
CM CM
§vO
0
o o
ce LJ
I I
XX X X
to to to to
X 1> 01 01
N D D D
X CO CO CO
r-r^cNCMo ONOr~rt--oco
-------
-D
QJ O
CC -r-
QJ ro
3 Q;
CO
r-H CM
CT> CO ^ OO i—I
CO CO CM i-H O
O O O O •-*
I I
CTl
CO
CM
CM ^ CO O
CO i—l OO • CO
O ^ 1—' O f—<
o o o i I-H
• • • r**- •
o o o co o
+1 O
o
CO CM
oo c>
^^ C21
CO
UJ
Q
~*^
h-
co
UJ
to
«c
o
0
CM
U_
OO
t— 1
00
>-
-J OO
^c ^^
•z. o
< H-l
1—
UJ i— •
0 Q
•z. z.
< O
^" ^ **
2. V_J
Qi
O 0
1 1 ^y
UJ O
Q. oe:
CD
i xx
UJ CJ
Q «f
o oo
>_
1 » >N
^^ V /
_j D;
1-1 4->
ro CD
n< c~
QJ
4-> QJ
••- >
to ro
C 3
OJ -^
C
1—4
^
**/
to
^^^
QJ
3
51
o
0
•5J- CO OO O •
o oocoo r~- i — CMO r— i •— i CM o
OO COCO^CMCM r-~lJ30O|C7> COOOCMCO
CD CTiOCMOO O O O •— < •— 1 OOCDCM
Oi- or
UJ
*- Q
ex. -z.
UJ
CO
QJ
O C
tf) ro QJ
Q£ (_ U
3 -i- O) QJ
O
O
1C
Q.
QJ
O
(O
QJ
C
O Q)
QJ t_ O 4->
t- O r- C
O- O CO i—i
ro
(_
QJ
O
•i- QJ
4J O
tf> C
•i- QJ
4J C
ro •-
4-> 4—
co c:
o
r— O
ro
3 tO tO
"O ro ro
Q)
t=
QJ
O
C I-
QJ O
"O t-
•r- i_
»»_ Q)
C
O QJ
O t-
ro
O
ro
CT
I
I/) CO XI
QJ
CC.
QJ t- ro
O ro QJ
c > E
ro I
•i- QJ 4J
3 O
O
O
•r—
tf)
ro
CO
ro
3
CO
QJ
T3
QJ
N
r- CO t—
> t— Qi
1-1 CM CM
COi UJ CO
to
o
ro
l_
4->
• r-
to
c
QJ
4->
e
• r—
>,
to
^^
QJ
E
3
r—
Q.
QJ
4->
£_
0
14-
-O
QJ
4->
to
-r—
r—
QJ
t_
ro
to
0
4->
to
•i—
4->
ro
4->
to
ro
3
HO
• r—
to
QJ
C_
-o
O)
tsl
•r—
ro
E
t_
O
C
c
0
c
-o
c
ro
C
O
.f—
to
to
QJ
CD
QJ
£_
t_
ro
OJ
C
—t
•K
QJ
OJ
t_
CVJ
to
-o
c
ro
CM
UJ
tf)
ro
QJ
i.
QJ
.C
S
„
(/>
O
tf—
4J
ro
£_
>>
•r—
to
c
QJ
4->
C
.r—
>>
tf)
*^
QJ
E
3
r—
0-
o
"*"*
•o
c
o
0-
to
QJ
l_
t
O
O
( — 1
to
•o
c
ro
i — t
UJ
to
0
.p—
4->
tf)
4->
ro
_4_>
to
,— .
ro
3
T3
.f—
to
QJ
t_
-rj
QJ
M
»r»
f^
ro
E
O
c
•
to
4->
10
ro
c_
4->
C
0
0
QJ
E
3
, —
o.
0
4->
JZ
4->
C
0
••-
ro
3
U
r—
ro
O
QJ
E
o
t_
^.
T3
QJ
0
to
., —
^3
QJ
t_
QJ
2
CM
O
•
O
C
(0
-C
"^
to
to
OJ
^—>
14-
o
4->
to
ro
1_
4->
C
O
0
QJ
E
3
'o.
QJ
4->
3
'o
tf)
JO
ro
•o
Q>
t_
3
to
ro
_c
4->
• r-
s
to
OJ
to
ro
O
4-
•
>>
'QJ
>
•r—
(j
OJ
Q.
to
QJ
<_
M
E
O
OO
CO
-^
c
ro
M
O
CO
CO
n
o
CO
«^-
CO
o
^*
4->
ro
CM
---)
-^
c
ftJ
A
r^
«— H
«
^^
OJ
«
o
CM
QJ
QJ
•s.
to
QJ
tf)
ro
O
4-
O
I_
QJ
JD
E
3
f~
r—
ro
4->
O
4->
28
-------
TABLE 8. ERT VISIBILITY MODEL PERFORMANCE ANALYSIS OF 2 CASE
STUDIES UNDER DARK-MOUNTAIN BACKGROUND CONDITIONS--
NORMALIZED RESIDUAL STATISTICS
Plume/Sky Intensity Ratios;
Plume Contrasts
(Wavelength--nm)T
Performance Measure
Measured mean value
Predicted mean value
El
Si
E2
s,
405
1.045
0.801
0.237
+0.237
0.945
+0.51
450
1.112
0.861
0.208
-0.198
1.120
+1.120
550
1.243
1.148
0.267
+0.041
2.128
-1.069
630
1.302
1.437
0.373
-0.161
3.177
-2.541
Blue/Red
Ratio
0.857
0.604
0.294
0.294
—
TABLE 9. ERT VISIBILITY MODEL PERFORMANCE ANALYSIS OF 5 CASE
STUDIES UNDER HAZY-SKY BACKGROUND CONDITIONS-
NORMALIZED RESIDUAL STATISTICS
Plume/Sky Intensity Ratios;
PIume Contrasts
(Wavelength--nm)T
Performance Measure
Measured mean value
Predicted mean value
El
Si
E2
S2
405
0.874
0.850
0.028
+0.028
0.204
-0.201
450
0.862
0.787
0.087
+0.087
0.58
-0.58
550
0.944
0.680
0.279
+0.279
5.578
-5.578
630
0.974
0.635
0.348
+0.348
8.384
-8.384
Blue/Red
Ratio
0.885
1.234
0.406
-0.406
--
E^ and S} correspond to plume/sky intensity ratios, and
plume contrasts.
and $2 refer to
29
-------
The correlation coefficient for measured and predicted values for the
overall model evaluation under clear-sky background is very low in the
blue light range--0.066 and 0.099 at 405 and 450 nm, respectively; average
(0.587) at 550 nm, and negative (-0.243) at 630 nm. These results suggest
that there is little overall correlation between telephotometer measure-
ments and model predictions. This lack of correlation can be caused by
various components of the model, i.e., the treatment of dispersion,
chemistry, or optics, or any combination of these elements. A negative
correlation coefficient of -0.241 is calculated for the blue/red ratio.
Other statistical performance measures of particular interest are the
bias and gross errors. Ue discuss here the values of the normalized bias
($1 and $2) and gross errors (E^ and £2), though this type of analysis
could be applied to the normalized statistics as well. The average
absolute normalized errors in plurre/sky intensity ratios are 8.2, 6.5,
4.4, and 8.1 percent at 405, 450, 550, and 630 nm respectively. These
values are comparable to the performance values of the PLUVUE, which were
calculated to be 8.3, 8.8, 5.4, and 5.6 percent at these same wave-
lengths. The average absolute normalized errors in plume contrast are 42,
37, 30, and 262 percent at 405, 450, 550, and 630 nm, respectively. For
comparison, the errors in plume contrast for PLUVUE were 69, 83, 69, and
154 percent at these same wavelengths.
It appears from the values of the bias that for clear-sky background
cases, the ERT Visibility Model tends to underpredict plume visual
effects, whereas PLUVUE tends to overpredict these effects. To determine
the cause of the overall behavior of the ERT Visibility Model, it is
necessary to individually evaluate the dispersion, chemistry, and optics
modules.
Only two cases having dark-mountain backgrounds were studied—too few
to provide much statistical inforrration regarding model performance under
these conditions. However, it is of interest to qualitatively evaluate
the capability of the model to reproduce the main characteristics of the
telephotometer measurements. The values for performance measures Ej, S2,
30
-------
E£, and $2 are presented in table 8 and for these two case studies, model
performance is poor. A primary source of uncertainty in these simulations
is the reflectance of the background object. The model shows good results
for case study 5, predicting a darker plume at 400 and 450 nm and a
brighter plume at 550 and 650 nm, both results that are in qualitative
agreement with the measurements. Agreement is not as good for case study
7, for which a brighter plume was measured but a darker plume was pre-
dicted at all wavelengths except 630 nm.
For the overall evaluation, 5 case studies corresponded to hazy-sky
background conditions—al so an insufficient number with which to obtain
much statistical information regarding model performance. The values for
performance measures E^, S^, ^2> anc' $2 were calculated and are presented
in table 9. The poor model performance indicated in the table is probably
caused in part by the use of a light gray reflector to simulate the
background haze layer.
31
-------
APPENDIX A
EVALUATION OF THE MODULES OF PLUVUE
A.I EVALUATION OF THE DIFFUSION MODULE
The diffusion module of the plume visibility model is based on a
Gaussian formulation of the dispersion of nonreactive pollutants. Several
options are available for defining the horizontal and vertical dispersion
coefficients: the stability class can be specified and the coefficients
can then be computed from the Pasquil1-Gifford-Turner (PGT) or the TVA
dispersion curves; or the dispersion coefficients can be specified
directly at specific downwind distances.
The dispersion curves of PGT have been deduced from empirical
correlations of experimental data obtained from a release of tracers near
ground level over flat terrain. Thus, one should not expect these
empirical formulas to hold for elevated emission sources over complex
terrain. However, since the PGT dispersion coefficients are likely to be
used when plume visibility simulations are performed to assess the
possible impact of new power plants or smelters on atmospheric visibility,
it is of interest to present a brief comparison of the PGT dispersion
estimates with the measurements obtained for the plume of the Navajo
Generating Station.
Plume widths can be obtained from aircraft measurements by consider-
ing the variation in S02 concentrations. For this study, measured S02
concentrations above 20 ppb defined the plume. Corrections were made for
variations in the altitude of the aircraft and for the angle between the
plume centerline and the flight path. A comparison of the plume widths
32
-------
predicted by the PGT dispersion curves with those measured by the aircraft
is presented in table A-l for the cases used in model evaluation (plume
width is assumed to be equal to 4 oy)'. In the three cases for which the
predicted plume width is greater than the measured plume width, the
airplane probably missed the center of the plume; thus, the measured plume
width is most likely smaller than the actual plume width. Among the
remaining cases, 28 June and 13 July are the only days on which the plume
width is overpredicted. It should be noted, however, that the use of SC^
concentration measurements can lead to some uncertainty in the measured
plume width because of the slow time response of the SC^ monitor (Seigneur
et al., 1981).
Because such a comparison of predicted and measured plume widths is
strongly dependent on the assumption of a Gaussian plume model, it is
interesting to evaluate the validity of the Gaussian approximation.
Figure A-l shows the measured NOX concentration profile for the flight
path traverse of 13 July, along with the Gaussian profile, which was
determined from the maximum NOX concentration and the total plume NOX
concentration integrated along the flight path. It appears that for this
case the Gaussian profile provides a fairly good approximation of the
actual profile. However, in some other cases, wind shear distorted the
plume and the measured profiles were then non-Gaussian (e.g., the 15
December case studies).
Other useful indirect measures of plume dispersion are plume concen-
trations. Concentrations of N02 and the value of bsp (particulate
scattering coefficient) integrated along the sight path are shown in table
A-2. The measured values were corrected for the angle between the flight
path and the sight path. These values are of particular interest in
conducting plume visibility model performance evaluation because they are
the input to the optics module. Concentrations of NOX at the plume
centerline are presented in table A-3, along with the corresponding
N02/NOX ratios, which will be considered next in the discussion of
evaluation of the chemistry module. Statistics concerning model perfor-
mance for these variables are listed in table A-4.
33
-------
TABLE A-l. COMPARISON OF PREDICTED AND MEASURED PLUME WIDTHS
Day
28 June
Flight
Number
5
7
Measured
Plume Width
(km)
S(>2 > 20 ppb
0.52
2.80
Predicted
Plume Width
(km)
2.42
4.30
13 July
4 December
7 December
(morning)
7 December
(afternoon)
15 December
1
2
3
4
5
12
2*
3
4
7
8
9t
10
11
1
2
3
4t
5*
7
2
3
7
9
3.70
12.18
9.34
6.79
8.53
18.10
1.86
6.30
2.90
5.11
4.95
2.75
7.56
7.74
8.66
9.22
8.50
0.50
1.19
13.0
10.92
11.10
8.26
15.70
5.73
1.73
2.86
3.25
3.86
2.52
3.53
3.53
2.23
3.46
3.56
2.54
1.88
1.12
1.58
"2". 20
2.28
2.90
3.64
3.61
2.90
2.60
3.65
4.54
Predicted plume widths are assumed to be equal to 4 Oy, where
-------
MEASURED
CONCENTRATION PROFILE
FITTED GAUSSIAN CON
CENTRATION PROFILE
-2000
-1000
0 1000
Distance (meters)
2000
Figure A-l. Comparison of measured and Gaussian-fitted concentrations profiles
for the 13 July 1979 case study.
35
-------
TABLE A-2. COMPARISON OF MEASURED AND PREDICTED INTEGRATED N02 CONCENTRATIONS
AND SCATTERING COEFFICIENT VALUES ALONG THE SIGHT PATH
Day
(1979)
28 June
13 July *
4 December
7 December
(morning)
7 December
(afternoon)
15 December
Flight
Number
5
7
1
2
3
4
5
12
3
4
7
8
9+
10
11
1
2
3
Measured and Predicted Values
;L [NO ] .
o <•
Measured
0.085
0.170
0.170
_ _
0.355
. 0.318
0.384
0.880
0.044
0.158
0.069
0.190
0.174
0.048
0.281
0.274
0.236
0.250
0.234
0.007
0.029
0.482
_ •»
0.415
0.520
0.302
dy (ppm • km)
Predicted
0.566
1.051
0.245
0.573
0.512
0.436
0.358
0.279
0.265
0.268
0.217
0.268
0.284
0.207
0.203
0.183
0.149
0.171
0.177
0.206
0.336
0.352
0.683
0.231
0.470
0.476
( V5!
Measured
0.121
0.200
0.01
0.349
0.154
0.131
0.149
0.438
0.024
0.094
0.067
0.157
0.123
0.024
0.206
0.246
0.097
0.106
0.088
0.003
0.010
0.161
1.409
0.290
0.323
0.337
50 nm) dy
Predicted
1.028
1.192
0.153
0.589
0.411
0.349
0.282
0.304
0.166
0.161
0.182
0.166
0.170
0.166
0.206
0.304
0.176
0.149
0.149
0.387
0.171
0.192
0.737
0.324
0.361
0.364
Sight path from the Kaibito City telephotometer site.
^ The aircraft probably did not fly through the plume center,
36
-------
TABLE A-3.
COMPARISON OF MEASURED AND PREDICTED NOX MAXIMUM CONCENTRATIONS
AND N02/NOX RATIOS
Measured and Predicted Values
Day
(1979)
28 June
13 July
4 December
7 December
(morning)
7 December
(afternoon)
15 December
F1 ight Number
5
7
1
2*
3
4
5
12
2t
3
4§
8§
9t
10
11
1
2
3
2"
3
7
9
Measured
0.122
0.061
0.282
_ _
0.766
0.427
0.144
0.483
0.217
0.251
0.698
1.021
1.288
0.636
1.215
0.926
0.393
0.238
0.209
0.044
0.044
0.223
— .
1.496
0.786
0.457
NOX
(ppm)
Predicted
1.410
0.219
0.105
2.413
1.135
0.819
0.433
0.821
0.456
0.458
0.900
0.478
0.460
0.752
1.189
2.632
1.286
0.780
0.744
0.518
0.373
0.450
0.605
0.712
0.435
0.317
[N02]/[NOX]
Measured
0.402
0.361
0.181
„ _
0.111
0.124
0.278
0.128
0.143
0.175
0.138
0.112
0.092
0.113
0.091
0.075
0.117
0.189
0.211
0.295
0.455
0.206
•. mm
0.123
0.155
0.184
Predicted
0.191
0.210
0.400
0.227
0.272
0.258
0.222
0.173
0.184
0.188
0.147
0.184
0.189
0.152
0.125
0.082
0.103
0.131
0.134
0.162
0.196
0.196
0.183
0.173
0.209
0.237
NO/NOX measurements are unavailable.
' The aircraft probably did not fly through the plume center.
§ Uncertainty in the NO/NOX measurements; N02 measured value is likely
to be too low.
37
-------
GO
UJ
1
— ^
Q
O
>-
oL
L, „
r~"
OO
!__(
2:
UJ
"T"
O
Q
Z
^c
z
0
»— 1
to
C£
LU
Q_
to
^H
0
UJ
33
^>
— i
— 1
O-
u.
o
GO
I_H
GO
>-
1
Z
UJ
t_3
^^
^£
y
D£
O
Li—
at.
LU
Q.
,)
"sS"
1
eC
UJ
1
CO
^£
1—
GO
Z
o
t— 1
L -.
O
z
o
o
z
o
a;
o
o
^c
CO
^.
XX
GO
1
QL
^£
UJ
i
0
OL
UJ
O
z
—^
GO
UJ
(— t
^2
— s
1—
GO
UJ
GO
^c
O
LO
t-H
Q,
O
LL.
*
>^
-O
•
O
co
.a
*— *
LO
i— l
«
O
4-
>~
-^
i — i
CM
O
Z
1 1
1 — t
LO
CM
•
O
4-> D
re c
OJ •.-
x E r-
031-
Z i— OJ
***•«• ^ 4 >
CM C
O OJ
Z 0
^~
I— t
•
o
-f-
c:
(LI O
0) C T-
E •*"" •!-*
3 i — ro
i— t_ t-
o_ ai 4->
-•-> d
x c a>
O OJ O
z o c
o
c ^
CO
LO
0
*
^*
I *
-o
3
QJ
^3
^~
Q-
^>
CM
•
00
ai
t.
3
OO
re
QJ
•s.
QJ
O
C
re
E
C_
0
(_
QJ
Q.
(U
r_
ro
c >
o
•r- C
oo re
oo O)
0) E
i_
CT ^3
O) QJ
ai t_
3
t- 00
re re
QJ O)
c. ^.
•r~
_J
CM
CM
•
O
LO
CM
•
O
oo
•
o
oo
r— •
•
O
cn
en
•
CM
01
3
re
c
to
01
•^3
QJ
O
•o
OJ
O-
CM
rH
cn
4>
O
0
LO
ro
•
0
t-H
0
•
0
LO
CM
CM
o
en
o
o
•
o
•4_>
c
OJ
• ^
o
•t—
14-
14-
OJ
0
o
c
o
>^-
-!->
re
QJ
^.
0
o
co
U
•r-
| '
CO
•r»
^_>
re
i <
GO
re
3
•o
•r"
co
O)
0£
«^
LO
O
4J>
O
1
LO
O
0
o
i
CM
CO
0
o
1
^-
0
•— 1
•
o
1
0
LO
CM
•
LO
CO
re
«^
CO
^*
LO
O
9
0
-H
,_<
r^N.
O
•
O
+1
CM
^-
0
o
+1
^^
vo
ro
•
0
+1
1— 1
p^
LO
•
.— 1
+ 1
^_
re
>
u
OJ
4_>
C
•r—
c_
0)
.,_
HI
o
c
QJ
^3
• r-
14-
C
o
o
O)
u
c
re
•i—
c_
re
>
01
3
t_
1—
LO
LO
0
«
o
LO
r— 1
! 1
•
O
LO
p**
0
0
LO
LO
LO
O
CO
ro
•
LO
L_
O
L.
t_
QJ
a>
c_
re
3
CT
1
C
re
QJ
f
jj_>
O
o
C£
.c.
o
JZ.
C
•r—
00
OJ
oo
re
o
•
CO
co
P_
re
c
re
OJ
e~
•*->
C
T3
QJ
CO
3
QJ
!_
QJ
•*
CO
QJ
00
re
o
•o
c
3 •
0 T3
{_ OJ
cr> -a
n t-
LJ re
ro o
-Q oo
.^-
>> -o
co OJ
1 (_
t- OJ
ro 2
O)
i— OJ
o c
QJ -^
x: t_
•(-> O)
i »
E c
t- 0
*i
"O E
O) 3
•(-> 1—
O 0.
QJ
r^ OJ
0) J=
CO 4_>
** "O
co OJ
OJ 00
•i- 00
TJ ••-
3 E
oo ai
c
O) re
CO , —
re Q-
u
a>
LO J=
i-H JJ_>
•K
x:
0
•r—
c~
2
C
00
QJ
CO
re
£_3
•
OO
co
^_
ro
c
ro
QJ
e—
4->
c
•r—
-o
QJ
00
3
O)
C-
1
m
CO
OJ
CO
ro
o
-a
c
3
O
c_
C71
NX
O
ro
JD
>,
CO
1
L.
ro
O)
f^
o
QJ
C"
4->
t_
T3
QJ
-(->
U
OJ
r-~
QJ
oo
M
00
0)
•r—
-o
3
4_>
CO
QJ
CO
ro
o
<^-
r— i
4-
•
T3
QJ
T3
L.
re
0
CO
-o
OJ
i_
QJ
QJ
f-i
ro
>r_
re
>
ro
QJ
c_
QJ
*
00
c:
QJ
OJ
£_
3
CO
ro
OJ
E
X
0
z
o
•z.
o
c
c**
O
>r_
^
5
c
O
-o
C
ro
a>
c
in
t_
QJ
4_3
r*
OJ
o
QJ
E
3
^~
D.
O)
JC
4->
T3
QJ
CO
00
•^
E
QJ
C
re
'a.
QJ
^
4_>
38
-------
The following conclusions can be drawn:
> Plume widths are generally underpredicted by the model.
Cases for which the plume width is overpredicted corre-
spond to those for which the aircraft probably missed the
center of the plume; thus, the measured plume width is
most likely smaller than the actual plume width. Among
the remaining cases, 28 June and 13 July are the only days
on which the plume width is overpredicted.
> The NOX plume concentrations at plume center are generally
overpredicted by the model, which is consistent with the
underprediction of plume dispersion. In some cases,
however, the NOX plume cncentrations are underpredicted,
suggesting that vertical plume dispersion was highly
overpredicted or that the concentration distributions were
non-Gaussian for these cases (on 13 July when plume width
was overpredicted, and cases 7, 8, and 10 on the morning
of 7 December, and all cases on 15 December).
> The correlation coefficient for the plume width is an
extremely low value of 0.09; the correlation for the NOX
plume center-line is better, with a coefficient of 0.225.
The treatment of horizontal plume dispersion is not
critical in plume visibility modeling because the horizon-
tally integrated plume values are used as input to the
optics module. Horizontal plume dispersion thus affects
model prediction mainly through the description of plume
chemistry, which depends on the mixing of plume concentra-
tions (e.g., NO) with background species (e.g., 0^).
> Tables A-l, A-2, and A-3 show that the agreement between
measured and predicted plume integrated values is actually
better than those for plume widths and NOX plume center-
line concentrations. The correlation coefficients are
39
-------
0.360 and 0.912 for the NO;, and bsp integrated plume
values. Overall, the N02 "ntegrated plume values are
slightly underpredicted by the model, whereas overall, the
bsp integrated plume values are overpredicted.
A.2 EVALUATION OF THE CHEMISTRY MODULE
The chemistry module of PLUVUE involves nine reactions among NO, N02,
03, 02, S02, OH, H20, and O^D) (Latimer et al., 1980). This mechanism
describes the main chemical process.es occurring in power plant plumes in a
clean background atmosphere. If notable concentrations of reactive
hydrocarbons are present in the background, a more detailed mechanism
should be used (e.g., Stewart and Liu, 1981). Hydrocarbon concentrations
in the vicinity of the Navajo Generating Station were low enough to apply
the plume visibility model mechanism for a clean atmosphere (Sheppard et
al., 1980).
Since N02 absorbs light in the blue range, it is important to compare
the predicted and measured rates of conversion of NO to N02. Oxidation of
NO to N02 occurs both by molecular oxygen in the high-temperature flue
gases and by ozone in the atmosphere.
It has been shown (Bergstrom et al., 1981) that the molecular
oxidation of NO to N02 by 02 accounts for an N02/NOX ratio of about
5 percent. This agrees well with the values of 3 to 6 percent deduced
from the airborne measurements (Richards et al., 1981).
Because oxidation of NO by Oj in the plume is limited by the mixing
of the plume with ambient air, the validity of the NO-N02-03 model depends
to a great extent on the description of the plume dispersion processes.
It can be evaluated by comparing the value of the N02/NOX ratio predicted
by the model with that measured in the plume. The predicted and measured
N02/NOX ratios are presented in table A-3 for the 27 case studies consid-
ered in the plume visibility model evaluation. It is appropriate to
40
-------
discard cases in which the aircraft most likely did not fly through the
plume center (cases 2 and 9--7 December morning; cases 4 and 5--7 December
afternoon) and cases showing uncertainty in the N02 measurements (cases 4
and 8--7 December morning). For the remaining 18 cases for which NC^ and
N0y measurements are reliable, there is good agreement between observed
/\
and predicted values because the average absolute relative error of the
model is 45 percent. There is some bias toward overestimation of the N02
formation rate since the bias is -0.032 and the predicted N02/NOX ratio is
larger than the measured value in 11 out of 18 cases.
Plume chemistry measurements were also carried out by the MRI
aircraft during one-hour orbit flights in the plume ranging from downwind
distances of 25 to 100 km. It is interesting to evaluate the chemistry
module with these time-averaged data because they were obtained at larger
downwind distances than were those considered for the plume visibility
model evaluation. To compare model predictions and measurements, it is
necessary to make some assumptions for calculations of the plume average
concentrations. The aircraft measures plume concentrations for one hour
while performing an orbit in the plume. Since plume concentrations vary
according to the location of the air sample in the plume, a procedure for
calculating the plume averages from the model predictions must be
defined. The plume generally becomes well mixed vertically at large
downwind distances, and the uncertainties in the plume-concentration-
averaging technique should not have a significant effect on the validity
of the comparisons. In this study, the plume model averages were defined
as the crosswind concentration average at the height of the plume center-
line, and the plume dispersion was adjusted so that the average NOX
concentrations fit the measurements.
Table A-5 presents a comparison of the predicted and measured plume
average N02/NOX ratios for 10 cases selected from the December 1979 VISTTA
field program. The agreement between model predictions and airborne
measurements is good, especially when one takes into consideration the
uncertainties associated with the averaging techniques. The average
absolute relative error between the predictions of the plume visibility
41
-------
model and the measurements of N02/NOX is 25 percent. The average signed
relative error is 19 percent. These results show that the N02/NOX ratio
is slightly overpredicted by the model.
TABLE A-5. COMPARISON OF MEASURED AND PREDICTED N02/NOX RATIOS—
ONE-HOUR TIME-AVERAGED PLUME CONCENTRATIONS
Concentrations
(pprn)
Day
(1979)
5 December
9 December
12 December
13 December
Downwind
Di stance
(km)
33
80
30
84
45
77
93
25
65
82
Measured
0.35
0.71
0.20
0.62
0.23
0.48
0.60
0.23
0.48
0.68
Predicted
Plume Vi sibil ity
Model
0.30
0.61
0.30
0.71
0.36
0.65
0.68
0.29
0.60
0.67
Predicted
Reactive Plume
Model
0.33
0.73
0.29
--
—
Three case studies were also modeled by the reactive plume model,
which involves 70 reactions among 35 chemical species and provides a
fairly detailed treatment of the chemical interactions between the plume
and the background (Seigneur, 1982; Stewart and Liu, 1981). The results
shown in table A-5 indicate that the plume visibility model with its nine-
step reaction scheme compares favorably with the more detailed reactive
plume model. This good agreement between model predictions and measure-
ments suggests that the nine-step chemical mechanism of the PLUVUE is
suitable for describing plume chemistry in clean environments in which
reactive hydrocarbon concentrations are low.
Treatment of plume dispersion is likely to be the primary source of
uncertainty in predicting chemical rates, because these rates depend on
42
-------
the mixing of the plume material with the ambient air. Modeling the
interactions of plume dynamics and chemistry, a subject mentioned by
Bergstrom et al. (1981), requires further investigation.
A.3 EVALUATION OF THE OPTICS MODULE
The optics module calculates the plume's radiative properties and
their optical effects. This module was evaluated using the plume concen-
trations of NOo and particulates measured by the airplane along a flight
path aligned with the telephotometer line of sight. The best data were
obtained on 7 December, the day on which 24 telephotometer sight paths
were flown. The comparison of measured and predicted values for the 10
remaining flights are shown in table A-6 for the four wavelengths.
The results of the optics module performance evaluation are shown in
table A-7. There is good correlation between measured and predicted
plume/sky intensity ratios at 405, 450, and 550 nm, with correlation
coefficients of 0.697, 0.843, and 0.768, respectively. There is a
negative correlation (-0.210) at 630 nm; however, because plume contrasts
are small at that wavelength, predicted and measured values are sensitive
to uncertainties. Good correlation (0.888) is obtained for the blue/red
ratio. Values of the bias show that PLUVUE, on the average, overestimates
the plume optical effects at all wavelengths.
43
-------
Q
1
tn
CJ
Cu
o
ac
>•
in
•z.
Ld
_J
O.
Q
LJ
ce
o_
o
LJ
a:
ce
«
O-
I
SO
I
to
I—I
o.
t— 1
u
CO
co
CO
.0
o
^^
g
c
1
,3
0-
•D
c
s
X
O
O
CO
U
•H
•o
Q)
l-i
Q.
CO
D
CO
CO
3:
CO
u
•H
•D
CO
Ui
0.
-D
CO
^
co
CO
•o
CO
u
•o
CO
u,
Q.
•O
CO
CO
CO
CO
u
H
T>
CO
f_,
o.
CO
3
CO
(D
CO
•D
5
o
cr
u
CO
CO
CO
0
c
CO
to
H
Q
CO
O
c
CO
4-J
CO
•H
O
CO
E
H
O
lA
SO
""*
O
PA
SO
3
lA
lA
*""*
0
lA
LA
O
lA
•^
~~"
0
lA
O
-1*
""'
^
LA
O
CO
g
H
I i
H
TJ
O
J
Sf
^-^
E
^
h—
to
^
CO O
Os Os
o o
PA -^
r~ IA
o a
03 PA
T— lA
OS Os
O O
CM -*
ON ON
0 0
CO CD
KN O
CO ON
0 0
\O lA
CM «—
ON ON
0 0
SO GO
CO ON
o a
lA -3
CM OS
ON CD
0 0
o r-
SO «-
1A 1A
CM r—
,_ ^_
CM ^~
OS ON
CM
0 O
0 0
so t~
r-- r~
o o
t- PA
lA f^
CO l~-
o o
ON PA
T- SO
r- CM
O 0
lA O
CSI CM
T- CM
CO O
PA -3-
. r-
co os
Q «-
CD "H
O C
& 8
-i
0)
O
CO
44
-------
oo
UJ
A
— ^
1 —
oo
1 1 1
oo
C_J
o
t— t
t 1
o
OO
00
>-
1
2:
*f
1 1 1
t_3
2;
<:
^
Q£
o
1 1
UJ
Q.
UJ
1
— ^
Q
y
oo
o
(—
Q.
O
•
I
UJ
QQ
(—
4t
to
4J
to
03
4-J
C
O
o
Ol
E
3
o
• «\
00
O
•r—
+J
TO
Q£
>^
^j
*r—
t/1
C
0)
c:
•— t
\^
oo
*""»•.
OJ
E
3
r—
D-
T3
CO O
OC -r-
"* — 4-J
d) (O
3 QC
,_
co
oo
00
o
^~>
E
c
1
1
JZ
1 *
en
c
a;
^—
QJ
>
^j
2
*. — •*
o
ro
LO
ro
O
•
o
1
O
LO
LO
oo
o
•
o
1
O
LO
^}-
LO
1— 1
•
o
1
LO
O
^*
.— 1
•
0
1
CO
c.
3
to
03
O)
s:
QJ
U
c
03
o
14-
^
ai
Q-
0)
3
r—
03
c >
o
•r- C
00 "3
tO <1>
0) E
t_
cn T3
0) (!)
QC l-
3
L_
C
cu
0) -i-
3 O
r— *r-
T3 tt_
> **—
TJ
O . —
•r- dJ
T3 t-
d) t.
^«S
oo
O
•r-
4^
t/)
•^>
4J
t)
4.)
OO
^_
O3
3
•o
to
ai
C£.
01 oo oo
O rH O
CM «^- O
O O O
• • •
o o o
+ -H
LO LO CTi
•—I LO O
*3- r-i 0
O O O
• • •
O O O
-H
ro o\ LO
«d- LO o
CM CM O
000
• • •
0 O 0
1 -H
r— H ^^ ^"i
LO *3- O
LO •!)- O
000
o o o
-H
r-H LO ^~
en oo *-i
^1- ro o
0 O 0
• • •
000
-H
r—
fa
c.
0)
c
•r"
a>
u
c
CM
O
O
• ro
O ro
i O
^" •
O 0
o
0
•
o
ro
CM
o
o
• o
O LO
1 0
LO •
0 0
o
o
•
o
ro
*-H
o
o
• CM
O ro
1 0
ro •
0 0
o
0
t
0
LO
CM
o
f">
• CM
O LO
1 O
un •
0 0
0
0
0
f**.
ro
o
0
• 0
O LO
1 O
r*>* •
0 0
o
0
•
o
'—
>
(_
ai
4_>
C
•p™
dJ
0
> E
I
a> 4->
3 O
1- O
h~ Of
l/l
(_)
«P>
^_j
to
•r-
O3
1 %
OO
t
(O
3
T3
•r*
to
dl
a:
^^
dj
Nl
>r.
^_
OJ
E
O
i — 1 LD
ro CM
o o
• •It
O O 1 1
oo ro
O O LO LO
• • • •
O O r-4 rH
1
^^ r***
ro CM CTI ro
O o ro ro
• • • •
0000
^~ ^*
LO LO LO LO
O O «i- ^d-
• • • •
o o o o
t
o vo
LO LO r^ LO
o o *i- «*
o o o o
4^. 4_
t— 1 i— 1 CM CM
UJ OO UJ OO
1 »
o
40
ai
rH JZ
OO 4J
T3
C C
03 O
•r-
UJ OJ
f—
3
tO CJ
4-3 r—
to 03
03 O
L. •
4-> IO d)
O OO 4->
O 03
e E
OJ 4-> O
EC t-
30 <*-
r— O
QL -a
dJ CU
dJ E "O
J= 3 C-
4J , — 03
a. o
t_ to
00 ••-
"^ E
T3 T3 OJ C
d) C t-
4-> O CU O
to o_ 5 ro
•r- 00 LO
f— d) CM
t- o -o
CU t- • C
I- 0 0 03
O3 O
c •<
IO CM OS O
O 00 J= LO
•i— 4-> LO
4-> "O
IO C tO *
•r- 03 tO O
4-> Ol LO
03 CM r— ^3"
4-> UJ
00 <4— «
•o o LO
r- C O
03 O3 •(-> «*•
3 to
*^7 ** 'O | ^
•r- 00 £-O3
tO O 4-*
dJ -r- C 00
t- 4-> 0
O3 O -O
•o c. c
dl d) (tJ
.^! *» E
-^ -r- ^ O
O3 to Q. t— I
E C
t_ dJ d> «
O 4-> 4-> O
C C 3 rH
C 't- i —
o o •
c >, to o
T3 to 03
c •*-. a;
IO dJ "O t-
E d) d)
C 3 t- 2
Or- 3
•r- Q. to 00
tO 03 d)
to o dJ to
dJ i * E 03
f- O
O) C 4-> <4-
t- O -r- O
a. S
t- to c_
03 cu oo dJ
aj t. ai ja
c t_ to e
•r- O 033
_j o o c
# -t-
45
-------
APPENDIX B
EVALUATION OF THE MODULES OF THE ERT VISIBILITY MODEL
B.I EVALUATION OF THE DISPERSION MODULE
The dispersion module of the ERT Visibility Model employs a Gaussian
formulation of the dispersion of nonreactive pollutants. In this section,
we evaluate the dispersion module using a comparison of selected measured
and predicted plume values. Values that can be used to evaluate the
adequacy of model treatment of plume dispersion include plume width and
plume NOX concentrations at centerline: plume concentrations at the
center-line of the plume are an indirect measure of total plume dispersion.
In addition to plume widths and plume centerline concentrations,
plume concentrations integrated across the plume are also of interest
because they provide an indirect measure of plume vertical dispersion.
Table B-l compares measured and predicted values for plume widths,
NOX concentrations at plume centerline, the N02/NOX ratio at plume
centerline (which is used in the plume chemistry module evaluation),
integrated N0£ concentrations, and the particle-scattering coefficient
along the line of sight in the plume. The results of a linear regression
analysis of these values along with residual statistics, are presented in
table B-2. The results presented in these tables suggest the following
conclusions:
> Plume widths are generally underpredicted by the model.
Cases for which the plume width is overpredicted corre-
spond to cases in which the aircraft probably missed the
46
-------
in
«*•
_)
tf
>•
%
_i
Q-
8
ce
S
^
0
I
8
t_>
Q
UJ
C
Cv
8
L3
u
_i
Ul
in
U_
o
^
1— I
cc
0
H
V
0.
~v
D1
00
to
2
t!
• H
"S
^J
0-
^
5
03
co
S
"S
CD
4-1
u
-H
"D
O-
-D
S
CD
(0
2
u
• l-t
0)
O.
•o
£
0)
CO
5
CJ
«
u
0.
"8
3
09
CO
i
z.
s~*
^
lA
•^
r^
«-
^
o
CM
a
o
lA
CO
o
o
^
o
o
CM
o
o
co
CM
CM
CM
CM
0
CO
,_
CM
lA
o
lA
o
"-J
CO
CM
LA
LA
^
o
o
CM
o
co
*
o
o
o
o
CO
o
,_
£
o
ON
CM
o
t—
°
o
o
•a-
o
CO
CM
r-
K\
VO
C3
o
o
o
CM
o
o
o
f^
CO
o
_
CO
o
o
^
CD
CM
CO
CM
o
CM
lA
lA
o
r^
^
^
*
1— I
3
«—
r^
CD
o
ON
1^
o
CM
o
1
1
CM
o
o
]
^
°
K\
1
1
LA
,_
00
*—
CM
^
h
CD
u
&
•&
N^OCOCM ^•CfCM^OlA'AGD
lAr^tA^— cOK\c^*~ONr^LAr^
ONVOCOCO <-VOCMlA.*CMt^r-
4- 4-
I* O>
f-H
2 i
a *^
r~-
*^ CM ON
CM CM T-
o o o
r- so oo
O T- O
o o a
£ p £
o o o
CM CM CM
o o o
»- r~ ON
t~* on CD
o 5 o
000
f- OX "-
r- CO <~
o o o
-------
•o
CO
•c
3
o
c
CD
1
(4
CO
4J
c
&
E
t-4
0.
o
z
]
*
*
>,
•0
r-n
0~
z
•^
i
3
*H
D-
4J
CO
ox
z
CM
O
z
c
o
•H
4J
co
(4
4J
C
CD
U
C
o
CJ
< CD
C
• H
•-H
CO
JZ
=g
•H
3E
i
.— t
o_
t
>>
^
•
^
o
lA
lA
Q
^
X— v
E
J*
1
§.
^Q.
Q)
C
•-(
1— (
(-1
(U
.*-»
c
8
'E
Q.
Q.
^_^
E
J£
•o
CO
4->
U
• H
•o
co
u
0-
•o
CO
9
0)
CD
i
*
4_>
U
• H
•o
CO
(->
O-
•n
OJ
9
03
CD
£
•g
4-1
CJ
• H
-o
CO
|4
O,
•g
9
0)
CD
£
"g
4J
CJ
• H
•D
co
tl
O-
1
U
D
03
CD
i
•g
J->
o
•H
T3
CD
U*
a.
•g
€_(
5
CO
CD
£
4-> (4
-C IB
Dl J3
2 i
U- Z
X— X
ON
>. f-
co o
Q *-
§
T^
0
O
CM
co
o
•
o
KV
CM
•
O
r-
co
CM
-
vO
s
t—
CNJ
r^N
0
^
K^
•
o
0
CM
lA
O
CJN
1A
^—
O
lA
LA
*—
•
o
CO
lA
KV
d
\o
CO
r^
•
o
vO
KV
i
1
§-
»H
CO
4J
>s
4->
CJ
o
4->
• H
JS
• H
a
:*:
CO
jr
4J
§
l-l
U_
jr
4J
CO
a
4J
JZ
D»
•H
01
*
<_,
CO
4-)
c
g
§
^H
D.
CD
JC
4->
f
O
(4
j:
4J
>,
rH
U-
4J
O
c
•o
•o
>s
JD
TB
JO
o
H.
4J
U-
0
h
CJ
(4
.^4
co
CO
JZ
t—
4-
CD
r-H
J3
CO
•— (
0
>
co
c
3
CD
L,
CO
CO
4J
C
1
U
3
CO
CO
i
X
0
z
~-,
o
u*
rections were made when the flight path did not correspond to the
u
8
£
4->
CO
0.
4->
•&
CO
CO
r.
4->
cn
c
0
I-*
a
i
3
•-H
O.
CO
JT
4J
C
• H
T3
CO
4->
CD
U
&
CD
•s
-H
CO
3
^H
O
>
*
*
•
JZ
4J
CO
a
4->
£
ni
CO
48
-------
in
UI
H-(
o
rj
h-
en
UJ
S
CO
lf\
u
o
in
in
>-
i
z
<£
a
or
o
u.
ce
UJ
°- in
in z
LJ O
_J 1— 1
3 1—
O 1-1
i §
S S
K i
i— » 13
1 1
0 B
1 S
§ ft
I-H 1
in cc
ce «t
UI UI
81 d
*— t
O cc
. ui
»— o
cc z
UI =1
CM
1
CO
LJ
d
t—
ux
•g-
•
D.
CO
XI
*— >
+-
>.
•D
f— i
CM
O
Z
t"
^~i
K>
Sfl T- -^
<- r^ CO
d o o
l/S
\c *— ^
CM CM I*"!
• • •
o o a
4-
CO
C
• H
4-> -H
CD b,
CD
X Z
S c
^ cS
Z £
•— t
O-
^ CM VD
V— «— ^
d o o
4-
C
O
• H
•B
bi
Jj
§c
cu
= 1
X
0 CD
Z C
•H
i-H
U
CO
J-l
c
cS
ten
£
*
.H
f
i
t— <
a.
o
CD vO >A
VO T- CM
0 r CD
o
<» r^ CM
CM r- O
CO CM O
CD
*
Performance Measi
.*J
Linear Regression
Measured mean value
Predicted mean value
Correlation coefficien
o
o
*— Ov *— O
o in •» i CM
\r\ \o o -sr vo
v- O O CVI «-
• • • o •
o o o o o
1 Tl d
vo
«- o o CM
K\ CM CM O 00
ir\ \o «- • *"
o o o o *-
• • • 1 •
O O O O CD
+' S
o
o
GO
to
T— QV ^ T—
•* •* vO O °
CM ^ O • 00
O O O O O
• * • 1 '
o o o \o o
•kl K\
§
Q
CM
VO O «- CM
VO 00 CM •
CO CM CD K\ r~
<* VO •* 1 CM
. • . I~- •
0 0 «- r- •<-
i T! ^o
CO
o
0
f~- lA (Jv
O ov 1^ CM VO
ir\ vo vO 1 u>
l/\ r- K\ O VO
CD
co
^
d)
jj
c
"~i
•— ( CO
co u
Residual Statistics
Bias
Bias confidence interv
Variance
True variance confiden
Root -mean-square error
5 « ^
/— v/l
^J il vi*
4J
C CD
CO C t-H
JT CD o
x -c «
X 4J
CO
CD CO CO
CO CD CD
5 CO CO
5 ~
*c
. CO
8« "0
CO tj
CD CO CO
0 2 ^
o co co
CJ H
T! T3
C ^
gc co
3 t-i
i-> o co
& (-1 X
v CT
1 * ••
J3 CO -H
* «
J£ >^ ^
CO JS£ -rt
CO ID
|j >
CO bi CO
co co
rH CD CO
t) "-!(•!
U CD
CD S
£ <»
^j £ CO
JJ JJ
§ B I
(-t O E
U_ U ffi
•o . ^ *
0) /-v T: 03
jj \O (DO)
8 „ ' 15 1
i-H ^H CD
CO r> "1 X
CO "8 CO 0
.jj CO Z
CO ^
P CO CD 0
CD CD M Z
X CO CD
~ X 0
S-. C
T3 E
CO CO CO £
CO "O CO O
o t, co -i-i
co CJ p
. H *
CO .rt . I-i
•H -r> CO O
CO -H C
>, CD CO
-H t. >, T!
CO m •-< C
CX CO CO
co c
CD CO CO
1 .5 « 5
4J i-H JZ ^H
b. *J b,
C CO CO
•H •£ c c
•0 CO TJ CO
CD u CO 0
CO 05
3 u ^92
CO D CO 3
bi i-H bi -H
CO Q. CD Q.
* CD * »
CO £ CO £
CD jj CD 4->
•H -H
•O 'C "D "D
3 CD 5 CD
*J CO J-> CO
CO CO 00 CO
• -a • e
CO CO
CO CD CO O
CO c 0 C
CO . CO
l/V ^H -^ «H
>- a «- a
«7» H-
49
-------
center of the plume; thus, the measured plume width is
most likely smaller than the actual plume width. Among
the remaining cases, 28 June and 13 July are the only days
on which the plume width is overpredicted.
> The NOX plume concentrations at plume center are generally
overpredicted by the model, a result that is consistent
with the underprediction of plume dispersion,. In some
cases, however, the NOX plume concentrations are under-
predicted, suggesting that vertical plume dispersion was
highly overpredicted or thai; the concentration distribu-
tions were non-Gaussian for these cases (cases 7 and 8 on
the morning of 7 December; all cases on 15 December).
> The correlation coefficient for the plume width is an
extremely low value of 0.02; the correlation for the NOX
plume centerline is better, with a coefficient of 0.250.
The treatment of horizontal plume dispersion is not
critical in plume visibility modeling because the horizon-
tally integrated plume values are used as input to the
optics module. Horizontal plume dispersion thus affects
model prediction mainly through the description of plume
chemistry, which depends on the mixing of plume concentra-
tions (e.g., NO) with background species (e.g., 03).
> Tables B-l and B-2 show thai; the agreement between
measured and predicted plume integrated values is actually
better than those for plume widths and NOX plume center-
line concentrations. The correlation coefficients are
0.325 and 0.843 for the N02 and b$p integrated plume
values. Overall, the N0£ integrated plume values are
slightly underpredicted by the model, whereas overall, the
bs_ integrated plume values are overpredicted.
-------
B.2 EVALUATION OF THE CHEMISTRY MODULE
The chemistry module of the ERT Visibility Model is based on the
chemistry of NO, N02» and 03. Because N02 absorbs light in the blue
range, it is important that this evaluation compare the predicted and
measured rates of conversion of NO to N02» Oxidation of NO to N02 occurs
both by 02 in the high-temperature flue gases and by Oj in the atmosphere.
Since oxidation of NO by 0^ in the plume is limited by the mixing of
the plume with ambient air, the validity of the NO-N02-03 chemistry module
depends to a great extent on the description of the plume dispersion
processes; it can be evaluted by comparing the value of the N02/NOX ratio
predicted by the model with that measured in the plume. The predicted and
measured N02/NOX ratios are presented in table B-l for the 27 case studies
considered in the visibility model evaluation. The two case studies of 13
July are listed as only one in table B-l, because they correspond to the
same flight path. For this evaluation, it is appropriate to discard cases
in which the aircraft most likely did not fly through the plume center
(cases 2 and 9—7 December morning; cases 4 and 5—7 December afternoon)
and cases containing uncertainties in the N02 measurements (cases 4 and
8--7 December morning).
There is a slight trend toward underestimation of the N02/NOX ratio
since the bias is 0.024 compared with the mean measured value of 0.14. As
indicated in table B-l, most cases show good agreement between the
measured and predicted values of the N02/NOX ratio. It is difficult,
however, to evaluate individual uncertainties associated with the chemis-
try module because of its close interaction with the dispersion module.
B.3 EVALUATION OF THE OPTICS MODULE
The optics module of the ERT Visibility Model was evaluated using
10 case studies from 7 December. The plume and background concentrations
measured by the aircraft along the sight path were used as input. The
51
-------
values used were NC^ concentrations, and scattering coefficient values at
550 nm. Aerosols were assumed to be noncarbonaceous.
Average values were assumed for the background both before and after
the aircraft entered the plume. Measurements of NO^ concentrations and
scattering coefficient values were averaged over 10-second intervals,
which correspond to distance increments of 700 m along the sight path
because the aircraft speed was about 70 m sec . The results of the
10 simulations are presented in table B-3.
The results of the performance analysis are presented in table B-4
for the plume/sky intensity ratios, plume contrast, and blue/red ratio.
The results indicate good correlation between measured and predicted plume
contrast in the blue light range, with correlation coefficients of 0.816
and 0.823 at 405 and 450 nm, respectively. The correlation is less
satisfactory at 550 nm, with a correlation coefficient of 0.482, and it is
negative at 630 nm, with a correlation coefficient of -0.401. A possible
explanation for this poor agreement is that the plume visual effects at
550 and 630 nm are not as important as they are in the blue light range;
thus, the measurements and model predictions are more sensitive to
uncertainties at the greater wavelength. Reasonably good correlation is
obtained for the blue/red ratio. For comparison, the correlation between
the plume contrasts predicted by the PLUVUE optics module and the measured
values were 0.70, 0.84, 0.77, and -0.21 at 405, 450, 550, and 630 nm.
The values of the bias and S^ (normalized bias) indicate that the ERT
Visibility Model underestimates the plume visual effects. The absolute
average normalized errors, E£, in plume contrasts are 46, 34, 21, and 51
percent at 405, 450, 550, and 630 nm. For comparison, the PLUVUE optics
module overestimated the plume visual effects for most cases—the absolute
average normalized errors, E£, were 47, 45, 39, and 164 percent at the
same wavelengths.
Taking into account uncertainties in the measurements and input data,
performance of the ERT optics module appears to be satisfactory.
52
-------
o
1—4
(—
tj
•H
01
O-
•0
9}
ti
3
1C
T>
CO
*J
O
•H
"S
l^
CL
•D
1)
U
3
a
?
o
IA
VO
N_x
^_^
KS
so
'E
c
lA
LA
' '
___
|
f^
lA
lA
^_^
|
CD
lA
-^
^
"E
c
(**!
lA
D
ce
3
«t
o
in
£
o
> a>
t-> •;}
(D U
CO
o JJ
° %
•-H
o.
^3 ttJ
c o
-H C ^—s
S (D E
C -u J*
X 0) >_.
O 'H
& a
*-**
!i
Oi ^
OS
>, f-
O
so
lA so OS KS
SO
CM
CM
N
R
^
o
CO
o
CM
S
CM
O
OS
R
•*
tf\ O
r- B
r~ CM
o
^
S
CM
CM
OS OS O -^ tA <*
CM
53
-------
_
II
LLJ
^^
™
i^—
CO
1 1 1
CO
<_>
r~'
Li*
^^
CO
CO
•~2
z^
*^
^^
LLJ
O
"Z.
^i
•^
Q-
O
Lu.
P^
Lu
Ou
LLJ
— •
•^
L_J
°.
2.
CO
*— '
^^
h—
"-
^j
^~
o£
LU
*
*d"
1
CO
LU
-J
CD
^t
r—
*
t/>
to
*0
^
1 *
c
o
QJ
E
3
r\
• p
to
O
•r—
4_>
re
a:
>.
+->
V)
c
QJ
C
>,
00
^•^
QJ
E
3
a.
-o
O) O
i .(_>
QJ re
3 a:
^~
CO
oo
oo
•
0
o
CO
10
CO
o
•
0
1
0
4- LO
. — • m
E
c
JC.
oo
o
•
0
1
i *
en
c
QJ
r—
QJ
>
re
3 O
in
LO
i-H
•
0
|
LO
0
^-
1— t
e
0
1
o
•r- C
to re
to QJ
QJ E
t.
a> -a
0) QJ
on t-
3
I- tO
re re
^ ^*
>r.
_i
CTt
^~ r^-
• •
O O
1— 1
i-H O
O «d-
• •
0 0
1 1
CM
oo oo
o -cj-
• •
o o
1
CO
CTl CM
O CO
• •
0 0
1
VO
t — r-H
0 00
e f
o o
1
1 1
c
QJ
a> -i-
3 O
re <*-
QJ
C 0
re u
QJ
E c
o
"D ••"
QJ 4->
+-> re
O r—
•i- OJ
•o «-
QJ L.
1- O
QL. <_>
to
u
•^
4J
to
•^
re
1 *
CO
re
3
•o
U)
QJ
a:
^" ^" CM
00 in cvj
ID CO O
o o o
• • •
o o o
1 +1
o co *i-
•-i O O
t— 1 r— 1 O
O O O
• • •
o o o
1 +1
LO r-H 00
I-H 00 O
O i-H O
000
• • •
000
1 +1
CM rH, 00
00 CO r-H
m co o
0 0 O
• • •
000
1 +1
co i— i «d-
oo en i-i
IO CM O
000
• • •
000
1 +1
,__
t3
>
^
OJ
4_>
c
•r—
QJ
O
C
QJ
-o
<4-
c. a>
o o
o c
re
to tn -r-
re re (-
•F- -t~ re
CO CO >
o
o
• r*^
o 10
1 0
CM O
i— 1 .
0 0
0
•
o
CM
T-t
o
0
• in
0 0
1 O
CM O
O •
0 0
0
•
o
o
CM
O
o
• I—-
0 0
1 O
«3- 0
o •
0 0
o
•
o
oo
^f
0
o
. 0
o in
1 O
en o
o •
0 0
o
•
o
p>»
CO
o
co en
. in
0 0
1 O
i — •
0 0
0
•
0
"re
L.
ai
c
f
QJ 4->
3 0
t- O
V— DC
4-
to
0
tp_
4J
to
4J
re
4^
OO
^_
re
3
TJ
•r—
to
QJ
OC.
T3
QJ
IM
f—
re
^
c
o
^
in ^-1
oo oo
o o
• •ii
O O 1 1
<£> CM i— 1 •— 1
I-H 1 — 1 r—4 I— t
o o in in
• • • •
o o o o
f-H CO O I— 1
CM O •— I ^O
o o CM o
• • • •
o o o o
1 1
t-H i— t CM CM
r--. r-- «3- «3-
O O CO CO
• • • •
0 0 0 O
1
O O CM CM
CO 00 IO VD
O O «3" ^"
• • • •
o o o o
1
t^l to^
,— 1 •— 1 CM CM
LU OO LU OO
o
m
VD CM
oo
C T3
re c
re
0 CM
in LU
in
to
•> re
O QJ
un t_
• ^- QJ
to .c
+-> « s
tO O
re o •>
t- <* to
4-> O
o re +->
o re
QJ <-
QJ C-
E re >,
3 4->
r— to ••—
a. c to
o c
QJ -r- QJ
+-> o c
t- -o "~
O QJ >,
l*- t- ^
O. (/)
T3 -^
QJ , — QJ
+-> QJ E
QJ (-
re E 0
-U 0
to 4—
0 i-H
r— OO
re ^:
3 4-> T3
"O CD C
•c- c re
to QJ
QJ r— r-t
t- QJ LU
>
T3 re i/>
QJ SO
•r- O +->
i — -P tO
re ••-
E T3 4->
t- c re
O O 4->
c o. in
C (/) •
O QJ i — to
c t_ re 4->
t- 3 to
•a o T3 re
C O •!— t-
re 01 4->
T3 QJ C
c: QJ t_ o
O 4-> O
•r— to T3
to •!— QJ OJ
U1 r— |S4 E
QJ -r- 3
f- tO r- r—
CD _^* ^ ^^
QJ +J E
t- cn t_ o
C O +J
C_ QJ Z
re i — t-
QJ QJ QJ
C > . 14-
••- re E a>
—1 3 C t-
* +~
^_
re
40
o
QJ
JT
.«
c
0
1^
j ^
re
^
o
re
o
QJ
^,J
E
O
t^^
•^
QJ
T3
re
O
to
.r—
T3 •
E
£_
QJ O
3 CO
VO
CM
0 -o
• c
O re
C~ M
re o
.c in
4-> LO
to •
to o
QJ LO
r~~ ^"
4— "
O in
0
to
re 4->
«- re
c oo
0
0 -0
c
QJ re
3 -
r— O
Q. i-H
QJ .
4-> 0
3 i— <
r—
O «
tO O
/^ i-H
re
QJ
^O L~
QJ QJ
s~ 3
3
to to
re QJ
QJ VI
E re
o
.c
4_> 14.
•r- O
S
c_
to QJ
QJ -D
in E
re 3
o c
«7>
54
-------
APPENDIX C
SENSITIVITY/UNCERTAINTY ANALYSIS OF PLUVUE
The degree of agreement that can be obtained between model predic-
tions and measurements is limited by the uncertainties associated with the
measurements. It is, therefore, of interest to evaluate the uncertainties
existing in the various measurements that constitute the data base used
for model evaluation and to determine their effects on visibility model
performance evaluation.
The procedure for conducting sensitivity/uncertainty analysis is as
follows. First, the uncertainties in the input data used in a model
simulation are defined. Ttien, the sensitivity of the model predictions to
each model input parameter is determined. The combination of the degree
of model sensitivity to an input parameter and the degree of uncertainty
associated with this parameter provides a measure of the effect of
uncertainties in an input parameter on the accuracy of model predic-
tions. In practice, the sensitivity analysis of the model is directly
combined with the uncertainty estimates, i.e., the sensitivity of the
model to parameter uncertainties is calculated by introducing perturba-
tions in the input parameter equal to the parameter uncertainties.
Flight 1 on the afternoon of 7 December was chosen for the
sensitivity/uncertainty analysis. This case study presents a well-defined
plume against a blue sky and a good telephotometer scan was obtained. A
qualitative study of the plume visibility model sensitivity and of the
uncertainties in the input data has been given elsewhere (Seigneur et al.,
1980). On the basis of the results of the uncertainty/sensitivity study
55
-------
conducted by Seigneur et al. (1980), seven input parameters were
selected. These parameters are listed in table C-l with their estimated
uncertainties.
TABLE C-l. INPUT PARAMETERS FOR PLUVUE SENSITIVITY/UNCERTAINTY ANALYSIS
Model Input Parameter
NOX emission rate
Primary aerosol emission rate
Wind speed
Wind direction
Visual range
Background ozone concentration
Aerosol size distribution
Coarse mode mean radius
Degree of
Uncertainty
13%
20%
40%
25%
5 ppb
0.25 urn (plume)
0.5 HTI (background)
The uncertainties in the NOX and primary aerosol emission rates were
chosen to be the standard deviations for the corresponding emission rates
for the winter VISTTA field program. The wind speed was calculated from
pibal data. Uncertainties were due to the temporal and spatial resolution
of these data. From the variation in wind speed between different levels
and measurement times, an uncertainty value of 40 percent was deter-
mined. The plume centerline location was defined from the S02 concentra-
tion profile measured by the aircraft. From the time response of the S02
monitor and aircraft speed, an uncertainty value on the order of 1 km was
estimated. This corresponds to a 7° uncertainty value in the effective
wind direction for the downwind distance considered.
56
-------
The visual range is calculated from background nephelometer measure-
ments. Uncertainties in the visual range value are due to measurement
uncertainties as well as to the nonhomogeneous nature of the atmosphere
(Babson et al., 1982). An uncertainty value of 25 percent was chosen for
the visual range. The background ozone concentration affects the rate of
formation of N02 in the plume. The uncertainty has been estimated to be
about 5 ppb (Richards et al., 1981). The aerosol size distribution has
some effect on the optical properties of the plume because light scatter-
ing depends on particle size. The uncertainty in the aerosol size
distribution was limited to the coarse mode because there was no appre-
ciable difference between the plumes and background accumulation mode
aerosol size distributions. The uncertainties in the mean radii of the
coarse mode were chosen to be 0.25 \tn for the plume aerosol* and 0.5 un
for the background aerosol (Whitby and Sverdrup, 1978).
Sensitivity studies were carried out by independently perturbing each
parameter by the uncertainty value listed in table C-l. For simplicity in
analyzing the results, the parameters were perturbed to decrease plume
visual effect. Thus, the wind speed, plume aerosol mean radius, and wind
direction were increased, whereas the NOX and primary aerosol emission
rates, background ozone concentration, background aerosol mean radius, and
visual range were decreased. A sensitivity study was also carried out by
varying all seven parameters. The results of these eight sensitivity
studies are shown in table C-2 for the ratios of plume/sky intensities at
the four wavelengths.
It is apparent from the results shown in the table that uncertainties
in the wind speed and visual range have the greatest effect on model
predictions. Uncertainties in the wind speed are higher in the blue range
(at 405 and 450 nm), where light absorption by NOo occurs, in part,
because N02 concentrations are more affected than primary aerosol concen-
trations by a change in wind speed. A change of 40 percent in the wind
G. E. Palomino, private communication, 1981,
57
-------
Oi
1 —
en
T— 1
oc
LU
£Q
51
LU
O
LU
D
r^
•N
Z
O
o
Z
a:
i i i
c
c
o
4_>
ID
•r—
CU
Q
CU
en
. 3
c
CU
U
c_
CU
Q-
oo
O
rt3
Q£
>,
4_J
.^
CO
C ^~»
CU E
•U C
»-H
^,^
s^-
CO
^•^
a>
E
3
a.
O
ro
^•O
O
«— (
•
O
CD
LO
LO
^s
<3-
^"
*
o
o
LO
^~
CM P^
• - •
C 0
P^ IO
• •
C 0
CM U3
P^ O
• •
c c
oo
0
•
^J-
01
^~
•
LO
ro
•
oo
G">
*3"
•
r^*
CU
3
03
>
T3
L.
3
4->
t_
cu
Q-
>^
fO
T3
«x^i
LO
O
CO
•
^^
oo
>^
03
•a
to
c
o
CTl
•
VO
r— 1
1
O
CU
CO
E
o
•a- o
• i — I
ro i— i
E
p~.
•
LO
O
CM
§.
a.
CT^
CM
O
•
O
T3
C
3
O
t_
en
N^
U
o
T3
CO
=s:
O
00
Q
1—
OO
>-
1—
Z
>— 1
^c
1—
ct
LU
O
SE
^^
^
CU
3
(O
^_
'O
c
CD
O
>^
ID
•o
CO
C
o
ro
P^_
CTi
>^
ITJ
T3
CO
C
o
ro
•
OO
1— 1
1
o
CU E
CO E CL
— j* a.
E CM ^
«i- o • ro
• CO ** O
CM C r^ •
r-l CM O
C
^
0
c.
CD
^^
O
ID
^3
«^-*
^
O
•
ro
, — _
cu
E
3
< —
CL
*^_^
^Z
LO
P-v
•
1 — 1
cu
>
o
J3
03
CO
^
t_
IB
E a>
t_ 03
Q. (_
"O
0)
cu
Q.
CO
•o
C
•n-
2
C
o
4^
O
CU
•r-
•o
•o
C
•r-
2
CU
CD
C
IB
L.
,-^.
IB
3
co
• r«.
=>
CU
c
o
IM
0
T3
C
3
O
t
CD
^
U
CO
c
0
• r—
.|-3
IB
L.
4_}
C
CU
o
c
0
U
c
o
•r—
4J
3
_o
•^
t_
4->
CO
•^
T3
CU -r-
IM -f-
••- -a
in to
L.
r—
o c
to 03
O (11
L_ y
cu
cu
E
t_
(B
Q-
^~
r^
*£
58
-------
speed leads to an equivalent change in primary aerosol concentrations, but
to a 50 percent change in N02 concentrations because of the nonlinear
relationship of NC^ concentrations to NOX emissions.
Uncertainties in the visual range also lead to larger deviations in
plume/sky intensity ratios at shorter wavelengths. This result occurs
because the plume visual effects at Navajo Generating Station are more
important in the blue range and the model is, therefore, generally more
sensitive in this range to perturbations that affect the whole visible
range.
Changes in the NOX emission rates have greatest effect in the blue
range, which corresponds to the range of light absorption by NC^.
Negligible effect appears at 630 nm. Although emission rates of NOX are
known with reasonable accurcy, about 13 percent, the effect on model
predictions is not negligible.
j
Uncertainties in primary aerosol emission rates are greater.
However, aerosols do not contribute as much as N0£ to the overall visual
effects of the plume at the downwind distance considered. The effect of
these uncertainties is at most 1.17 percent in the red and decreases with
decreasing wavelength as the effect of N02 absorption becomes predominant.
Uncertainties in the mean radii of the coarse mode aerosols have a
comparable effect on the model predictions. The value of the background
ozone concentrations affect the formation of NC>2 in the plume. An
uncertainty value of 5 ppb in the ozone concentrations corresponds to
about 15 percent uncertainty. Thus, the model is about twice less
sensitive to Oj concentration than to NOX emission rates.
The uncertainty in wind direction correspond to an uncertainty in the
plume-observer distance of 6 percent. This does not induce any major
perturbation in the model predictions.
59
-------
When all these parameters wera perturbed, large deviations occurred
in the values of the predicted plume/sky intensity ratios. It should be
noted, however, that this represents an upper limit of the total uncer-
tainty in model predictions due to uncertainties in input data, because
input data uncertainties may well compensate each other to some extent,
thus leading to a lower deviation in model predictions. The perturbations
in the plume/sky intensity ratios correspond with perturbations in plume
contrast of 50, 50.5, 54.4, and 55.5 percent at 405, 450, 550, and 630 nm,
respectively.
Comparison of telephotometer scans conducted at 5-minute time
intervals showed that uncertainty in the telephotometer measurements of
plume visual effects ranges from C to 5 percent. Since the relative
deviations between the predicted and measured concentrations [(Predicted -
Measured)/Predicted] are 4.5, 1.2, -2, and -4 percent at 405, 450, 550,
and 630 nm, respectively, the agreement between model predictions and
measurements is within the uncertainties in the input data and measure-
ments for a run for which the plume is well defined.
60
-------
APPENDIX D
MODIFICATIONS TO THE ERT VISIBILITY MODEL COMPUTER CODE
To evaluate the modules of the ERT Visibility Model, it was necessary
to modify the computer code to obtain those output functions not provided
by the original model output. Additional output functions needed for
evaluation of the chemistry and dispersion modules included
> Centerline concentrations of NOX, f^, and
particulates.
> Values integrated across the plume (integral concentra-
tions) for N02 concentrations.
> Values integrated across the plume (integral values) for
the scattering coefficient of particulates (bsp).
D.I CENTERLINE CONCENTRATIONS
Integrated into the ERT Visibility Model was a method of determining
pollutant concentrations at various distances, y, from the plume center-
line (e.g., for equally spaced points along a given line of sight). By
entering a particular downwind distance, DWDT, and setting y = 0, center-
line concentrations for noncarbonaceous particulates, carbonaceous
particulates, and N0£ could be obtained. In the cases considered,
center-line concentrations of NOX were also of interest. NOX concentra-
tions were already determined by the model as an initial step to calculat-
ing N0; therefore, only a unit conversion factor and a common statement
61
-------
remained to be added. To obtain the center!ine concentrations of particu-
lates, N02, NOX, and 03, the following steps were added to SUBROUTINE
PLUME:
(1) The downwind distance, DWDT, was read in.
(2) SUBROUTINE SIGMA was called using DWDT to obtain SY and
SZ, the horizontal and vertical dispersion parameters.
(3) SUBROUTINE VERT was called using SZ to obtain the
vertical distribution function, VERT. Because the plume
height at DWDT was unknown, the final plume height
calculated was used.
(4) SUBROUTINE YDIF was then called, with the distance from
the plume center-line set equal to zero. From this, the
horizontal distribution function, YDIF, was obtained.
(5) CHIQ was then set equal to ZDIF*YDIF/UHP, as in the
model, with UHP equal to the wind speed at stack height.
(6) CHIQ, DWDT, and the plume height above sea level were
then used in a call to SUBROUTINE QF to obtain center-
line concentrations for the particulates N02 and NOX>
D.2 INTEGRAL PLUME VALUES
To obtain the integral values of b and N02 concentrations across
the plume along the line of sight, indices corresponding to the plume
edges had to be determined. This was done in SUBROUTINE PLUME while the
concentration field was being recorded along the line of sight from the
observer to the visual range. The method used was to compare YABS, the
absolute value of the distance from the plume centerline, with 2*SY at the
corresponding downwind distance. When YABS was first found to be less
62
-------
than, or equal to, 2*SY, the iteration number was set equal to NNBEG. As
the concentration recording continued along the line of sight, a check for
YABS > 2*SY was imposed. When YABS again became greater than, or equal
to, 2*SY, the iteration number was set equal to NNEND. NNBEG and NNENO
were then used as signals for the plume boundaries in SUBROUTINE RADTRA.
SUBROUTINE SIMPS was called at 550 nm using NNBEG and NNEND as
integration/summing boundaries for noncarbonaceous particulates, carbona-
ceous particulates, and N02« The summed concentrations obtained were then
multiplied by corresponding scattering and/or absorption coefficients.
(/ N02 dy values were later divided by the absorption coefficients to
obtain / N02 dy in units of km .)
Centerline N02 and NOX concentrations and / N02 dy values were
converted from units of g/m^ to ppm using 2.46 x 10*° ppm/(mole cm~3),
along with the respective molecular weights. The resultant values were
then divided by 0.86 atmospheres to compensate for pressure.
63
-------
REFERENCES
Babson, B. L., R. W. Bergstrom, M. A. Samuelson, C. Seigneur, A. Waggoner,
and W. C. Malm (1982), "Statistical Analysis of Nephelometer Regional
Field Data," Atmos. Environ., in press.
Bergstrom, R. W., B. L. Babson, and T. Ackerman (1981), "Calculation of
Multiply Scattered Radiation in Clean Atmospheres," Atmos. Environ.,
Vol. 15, pp. 1821-1826.
Bergstrom, R. W., C. Seigneur, B. L. Babson, H. Y. Holman, and M. A.
Wojcik (1981), "Comparison of the Observed and Predicted Visual
Effects Caused by Power Plant Plumes," Atmos. Environ., Vol. 15, pp.
2135-2150.
Environmental Research and Technology, Inc. (1980), "ERT Visibility
Model: Version 3--Technical Description and User's Guide," Document
M-2020-001, Environmental Research and Technology, Inc., Concord
Massachusetts.
Fox, D. G. (1981), "Judging Air Quality Model Performance," Bull. Am.
Meteor. Soc.. Vol. 62, pp. 599-609.
Friedlander, S. K. (1977), Smoke, Dust and Haze, Fundamentals of Aerosol
Behavior, p. 143 (John Wiley & Sons, New York, New York).
Hudischewskyj, A. B., B. L. Babson, R. W. Bergstrom, and C. Seigneur
(1981), "Coal-Fired Power Plant Contribution to Visibility Impairment
in Western Pristine Areas--Index to the Photographs Taken from
Telephotometer Sites during the June-July and December 1979 VISTTA
Field Programs," 6-ES81-025, Systems Applications, Inc., San Rafael,
California.
Johnson, C. D., D. A. Latimer, H. Hogo, and R. W. Bergstrom (1980),
"User's Manual for the Plume Visibility Model (PL.UVUE)," EPA-450/4-80-
032, U.S. Environmental Protection Agency, Research Triangle Park,
North Carolina.
64
-------
Latimer, D. A., R. W. Bergstrom, C. D. Johnson, and J. P. Killus (1980),
"Modeling Visibility," Second Joint Conference on Applications of Air
Pollution Meteorology, 27-29 March 1980, New Orleans, Louisiana.
Latimer, D. A., R. W. Bergstrom, S. R. Hayes, M. K. Liu, J. H. Seinfeld,
G. Z. Whitten, M. A. Wojcik, and M. J. Hillyer (1978), "The Develop-
ment of Mathematical Models for the Prediction of Anthropogenic
Visibility Impairment," EPA-450/3-78-110a, b, and c, U.S. Environmen-
tal Protection Agency, Research Triangle Park, North Carolina.
Richards, L. W., J. A. Anderson, D. L. Blumenthal, A. A. Brandt, J. A.
McDonald, and N. Waters (1981), "The Chemistry, Aerosol Physics, and
Optical Properties of a Western Coal-Fired Power Plant Plume," Atmos.
Environ., Vol. 15, pp. 2111-2134.
Seigneur, C. (1982), "A Model of Sulfate Aerosol Dynamics in Atmospheric
Plumes," Atmos. Environ., in press.
Seigneur, C., A. B. Hudischewskyj, B. L. Babson, R. W. Bergstrom, and
S. Eigsti (1981), "Coal-Fired Power Plant Contribution to Visibility
Impairment in Western Pristine Areas-VISTTA 1979 Interim Report-Part
2-Case Studies for Plume Visibility Model Validation," MRI 81 IR 1829,
Meteorology Research, Inc., Santa Rosa, California.
Seigneur, C., R. W. Bergstrom, D. L. Blumenthal, L. W. Richards, E. S.
Macias, W. E. Wilson, and P. S. Bhardwaja (1980), "The Data Base for
Visibility Model Evaluation from the June-July and December 1979 EPA-
VISTTA Field Programs," Symposium on Plumes and Visibility--Measure-
ments and Model Components, 10-14 November 1980, Grand Canyon,
Arizona.
Sheppard, J. C., M. Campbell, P. H. McMurry, and G. L. Kolk (1980), "Trace
Gas and Particle Data from a Remote Site in Northern Arizona,"
unpublished results, Department of Civil Engineerinig, University of
Washington, Seattle, Washington.
Stewart, D. A., and M. K. Liu (1981), "Development and Application of a
Reactive Plume Model. Atmos. Environ., Vol. 15, pp. 2377-2393.
Whitby, K. T., and G. M. Sverdrup (1978), "California Aerosols: Their
Physical and Chemical Characteristics," ACHEX Hutchinson Memorial
Volume, Particle Technology Laboratory Publication Number 347,
University of Minnesota, Minneapolis, Minnesota.
65
-------
TECHNICAL REPORT DATA
(Please read Instruction* on the reverse before completing)
1 REPORT NO.
EPA-450/4-82 -008
3. RECIPIENT'S ACCESSI Of* NO.
4. TITLE ANDSUBTITLE
Evaluation of the EPA PLUVUE Model and the ERT
Visibility Model Based on the 1979 VISTTA Data Base
5. REPORT DATE
June 1982
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
Christian Seigneur, A. Belle Hudischewskyj, and
Robert W. Bergstrom
8. PERFORMING ORGANIZATION REPORT NO,
82190
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, California 94903
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3225
68-02-3582
12. SPONSORING AGENCY NAME AND ADDRESS
U. S. Environmental Protection Agency
OAQPS-MDAD-SRAB
(MD-14), Research Triangle Park, N. C.
13. TYPE OF REPORT AND PERIOD COVERED
Final
27711
14. SPONSORING AGENCY CODE
EPA-450/4-82 -008
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report presents the results of the evaluation of the EPA PLUVUE model and the
ERT Visibility model based on the 1979 EPA VISTTA data base which was obtained during
the June-July and December 1979 field orograms conducted at the Navajo Generating
Station. The overall evaluations were performed for three different types of back-
ground conditions: clear sky, dark mountain and hazy sky.
Results of the evaluation of the two models indicate: (a) PLUVUE tends to overestimate
Dlume visual effects; the average absolute relative errors in plume contrast are 69,
83, 69 and 154 percent at 405, 450, 550 and 630 nm, respectively; (b) the ERT Visibilit.
model tends to underpredict the plume visual effects; the average absolute relative
errors in plume contrast are 42, 37, 30 and 262, percent at the above wavelengths;
(c) the dispersion module in each model appears to contain the primary source of
uncertainty in model predictions; (d) the chemistry module in each model shows reason-
able agreement between predictions and measurements of the NCL/NO ratio; (e) the
optics module in each model compares satisfactorily with the casexstudies.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b, IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Atmospheric Chemistry
Mathematical Modeling
Meteorology
Nitrogen Oxides
Power Plants
Radiative Transfer
Visibility
New Source Review
Class I Areas Visibility
Impairment
13B
13. DISTRIBUTION STATEMENT
Release to the Public
13. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
70
20 SECURITY CLASS (This page/
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
------- |