Environmental Monitoring Series
REFINEMENT AND VALIDATION OF AN
URBAN METEOROLQGICAL^POLLUTANT MODEL
Environmental Sciences Research Laboratory
Ipce of Research and Beveloproent
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service. Springfield. Virginia 22161.
-------
EPA-600/4-76-037
July 1976
REFINEMENT AND VALIDATION OF AN URBAN
METEOROLOGICAL-POLLUTANT MODEL
by
Joseph P. Pandolfo
Clifford A. Jacobs
Robert J. Ball
Marshall A. Atwater
Joseph A. Sekorski
THE CENTER FOR THE ENVIRONMENT AND MAN, INC.
Hartford, Connecticut 06120
68-02-1767
Project Officer
Jason Ching
Meteorology and Assessment Division
Environmental Sciences Research Laboratory
Research Triangle Park, N.C. 27711
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, N.C. 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research Lab-
oratory, U.S. Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect the views and
policies of the U.S. Environmental Protection Agency, nor does mention of
trade names or commercial products constitute 'endorsement or recommendation
for use.
ii
-------
FORWARD
Understanding of the phenomena and processes forming the reciprocal rela-
tionships between man and his environment has become essential. The Center
for the Environment and Man, Inc. investigates these relationships in programs
which are concerned with
• quantitative analysis and simulation of the physical characteristics
of atmospheric and oceanic systems on all scales of action;
• the development and application of tools for the optimum use of en-
vironmental resources for mankind.
In this report, both major activities of the Center are reflected. It is
concerned with the refinement and validation of a unique computer model, which
simulates the joint development of natural meteorological-oceanic systems and
anthropogenic pollutant fields on the spatial scales of a major metropolitan
region. The model can serve both as a research tool for the quantitative study
of such systems, and as a component of future air quality management procedures.
F. C. Henriques
Chief Executive Officer
The Center for the Environment and Man, Inc.
iii
-------
iv
-------
ABSTRACT
New urban meteorological-pollutant model forecasts for the case described
in Pandolfo and Jacobs (1973) were obtained and validated by the evaluation
method proposed by Nappo (1974). Three new forecasts are described. The first
retains the coarse {8 mile) horizontal resolution of the source emissions in-
ventory previously used, but incorporates improved meteorological and pollutant
.input data. The second is like the first, but defines the source emissions in-
ventory with finer (2 mile) horizontal resolution. The third is like the se-
cond, but incorporates increased vertical resolution below the prevalent inver-
sion base level.
It is noted that the model should be credited with the production of a
meteorological forecast as well as a pollutant forecast, in contrast to the
others surveyed by Nappo. Most of these other models must be provided.with a
meteorological forecast as input. The evaluation statistics for the pollutant
forecast by this model indicate that its meteorological forecast can be ex-
pected to be well suited as input to any pollutant forecast model.
Contrary to Nappo's suggestion, and our expectation, the evaluation statis-
tics show that increasing the degree of horizontal detail of the source emis-
sion inventory did not, in this case, significantly increase the sensitivity
and. accuracy of the pollutant concentration forecast. In conformance with
Nappo's previous finding, however, was the result that the detail of the verti-
cal diffusion calculations was not very significant in determining the accuracy
of the pollutant forecast. The validation statistics also indicate that making
fuller use of the original capabilities of the model did significantly increase
the accuracy of the forecast.
The results of the first ("new coarse") forecast prepared for this report
show that the model is the least expensive of those surveyed by Nappo to run
(except for the "persistence" model) when not charged with the cost of the
meteorological prediction. This pollutant forecast is among the most accurate
of those surveyed in terms of Nappo's overall average concentration and temporal
correlation statistics. It is approximately accurate as other primitive equa-
tion models surveyed in terms of Nappo's standard error and spatial correlation
statistics.
Because of these results, it is recommended that:
« the model described here be applied to generate meteorological
predictions for input to more accurate pollutant forecast models
where greater accuracy in spatial detail at smaller than metro-
"^s-v, politan scale is desired.
-------
• the model described here be applied as a pollutant forecast
model where less accuracy i-n spatial detail is acceptable in
return for low implementation costs; and
• a more accurate (higher order) computational scheme for the
horizontal advection calculations be incorporated in the model.
The development effort required to do this is relatively small,
and the accuracy of spatial detail on the intra-metropolitan
space scales could well be significantly enhanced.
This report was submitted in fulfillment of Contract No. 68-02-1767 by
The Center for the Environment and Man, Inc. under the sponsorship of the U.S.
Environmental Protection Agency. This report covers a period from February
1974 to November 1976, and the work was completed as of March 1976.
VI
-------
CONTENTS
Foreward iii
Abstract v
Tables vii
Figures viii
1. Introduction 1
2. The Planetary Boundary Layer Model 4
3. The Input Data for This Experiment 5
3.1 The modeled region 5
3.2 Initial Fields 7
3.3 Modifications of the thermal radiation by
atmospheric pollutants. ; 8
4. The Results of the Experiments 9
4.1 Time series presentations 9
4.2 Verification in terms of Nappo's statistics 12
5. Summary and Conclusions 19
References 21
Technical Report Data Form 22
.TABLES
Number Page
1 Model Evaluation Based on Temporal Characteristics 13
2 Summary of Statistics for "New Forecasts" 18
vii
-------
FIGURES
Number Page
1 Basic equations used by the model 4
2 Locations of the horizontal grid used by Roth, et al. (1971)
and the smaller area grid (fine and coarse) used in this
experiment 5
3 Heights (depths) of vertical grid levels in meters ........ 6
4 Initial vertical temperature profiles used in forecasts 7
5-14 Time series of observed and forecast pollutant concentrations
at measurement sites.
5 Reseda . . . 9
6 Burbank ' 9
7 Azusa 10
8 West Los Angeles . 10
9 Downtown Los Angeles 10
10 Commerce 10
11 El Monte 11
12 Lennox 11
13 Whittier 11
14 Long Beach . . . 11
15 Spatial distribution of temporal correlation coefficients R(t)
for the "fine (horizontal) grid" and "new coarse (horizontal)
grid" forecasts 14
16 R(t)s versus R(s)*- average result for each model tested ... 15
17 Temporal variation of spatial correlation coefficients R(s) , . 16
18 o(t)s versus a(s) average result for each model tested ... 17
19 Average and spread of temporal and spatial ratios 17
viii
-------
SECTION- 1
INTRODUCTION
Nappo's (1974) survey indicated that a model previously developed at The
Center for the Environment and Man, Inc. (CEM) under the Environmental Protec-
tion Agency (EPA) sponsorship (Contracts CPA 70-62 and 68-02-0223) and de-
scribed in a report to EPA (Pandolfo, et al., 1973, EPA-R4-73-024a,b), pro-
duced pollutant forecasts with some verification statistics as accurate as
those produced by more complex models, and at costs commensurate with simpler
models. Nappo's analysis suggested again that a major source of inaccuracy
(as measured by other verification statistics) was one proposed in that re-
port—viz., the coarseness of the horizontal resolution of the model.
The abstract of that report follows:
The urban boundary-layer model, described in a previous report
(Pandolfo3 et al.3 1972), was modified and used in forty test runs.
Many of the runs varied the meteorological input about a standard
(observed), set. It has, therefore, been demonstrated that an eco-
nomical, objective, physically consistent, and precisely specified
(though with some arbitrary elements) procedure has been achieved
for obtaining and predicting the three-dimensional meteorological
fields needed. In several of the runs, the input topography, land-
water distribution, and other physical characteristics of the under-
lying surface were varied. The results demonstrate that ready gen-
eralization to other regions can be expected.
The modeled region was simulated with relatively coarse (8-
mile) grid spacing. This is in contrast to other models which deal
with pollutants only, and which are based on two-mile grid spacing
(Sklarew, et al., 1971; Eoth, et al., 1971). Nonetheless, the tem-
poral and spatial variations of air temperature, humidity, and wind
are simulated with an encouraging degree of realism. Temporal and
spatial variations of CO are also simulated fairly realistically,
with somewhat less accuracy than in the model described by Roth,
et al. (1971), and with accuracy equivalent to that shown by Sklarew,
et al. (1971). It is reasonable to expect improved simulation ac-
curacy with the finer horizontal resolution used in these other
models, and the performance of such simulations with this model is
strongly urged.
r
Nappo's (1974) results formed the basis of a proposed scope of work for
this project, which is defined as follows. This project was limited, as was
the previous project, to the consideration of carbon monoxide (CO) as the
-------
subject pollutant. Therefore, no attempt was made to model chemical changes.
It was further proposed that:
• The model described in the EPA report (EPA R4-73-025a,b) be
modified to allow 2-mile square resolution in the CO predic-
tion and CO input source distribution, with no other changes.
• A single test prediction be run. for the case previously
treated (Los Angeles, 14GMT, 29 September 1969 through 04GMT,
30 September 1969).
• The accuracy of this pollution prediction be evaluated and
compared to that of other pollution models using the method
of Nappo (1974).
• The computer costs for the prediction be fully documented.
• A summary report, supplementing our previous report, be sub-
mitted on the results of this single experiment.
As proposed, this scope of work would have resulted in an experiment in
which two pollutant forecasts would have been compared—those resulting from
a "coarse (horizontal) grid" model, and those resulting from a "fine (hori-
zontal) grid" model. During the course of the work, it was decided, with the
agreement of the contract monitor, to expand the scope of the work, at no ad-
ditional cost to the sponsor, to take advantage of several capabilities of
the coarse (horizontal) grid model which had not been previously used, and to
obtain forecasts with
1) greater vertical resolution below the prevalent temperature in-
version base level (at about 150m in height) and coarser verti-
cal resolution above this level;
2) initial vertical profiles of temperature in the soil and lower
atmospheric layer over the modeled region whose local shapes
were generated by the 24-hour forecasts previously obtained
with the model; but whose absolute values were determined by
the initial sparse set of temperature observations available
from the five weather stations in .the region;
3) calculated absorption/scattering of solar and radiation by
N02 and particulates; N02 and particulate concentrations used
in these calculations were assumed to be fixed fractions of
. the predicted CO concentration at any given place and time.
. The scope was additionally expanded through the use of hourly averages
of model-predicted pollutants in the validation, rather than the single-time
step values previously used (the observations are hourly averages). Also,
changes in several of the initial pollutant concentrations were made to pro-
vide better agreement with the observed values at the initial time and at
nearby stations.
-------
This expansion of scope now allowed three forecasts to be compared; viz.
• the "old coarse (horizontal) grid" forecast previously validated
in Pandolfo, et al. (1973), and Nappo (1974);
• a "new coarse (horizontal) grid" forecast; and
• a "fine (horizontal) grid" forecast.
Improvements in accuracy of the second forecast over the first can thus .
be attributed to fuller use of the meteorological capabilities of the original
model and in the changes made in the initial concentrations. Improvements on
accuracy of the third forecast over the second can be attributed to refinement
of the horizontal resolution of the source emission fields and pollutant pre-
dictions .
Finally, we refined the vertical grid still further to approximate the
vertical resolution used by Roth, et al. (1971), and obtained another set of
new forecasts. These "fine vertical grid new forecasts" did not exhibit bet-
ter verifications than the other "new" forecasts as will be shown later.
-------
SECTION- 2
THE PLANETARY BOUNDARY LAYER MODEL
The model equations remain as outlined in Section 2.0 and Appendix C of
Pandolfo and Jacobs (1973).
The only changes to the model are those which allow the optional use of
a finer horizontal grid for the solution of the conservation equation for one
of the pollutants, than that used in the solution of the conservation equations
for all other variables. The program, as described in an addendum to this re-
port, now allows the finer grid to have a grid length which is an integral
fraction [l]"1 of the basic meteorological grid length, where I can range
from 1 to 4.
A summary of the modeling equations is also included here for handy ref-
erence (see Figure 1).
FORM
r *c,i
KJ
* Aj
ATMOSPHERE
li-I)!r]* '[*-*]
rad
•S-*-
SOIL
If -*[«•«£•]
INTERFACE ENERGY BALANCE
O =Rs(i-a) » Ra - OT£ - LE - A -S «l
Figure 1. Basic equations used by the model.
-------
SECTION- 3
THE INPUT DATA FOR THIS EXPERIMENT
3.1 THE MODELED REGION
The specification of the modeled region remains as defined in Section
3.1.1 of Pandolfo, et al. (1973), with the following exceptions.
The fine horizontal grid cell dimension of 2 miles is specified for the
pollutant, along with the coarser 8-mile grid dimension used for the other
variables. The two grid arrays, and their location in the Los Angeles region
are shown in Figure 2.
25
24
23
22
21
20
19
18
17
1C
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
*»
-
ss
•••^
—
—
'Eft,
•--1
-2
-3
urn
j^Ci"
SA
"-"
X;\
'•1C
,' V'fi
•— — -
:TA,
HOU:
S
'...'.'
'"v.
• I "*
San fc
Uti-
1
;os
17. !
K
anta
nnic
^
—
—
O/1
—
CA
IS
.— _—
a
i"
\
—
*'.:•
f
v
:?:
-^ • -^
c
(_i
2
1
Jtrf
»rt
i'"
Ho
u
ilh
lyi;
—
y
30d
_..
Los Am;el;s
• Internationr.l-
Ai rport
^
^
Zife
^
.....
S-
V
;a
r
j-
'V
^
5^
Pasadena*
• Oo^mt
—105 A
—
-------
The vertical layer modeled extends, as previously, from 200m below the
air-soil(water) interface to 1500m above this interface. However, the grid
levels used in the new forecasts have been changed as described in Section 1
of this report. The old set and the new set of vertical grid levels are shown
in Figure 3. The "finer vertical" set of vertical grid levels is also shown.
Original
Model
1500.00
onn nr\ i
650.00
450.00
">f\f\ nn A
150.00
75.00
i
25.00 ^
t i
4.00 *
0.00
- 0.05
- 0.10
- 0.15
- 0.20
- 1.00
- 10,00
- 50.00
-100.00
-200.00
i
Vertical Grid for
"New Coarse- (horizontal)
Grid" and
"Fine-(horizontal) Grid"
Forecasts
On T
OUT /
U T / -«_
/125^_
Ono
X^
Vertical Grid For
"Finer Vertical"
Forecast
135
115 »
i 85 /65
/55
,, / flC
S. 15
\35
15
Figure 3. Heights (depths) of vertical grid levels in meters,
6
-------
3.2 INITIAL FIELDS
The initial fields of the variables specified in Section 3.1.3 of Pan-
dolfo, et al. (1973) were used with the temperature and pollutant fields modi-
fied as described in Section 1 above. A comparison of the originally used and
modified initial vertical profiles of temperature at the center of the modeled
regions is shown in Figure 4.
Height
(meters)
1400 -
1200 -
1000
200|-
0
0
-1.0
Depth
(meters) _2>Q
-3.0
ATMOSPHERIC TEMPERATURE
PROFILES
Original Initial
New Initial Shape
(predicted)
16 18 20 22 24 26 28 30 32 Temperature (°C)
__!__ J | | I [ t I 1
SOIL TEMPERATURE
PROFILES
Original Initial
r New Initial Shape
(predicted)
Figure 4. Initial vertical temperature profiles used in forecasts,
-------
3.3 MODIFICATION OF THE THERMAL RADIATION BY ATMOSPHERIC POLLUTANTS
The original model had the potential for including in the meteorological
forecast the radiative effects of selected pollutants. In its most complete
form, the model would include each of two radiatively active pollutants as
separate dependent variables whose concentrations are to be predicted. Atwa-
ter's (1971) work, based on a simplified version of this model, had, however,
identified the nitrogen oxide (NO2) and the particulate aerosols as the only
constituents with significant radiative effects. In. the present experiment,
the inclusion of these radiative pollutants was, therefore, carried out in a
much simpler manner than that originally developed. We first assumed that the
concentrations of the radiatively active pollutants could be represented as
fixed fractions of the CO, whose concentration was explicitly predicted. This
assumption was justified by analysis of the observed concentrations which were
available for study,* and which established concentration ratios.
These concentration ratios were then combined with actual values of the
solar absorption coefficient for NO2, the solar absorption, solar scattering,
and infrared absorption coefficients for aerosol particulates to obtain a cor-
responding set of "fictitious" radiative coefficients for the only pollutant
that was explicitly predicted—viz., the CO. These "fictitious" coefficients,
valid for 1 ppm NO2 and urban particulate aerosols, are:
• solar absorption coefficient = .15 km
• solar scattering coefficient = .15 km"
« infrared absorption coefficient = 0 km
• infrared scattering coefficient = 0 km
These coefficients are used with the simulated CO concentrations and the con-
centration ratios to compute the actual coefficient at each time step.
* Obtained from the Los Angeles Air Pollution Control District.
-------
SECTION 4
THE RESULTS OF THE EXPERIMENTS
4.1 TIME SERIES PRESENTATIONS
Time series of predicted CO concentrations at the 4-m grid level, at each
of the measuring stations are shown in Figures 5 through 14. Also shown are
the measured values, and the values previously reported and validated from the
"old coarse (horizontal) grid" forecasts (Pandolfo, et al., 1973; Nappo, 1974).
Two "new" forecasts are shown—one for the "fine (horizontal) grid" and one
for the "new coarse (horizontal) grid."
Figures 5 through 14. Time series of observed and forecast
pollutant concentrations at measurement sites.
18
Figure 5. Reseda (1,1)
Figure 6. Burbank (1,2)
§
12
10
£ 8
8
I
6
• • Pandolfo S Jacobs,
O-—O Fine Grid
D--O Coarse Grid
«•-•••* Observations
f—*-
i i i i
t i i i
18
16
12
,0
•J
o
8 10 12
Time (hrs)
16
8 10 12
Time (hrs)
lit
16
-------
18
16
14
12
10
Figure 7. Azusa (1,5)
Figure 8. West L.A. (2,1)
£ 8
•—• Pandolfo S Jacobs. 1973
O---O Fine Grid
D--O Coarse Grid
*—~* Observations
18
10 12
Time (hrs)
, , Pandolfo 6 Jacobs,
fj 1975
/ !: O---O Fine Grid
| ': o--o Coarse Grid
• • *••••••* Observations
/ =.
: { B Original Initial
j ': Concentration
/ : A New Initial
• 5 Concentration
1 !
J i
• 1
• K 1
• * \ \
_^\ : . p
"^\ *"~"\ / "
X, ; /
r R^-CL '• /
• _^Sv^:
••
1 1 1 1 1 1 1 1 • 1 1 1
16
14
'2 ,-.
I
Q.
"^ '
10 J
A4
(0
L.
*>
C
8 S
C
o
u
6
L
H
2
A
6 '8 10 12 14 16
Time (hrs)
18
16
14
12
10
S 8
Figure 9. Downtown L.A. (2,3)
Figure 10. Commerce (2,3)
I
•—• Pandolfo 6 Jacobs, 1973
O-"o Fine Grid
D--O Coarse Grid
Observations
Original Initial
Concentration
New Initial
Concentration
10 12
Time (hrs)
14
16
to 12
Time (hrs)
14
16
18
16
14
12
'I
to S
4J
(0
k.
*>
8 S
10
-------
18
Figure 11. El Monte (2,4)
Figure 12. Lennox (3,2)
12
10
S 8
• • Pandolfo I Jacobs, 1973
O---O Fine Grid
O--O Coarse Grid
•»•--•* Observations
8 10 12
Time (hrs)
18
16
12
10
8 £
8 10 12
Time (hrs)
16
18
. 16
Figure 13. Whittier (3,4)
12
a
o.
10
£ 8
V
Figure 14. Long Beach (4,3)
•—• Pandolfo S Jacobs, 1973
O—O Fine Grid
O--O Coarse Grid
Observations
8 10 .12
Time (hrs)
14
16
t I L
18
16
14
12 _
I
o.
10
10 1,2
Time (hrs)
14 16
11
-------
While it is difficult to summarize the comparative results from this pre
sentation, some qualitative features are evident. First is that the "new"
forecasts are slightly more realistic than the "old" forecasts. Second is
that only slight differences between the "new coarse (horizontal) grid" fore-
cast and the "fine (horizontal) grid" forecast are evident.
One change in the validation procedure should be noted. In the "old"
forecasts, "instantaneous" (4.8 minute average) predicted values on the hour
were compared with hourly average observed values centered on the half hour.
In the "new" forecasts, hourly average predicted values centered on the half
hour are used in the verification. Some of the improvement noted in the sta-
tistical verifications of the "new" forecasts that follow may be attributed
to this change.
4.2 VERIFICATION IN TERMS OF NAPPO 'S STATISTICS
The statistics suggested by Nappo in his 1974 paper for the evaluation
of pollutant predictions are
« the spatial average of temporal correlation coefficient, R(t)s ;
« the temporal average of the spatial correlation coefficient,
R(s)t ; -
• a "spread", or temporal average of standard deviation of the
radios of predicted-to-observed map values at a given time,
« a "spread" or spatial average of standard deviation o'f the
ratios of predicted to observed time series values at a given
place, o(t)s ,
« the spatial average of the ratios of predicted- to-observed
values as seen in time series at a given place, r(t)s ;
• the temporal average of the ratios of predicted- to-observed
values as seen on maps at given times, r(s)fc .
Nappo showed that, when a validation sample was distributed in both space
and time, the last two statistics were identical; i.e.,
rTt)S = rls")1" .
The graphs presented in this section are those of Nappo, with points add-
ed for the two new forecasts obtained in this project, which are identified
as "Pandolfo, et al. (1976)".
We first reproduce Nappo 's summary table of temporal characteristics
(Table 1) with the newly validated forecast summaries added. The station-to-
station variation of R(t) is shown in Figure 15. In Table 1, we show the
spatial average R(t)s .
12
-------
TABLE 1
MODEL EVALUATION BASED ON TEMPORAL CHARACTERISTICS
MODEL
Average
Temporal
Correlation
Coefficient
IBM 360/65
Computer Time
for 24-hour
Prediction
(minutes)
Computer Cost
for 24-hour
Prediction
(dollars)
MacCracken, et al. (1971)
multi-box
24-Hour Persistence
Roth, et al. (1971)
primitive equation
Hanna (1973)
ATDL simple model
Sklarew, et al. (1972).
particle-in-cell .
Pandolfo and Jacobs (1973)
primitive equation ("Old
coarse (horizontal) grid"
forecast)
Reynolds, et al. (1973)
primitive equation
Eschenroeder, et al. (1972)
trajectory
Lamb and Neiburger (1971)
trajectory
Pandolfo, .et al. (1976)
primitive equation
Coarse (horizontal) Grid
Fine (horizontal) Grid
Finer Vertical, Fine
(horizontal) Grid
0.37
0.47
0.52
0.60
0.65
0.66
0.73
0.73
0.90
0.75
0.79
0.80
106
none
60
none
49
20
30
15
35
18*
144*
150*
350
none
200
none
160
70
100
50
115
" 60 t
480 t
500t
* Converted from CDC 7600 running time using the ratio 25 rain (IBM 360/65)
= 1 roin (CDC 7600):
t Estimated at Nappo's average of $3.33/min (IBM 360/65). Actual costs
experienced in this project on CDC 7600 @ $2580/hour were slightly lower
than these.
13
-------
25
24
23
22
21
20
19
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
i
SAH
-=a=f
--.—•
— -
----
•*»
92
-I
-2
'
i:
j
AND
IK
*
SA
"•— "
^
'M(
J v/\
D
.85
s'TA
MOUM
V
. U V
S
V
-*'•£<
I*~%~*"'T-
'•*/,/•*'
I
KOH
TAI:
/fs
*
^••-
t
CA
IS
T
.50
L
55
Y
\
J
'1,
N;
i
.1
.5,
^
—
r KI ^
_ n t
*
';';'
,.
N.
2
UR
i*
--•*
—
1
h.
.88
s^
"^'N
i
-._.
)
:=.
^s
^
-. '•/"""**
i"
^v
—
DAJ
*
I
..«
5f
?
.73
.55
ON
-^r
¥
K
—
l/L
*
B
.52
-ii
—
ft
.73
"'** v^«
,
s ; . ; ,
'"tl
SAN (
4
— v
.81
\
Vt
66
—
^V'-T
x.
'AUK
~--
ELM
*
fHT
*
s
.85
^^_
R..
.75
^
-,,.
—
.82
--
PU
\^
—
k:.
X,
"Oy/,
H'AI
5
'•tort.
— »s
\
.82
_«.
^-_
—
^s
vs
=^
•iS-
J±%
—
%
^.
*«
-—
J — •**
1 2 3 4 5 6 7 8 9 10 11 12 13 14- 15 16 17 18 19 20 21 22 23 24 25
"Fine" indicated in italics at the left of the
measuring site.
"New Coarse" indicated in bold numbers at the right
of the measuring site.
Figure 15. Spatial distribution of temporal correlation coefficients
R(t) for the "fine (horizontal) grid" and
"new coarse (horizontal) grid" forecasts.
14
-------
Figure 16 shows R(t)s and Rfs)* for the models evaluated by Nappo,
with the two forecasts made in this project added. The best result that a
forecast could achieve would fall at the coordinates 1,1 in this figure.
1.0
.8
.6 _
.4 _
.2 -
.
1
~"
-
1
1
0 E J
F
1 ' 1 ' 1 ' /
X
X
A / -
r
f
i , • .- . .._i /
B c *. X'
^/'
c, *x/'
<£x' ~~
G/
X
t
f _
H/
X
/ ~
X
X
^ B
x I —
/
X
X
X
X
X
X
X
X
/ , 1
-
.1,1,1.
/
KEY
A LAMB AND HE I BURGER (1971)
B ESCHENROEDER, et a1.(1972)
C REYNOLDS, et al. (1973)
D PANDOLFO 4 JACOBS (1973)
E SKLAREW. et al. (1972)
F HANNA (1973)
G ROTH, et al. (1971)
H 24-hr PERSISTENCE
1 MacCRACKEN, et al.(1971)
PANDOLFO, et al. (1976)
• Fine Grid
• Coarse Grid
| 1 95% Confidence Limits •
.2
.4
.6
.8
1.0
Figure 16. R(t)s versus R(s) average result for each model tested.
Also added to Figure 16 are the 95 percent confidence limits obtained by using
Fisher's (1925) transformation for the correlation coefficient, which allows
the confidence limits to be evaluated for an aggregate of several correlation
coefficients, each obtained from a sample with a small number of pairs. These
confidence limits add greater perspective to the significance of the statis-
tics, R(t)s and R(s)fc , with the sample sizes available for Nappo's vali-
dation study.
From Figure 16 it appears that no significant improvement in these sta-
tistics was obtained by going from the "new coarse (horizontal) grid" forecast
to the "fine (horizontal) grid" forecast at a large increase in cost. However,
a barely (at the 95% level) significant improvement was obtained by the "new
coarse" forecast over the "old coarse" forecast at no increase in cost.
The hour-to-hour variations of the spatial correlation coefficient R(s)
are shown in Figure 17. An interesting feature of these variations is that
more skill is attained in the two forecasts tabulated here, for those hours
in which relatively high levels of CO concentration are present—viz., the
morning peak hours (see Figures 5-14) between 0630 and 0930 and the onset of
the evening peak at the last hour of observations (viz., 1630). The lack of
skill in spatial prediction is confined to the hours between 1030 and 1530,
15
-------
0630
0830
1430
1630
'Figure 17.
1030 1230
T I H E
Temporal variation of spatial correlation coefficients R(s)
during which background levels of 3 to 5 ppm are prevalent. This could be a
feature, common to all th-2 primitive equation models evaluated by Nappo> and,
if so, the apparent advantage of the other types of model in obtaining higher
temporal averages of the spatial correlations would be of little operational
significance.
The "spreads" or average standard deviations o(s)t and o(t)s of the
ratios (predicted-to-observed) are shown in Figure 18. The best result that
could be achieved by a model would fall at the origin of this graph. Again,
no significant improvement is achieved in going from the "new coarse (horizon-
tal) grid" forecast to the "fine (horizontal grid" forecast in a large addi-
tional cost. Significant improvement is obtained in going from the "old coarse
(horizontal) grid" forecast to the "new coarse (horizontal) grid' forecast at
no additional cost.
The "spreads" and average values of the ratios are plotted on Figure 19.
The best result that could be achieved by a forecast would be represented by
a point falling at 1,1 with zero spread in both coordinate directions. Here
the differences between the "new coarse (horizontal) grid" and "fine (horizon-
tal) grid" forecasts are so small that only one is plotted. These forecasts
did not improve the average ratio (predicted-to-observed) score obtained by
the "old coarse (horizontal) grid" forecast, which was the best of all the
models surveyed by Nappo.
The values of each of the statistics obtained in the two forecasts are
summarized in Table 2. Also shown in this table are the statistics obtained
in the "finer vertical grid, fine (horizontal) grid" forecast.
16
-------
z.u
1.5
1.0
.5
Q
1 1 1.
X
X
X
X
X
— x'' —
X
/£V''''
*/'' F
- /Cy'
*x
f
f
t
/ H
X
X
t
~ A ,'£ G ° ~
B . /O •
y'c
x
x
1 1 1
KEY
*^B^^B^HB_
A LAMB AND NE1BURGER (1971)
B ESCHENROtDER. ct al.{1972)
C REYNOLDS, et al. (1973)
D PANDOLFO 4 JACOBS (1973)
E SKLAREW. ct al. (1972)
F MANNA (1973)
G ROTH, et al. (1971)
H 21-hr PERSISTENCE
1 MacCRACKLN, et al.(1971)
PANOOLFO, et al. (1976)
0 Fine Grid
• Coarse Grid
0 0.5 1.0 1.5 2.0
-7-\t
our
Figure 18. o(t)s versus a(s)t average result for
each model tested.
KEY
a LAMB AND NEIBURGER (1971)
b MacCRACKEN, et al. (1971)
e SKLAREW, et al. (1972)
-------
TABLE 2
SUMMARY OF STATISTICS FOR "NEW FORECASTS"
GRID
New coarse (horizontal)
Fine (horizontal)
Finer vertical,
Fine (horizontal)
R(t) R(s)
.75 .23
.79 .24
.80 .20
o(t)S
.42
.41
.37
j_
o(s)-w
.50
.52
.44
r(t)~ = r(s) "
1.02
1.04
.98
18
-------
SECTION-5
SUMMARY AND CONCLUSIONS
In comparing the costs incurred by the models he surveyed (see Table 1 of
this report), Nappo ignored the fact that the cost of running the Pandolfo and
Jacobs (1973) model included the generation of four-dimensional (x,y,z,t) fields
of wind, temperature, and diffusivity. None of the other models surveyed were
charged with the presumably greater costs of subjectively and laboriously pre-
specifying these fields on a case-by-case from a sparse set of raw meteorologi-
cal observations. About half the computer costs (— $30/24-hour prediction) of
the coarse (horizontal) grid models described in this report are attributable
to the thermal radiation calculations alone. Of the remaining half of the com-
puter costs, about one-sixth is attributable to the CO prediction (this gross
upper limit is obtained by assuming that the remainder of the primitive equa-
tion prediction of u , v , Tair , q , TsoiiTwater • and co can be equally
apportioned). This means that the coarse (horizontal grid CO predictions
should, at most, be charged with $5/24-hour prediction for a valid comparison
of costs with the other models, except the "persistence" model.
The correct way to look at this factor is, however, the reverse. That is,
the primitive equation model described in this report generates the necessary
meteorological data in the appropriate form for something under $60/24-hr pre-
diction. All of the other models (except the "persistence" model) require, be-
fore 'they can be applied, the specification of spatially and/or temporally vary-
ing fields of at least wind and inversion height for each day and region.
These would have to be obtained with only about four man-hours of a $15,000
per year meteorologist, assuming a $15,000/year overhead load, or the equivalent
in additional computer costs, for our costs to be equal. Therefore, it would be
fairer to add about $60/24-hr prediction to Nappo's tabulated costs (Table 1)
for all models other than those described here, except the "persistence" model.
This having been said, the first conclusion to be stated is one that is
contrary to Nappo's suggestion and to our expectation, in the case of the model
used in this study. Table 2 shows that increasing the degree of detail of the
source emissions inventory did not, in this case, significantly increase the
sensitivity and accuracy of the pollutant concentration forecast. This result,
in hindsight, can only be attributed to the fact that the calculation scheme
used for the horizontal advection calculations—viz., the upwind-difference
scheme—has an inherent degree of approximation that masks the increase in ac-
curacy that might have been expected from treating the source emissions inven-
tory in greater detail.
If we compare results for the "fine (horizontal) grid" and "finer vertical,
19
-------
fine (horizontal) grid" forecasts in Table 2, however, we find that another of
Nappo's findings is confirmed in this case. This is that "the detail of the
vertical diffusion calculations is not very significant in determining a model's
accuracy."
Comparison of the results for the "new coarse (horizontal) grid forecast"
and the "Pandolfo-Jacobs (1973) forecast" shows that making fuller use of the
inherent meteorological capabilities of thd model used in these studies does
significantly increase the accuracy of the forecast. These features were ex-
ploited by combining model-generated vertical profiles of temperature with ob-
served surface temperatures to obtain initial temperature fields, and allowing
radiatively active pollutants to affect the predicted temperatures at no signi-
ficant increase in forecast cost.
Some of the improvement noted in the "new coarse (horizontal) grid" fore-
cast results may also be attributed to the validation of predicted hourly av-
erages instead of predicted single-time step concentrations as was done by
Pandolfo and Jacobs (1973). ' .
Finally, we restate some of the results presented in Table 1 and Figures
16, 18, and 19. These are that in the "coarse (horizontal) grid model" fore-
casts made here,
• the forecast is among the least expensive to run, and is, in
fact, the least expensive to run when it is credited with the
generation of the meteorological predictions needed by all the
other models except the "persistence" model;
® the pollutant forecasts are among the most accurate in terms
of Nappo's R(t)s and r(s)fc statistics; and
• the pollutant forecasts are-of about average accuracy in terms
of Nappo's R(s)fc , ofs)11 , and a(t)s statistics.
The adequacy of. the pollutant forecasts suggests that the meteorological
predictions yielded by the model are well suited for input to any pollutant
prediction model requiring such inputs—e.g., all the models surveyed by Nappo
except the "persistence" model.
The results suggest that, in the near future
• the model described here be applied to generate meteorological
predictions for input to more accurate pollutant forecast models
where great accuracy in spatial detail at smaller than metropoli-
tan region scales is desired;
• the model described here be applied as a pollutant forecast
model where less accuracy in spatial detail is acceptable in
return for low implementation costs; and
• a more accurate (higher order) computational scheme for the hori-
zontal advection calculations be incorporated in the model. The
development effort required to do this is relatively small, and
the accuracy of spatial detail on the intra-metropolitan space
scales could well be significantly enhanced.
20
-------
REFERENCES
Atwater, M.A., 1971: Radiation effects of pollutants in the atmospheric bound-
ary layer. J. Atmos. Sci., 23, 1367-1373.
Fisher, R.A., 1925: Statistical Methods for Research Workers, 1st Edition
(12th Edition, 1954), Oliver and Boyd, Edinburgh.
Nappo, C.J., Jr., 1974: A method for evaluating the accuracy of air pollution
models. Preprints, Symposium on Atmospheric Diffusion and Air Pollution,
Sept. 9-13, 1974, Santa Barbara, California,'American Meteorological Soci-
ety, Boston, Mass., 325-329.
Pandolfo, J.P, M.A. Atwater, and G.E. Anderson, 1971: Prediction by Numerical
Models of Transport and Diffusion in an Urban Boundary Layer, Final Report
Vol. I, The Center for the Environment and Man, Inc., Hartford, Conn.,
139 pp. - ".
, and c.A. Jacobs, 1973: Tests of an Urban Meteorological Pollutant Model
Using CO Validation Data in the Los Angeles Metropolitan Area. Final Re-
port EPA Contract No. 68-02-0223 (EPA-R4-73-025a), NERC, USEPA, Durham,
N.C.
Roth, P.M., S.D. Reynolds, P.J.W. Roberts, and J..H. Seinfeld, 1971: Develop-
ment of a Simulation Model for 'Estimating Ground Level Concentrations of
Photochemical Pollutants. .Final Report, 71 SAI-21, Systems Applications,
Inc., Beverly Hills, Calif., 51 pp., Appendices A-F.
Sklarew, R.C., A.J. Fabrick, and J.E. Prager, 1971: A Particle-in-Cell Method
for Numerical Solution of the Atmospheric Diffusion Equation, and Applica-
tions to Air Pollution Problems. Final Report, 3SR-844, Systems, Science,
and Software, LaJolla, Calif., 163 pp.
21
-------
TECHNICAL REPORT DATA
(Please read laiirucnuns un the reverse he/ur
I. RF.POHT NO.
EPA-600/4-76-037
2.
4. TITLE AND SUUTITLc
REFINEMENT ADD VALIDATION OF AN URBAN METEOROLOGICAL-
POLLUTANT MODEL.
3. RECIPIENTS ACCESSIOI»NO.
5. HEPOfIT DATE
July 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Joseph P. Pandolfo, Clifford A. Jacobs,
Robert J. Ball, and Marshall A. Atwater
8. PERFORMING ORGANIZATION REPORT NO
CEH 1182-542
9. PERFORMING OMCANIZATION NAME AND ADDRESS
The Center for the Environment and Man, Inc.
275 Windsor Street
Hartford, Connecticut 06120
10. PROGRAM ELEMENT NO.
1AA009
11. CONTRACT/GRANT NO.
68-02-1767
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park. M.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final Report 2/24 - 11/76
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report discusses the refinement of a grid point model for predicting at-
mospheric transport, and diffusion in an urban boundary layer. In earlier studies,
a coarse 8-mile horizontal square grid was used to'predict CO concentrations in the
Los Angeles Metropolitan area. In this study, a 2-mile grid for the CO emission has
been used to test a recent hypothesis that predictive accuracy ce.n be improved by in-
corporating a finer horizontal grid resolution to more accurately reflect the spatial
distribution of a pollutant emission pattern. However, contrary to expectations, the
evaluation statistics show that increasing the degree of horizontal detail in the
source emission inventory did not significantly increase the sensitivity and accura-
cy of the pollutant concentration forecast.
When compared to other mode-Is, this pollutant forecast is among the most accu-
rate when predicting the overall average concentration and temporal correlation,
and approximately as accurate as other primitive equation models in terms of spatial
correlation statistics.
KEY WORDS AND DOCUMENT ANALYSIS
DESCflJPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Meld/Group
*Air pollution
*Meteorologica1 data
*Mathematical models
*.Atmospheric models
*Models tests
*Forecasting
13 B
04 B
12 A
04 A
14 B
!. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SfcCUMITY CLASS (Tliil Report)
UNCLASSIFIED
21. NO. Of PAGES
32
20. SECURITY CLASS ITIiu page)
UNCLASSIFIED
F.PA Form 2220-1 (9-73)
22
------- |