-------
f"
TECHNICAL REPORT DATA
ccic rcaJ liiurueiiot't cn the ro tnc bdo'c for.'ili'nn'l
1. REPORT ,'vO.
iiPA-600/3-84-089
2.
3. RECIPIENTS ACCES? IO* i .0.
PBS U 236850
¦». TITLE ANO subtitle
K-VIEW OF THE ATTRIBUTES AND PERFORMANCE OF
SIX URBAN DIFFUSION MODELS
5. R£»OfcT DATE
AURUSt 1984
6. PEHFOPMING organization CODE
7. AUTHORlSI
Fred D. White, Editor
a. performing organization report no.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
American Meteorological Society
45 Beacon Street
Boston, Massachusetts 02108
10. program element no.
CDWA1A/02-0279 (FY-84)
11. CONTRACT/GRANT no.
810297-01
13. S*ONS Ofl'NG AGENCY NAME AND A DOR ESS
Environmental Sciences Research Laboratory
- RTP, NC
13. TYPE OP REPORT AND PERIOD COVERED
Final 10/82 - 9/83
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
14. SPONSORING AGENCY CODE
EPA/600/09
15. Supplementary notes
16. ABSTRACT
The American Meteorological Society conducted a scientific review of a set of
six urban diffusion models. TRC Environmental Consultants, Inc. calculated and
tabulated a uniform set of statistics for all the models. The report consists of
a summary and copies of the three independent model reviews conducted to evaluate
the models. General conclusions included: (1) all of the six models are very
similar to each other and represent simple approximations of the urban diffusion
situations in a given time period with no horizontal variability of the boundary
layer structure or depth; (2) none of the models can be considered state-of-the-art
since a great deal has been learned about the planetary boundary layer that could
be incorporated into such models; (3) the models all use an all or nothing approach
to plume penetration; either the plume penetrates the elevated inversion and is lost
to the compu*- w.ion or it is completely trapped; and (4) the four annual models all
produced good estimates of the observed concentrations, while, of the short-term
models, TEM-8A seriously overpredicted at night and RAM seriously underpredicts
during the day
17.
KEY WOROS ANO DOCUMENT ANALYSIS
a. descriptors
b. IOENTIF IE RS/OPE N ENDED TERMS
c. COSA > 1 Field/Group
is. DlSTAJdUTION STATEMENT
19. SECURITY CLASS (This Report)
UNCI ASSIFIED
21. NO. Of PAGES
106
RELEASE TO PUBLIC
:-:o. security class (Thispttt)
UNCLASSIFIED
32. PRICE
tPA rourt 2210-1 (VT J)
-------
NOTICE
This document has been reviewed in accordance with
U.S. Environmental Protection Agency policy and
approved for publication. Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
11
-------
ABSTRACT
The American Meteorological Society conducted a scientific review of
a set of six. urban diffusion models. TRC Environmental Consultants, Inc.
calculated and tabulated a uniform set of statistics for all the models.
The report consists of a summary and copies of the three independent model
reviews conducted to evaluate the models. General conclusions Included:
(1) all of the six models are very similar to each other and represent
simple approximations of the urban diffusion situations in a given time
period with no horizontal variability of the boundary layer structure or
depth; (2) none of the models can be considered state-of-the-art since a
great deal has been learned about the planetary boundary layer that could
be incorporated into such models; (3) the models all use an all or nothing
approach to plume penetration; either the plume penetrates the elevated
inversion and is lost to the computation or it is completely trapped; and
(4) the four annual models all produced good estimates of the observed con-
centrations, while, of the short-term models, TEM-8A seriously overpre-
dicted at night and RAM seriously underpredicts during the day.
iii
-------
CONTENTS
Abstract iii
1. Introduction and Summary 1
Fred 0. White
2. Urban Model Review 4
Maynard E. Smith
3. Review of Urban Dispersion Models 56
J. C. Weil
4. An Urban Model Review 82
J. C. Wyngaard
v
Preceding page blank
-------
INTRODUCTORY AND S'JMMARY
As one task of. the current Cooperative Agreement between the
American Meteorological Society (AMS) and the Environmental
Protection Agency (EPA), the AMS agreed to conduct a scien-
tific review of a set of six urban diffusion models. This
review was a follow-on to an earlier examination of rural
models which was summarized in a report to EPA, January 1983,
entitled Synthesis of the Rural Model Reviews.
A8 in the review of the rural models, EPA arranged with TRC
Environmental Consultants, Inc. to calculate and tabulate a
uniform set of statistics for all the models. These statis-
tics, which are included in TRC report "Evaluation of Urban
Air Quality Simulation Models," April 1983 , (EPA 450/4-83-020 )
provided the reviewers with a consistent set of measures for
evaluating model performance.
The data base to which model predictions were compared was
acquired with a 13-station network of continuous SO^ monitors
operated in metropolitan St. Louis. The data were obtained
from the EPA/RAMS/RAPS archive. Coincidental air quality and
emissions data for calendar year 1976 were used in this
8 t ud y .
The six urban dispersion models reviewed are all based on
the Gaussian plume and apply to area and point, stationary
sources. Two of these models, RAM and TEM-8A, provide short-
term concentration averages and the remaining four, AQDM,
CDM, ERTAC and TCh, provide annual average concentrations.
Three members of the AMS Steering Committee which oversees
this Cooperative Agreement performed this scientific review.
They were John C. Wyngaard, National Center for Atmospheric
1
-------
Research, Boulder, Colorado; Jeffrey C. Weil, Martin Marietta
Corporation, Baltimore, Maryland; and Mayna rd E. Smith,
Meteorological Evaluation Services, Inc., Amityville, New
York. Copies of their independent reviews form the basis of
this report.
The scientific reviews were prepared during the spring of
1983. On completion, the reviews were then transmitted to
other members of the AMS Steering Committee: Gabriel
Csanady, Woods Hole Oceanographic Institution, Woods Hole,
Massachusetts; Douglas G. Fox, U.S. Forest Service, Fort
Collins, Colorado; and Fred D. White, National Academy of
Sciences, Washington, D . . Collectively we met to discuss
the reviews of the six urban mo dels.
No synthesis of these reviews is planned since the complete
texts are included in this report. However, we would like to
call attention to the following points.
All of the six models are essentially very similar to
each other and represent simple approximations of the
urban diffusion situations in a given time period with
no horizontal variability of the boundary layer struc-
ture or depth.
None of the models can be considered state-of-the-
science since a great deal has been learned about the
planetary boundary layer that could be incorporated into
such models. All such improvements could be handled
easily by currently available computers.
The plume penetration problem that was discussed in the
earlier rural model evaluation is of equal concern in
this review. The models all use an "all or nothing
2
-------
approach;" either the pli>me penetrates the elevated
inversion and is lost to the computations or it is com-
pletely trapped (an unlikely situation).
The four annual models all produced fairly decent esti-
mates of the observed concentrations. The two short-
term models left a lot to be desired. TEM-8A seriously
ove - predicts at night. RAM seriously underpredicts
during the day. Of these two short-term models, RAH
appears to be better.
3
-------
URBAN MODEL REVIEW
Maynard E. Smith
Meteorological Evaluation Services, Inc.
Ami tyville, N.Y.
June 1, 1983
1' SUMMARY
At the request of the AMS-EPA Steering Committee, I have
prepared a review of the set of urban diffusion models
submitted to me, basing my evaluation on the description
of the models in the User's Guides and on the draft TRC
report of the model performance as judged againot the
RAPS/RAMS field data obtained In 1976.
From a scientific standpoint, these models are all very
simplified Gaussian representations of the urban diffu-
sion processes, differing in detail but not in basic
approach. None of the models can faithfully represent
many urban situations, because the complex wind fields,
heat island effects and recirculation patterns Chat are
known to exist in cities like St. Louis cannot be repro-
duced. Furthermore, even in les? complicated situa-
tions, spatial variations in the key input parameters
would have to be taken into account for a physically
realistic representation, and the models cannot do
this.
Another very important scientific weakness is the lack
of time resolution in these models. Pollutant concen-
trations in cities do not respond instantly to changing
meteorological conditions, and the time needed to reach
A
-------
equilibrium must be many hours in light-wind, poor-
ventilation situations.
These scientific limitations are included in my review
largely for completeness rather than for practical, im-
portance. The RAPS/RAM data base, although conceded to
be the best available, would not have perraittad very
sophisticated modeling representations, although
something might have been done to represent a more
detailed wind field.
In terms of performance, three of the four annual mod-
els, AQDM, CDM and TCM performed fairly creditably on
this data base, with the individual prediction8 general-
ly falling within the reasonable statistical limits.
However, ;he ERTAQ seems to be a modest overpredictor.
The short-term modeling evaluation, in which only two
models we re in"o1ved, reveals far more about model
strengths and weaknesses, and it even suggests some
possible reasons for poor model performance. On the
regulatory three-hour and twenty-four-hour time scales,
TEM-8A is an overpredictor and R>,M is an underpredictor.
Which might be chosen would depend upon whether one
wanted to maximize or minimize the calculated values.
The TRC tables summarizing the 25 highest, unpaired
observations show that the observed data do not differ
significantly when segregated into four time periods of
the day, but the predictions show marked variations for
these same time periods, with TEM-8A grossly overpre-
dicting. during nocturnal hours and RAM grossly under-
predicting during daylight houvs. Certainly careful
attention to the data in these cables might result in
5
-------
some helpful adjustments to the modeling concepts and
ass ump t i on s.
The TRC case study tables have been rearranged in my
review so that meteorological, observational and model-
ing data could be compared. A? with the rural models,
there was very little correspondence between a given
prediction and the observed data. This ia not a sur-
prising result, but it does reinforce some of my conten-
tions about the weaknesses of urban modeling as it now
stands. For example, several of the high concentration
periods represented rural, not urban diffusion, and they
should be so treated. In other cases, the models seemed
to have predicted high concentrations associated with
low-level stable cases that did not occur in the city.
6
-------
II.
MODELS AND MODELING COHCEPTS
A. Models and Their General Characteristics
In this review, four models providing long-term
(annual) estimates and two models providing short-
terra (one, three and twenty-four-hour) estimates
were evaluated. The TRC draft report (Apr. 1983)
describing the adjustments, data processing and
performance data for this study includes a reasona-
bly good description of the characteristics of the
model functions, so that there is no point in re-
iterating them here.
As in the case of the rural model review completed
earlier in 1983, there is little to choose among the
models from a scientific standpoint. All are rela-
tively simple applications based on the Gaussian
plume model adapted for broad scale, multi-scource
use. They contain, in various combinations, fa-
miliar algorithms for dealing with plume rise, re-
flection from the ground surface and from the top of
the mixed layer, wind speed profiles, and in some
cases pollutant decay and deposition. The models
have several different methods of dealing with area
sources, including the typical virtual sou re or
ir.itial area and depth representations, and they
also differ in their systems for handling lateral
diffusion, some using uniform concentrations within
angular sectors, others using techniques for smooth-
ing the boundaries between sectors, and one using
the Gaussian approximation itself for the lateral
distribution.
7
-------
One could dwell on these individual difference#,
and it is important to return to some of them in
describing the performance of the models in this
particular study, but fundamentally they are all
very similar to each other in concept, differing
essentially only in details.
* * Scientific Status of the Models
It may seem unfair to criticize these models sci-
entifically because there is no doubt that it would
currently be impossible to provide a sophisticated
model with the input data necessary for it to func-
tion properly. However, it does not seem unreason-
able to document some of the disparities between the
concepts used in the models and our knowledge of
urban flow and diffusion problems. One should at
least recognize that much more is known scientific-
ally than is included in these modeling systems.
1. Many Input Variables Are Assumed to b« Constants
in Space
None of the models permit spatial variation of
any of the following, parameters;
Wind Speed
Wind Direction
Temperature
Mixing Depth
Wind Profile
Diffusion Rates
8
-------
There is no question Chat in a city such as St.
Louis all of these parameters vary significantly
over the area included in this study. In fact,
the meteorology over this area varies from urban
to rural, and on many occasions the assumptions
of constancy in space must be seriously in
error.
The failure to take account of the variations in
mixing height may be especially important be-
cause of the assumption that pollutant rising
above the mixing depth no longer can contribute
to the concentrations.
2. Heat Island Effects
There is also no provision in any of the models
for dealing with the upwelling of air that is
known to take place over a city during very
light gradient winds. This vertical outflow
certainly must act as a ventilating agent,
tending to reduce the concentrations that would
otherwise accumulate.
3. Time Dependence is Mot Accounted For
These urban models use the same sort of rea-
soning applied to rural point source models,
assuming that the air flow extant during a given
hour or joint frequency class must transport
pollutant to the limits of the calculation area,
but that the next hour is an entirely new and
independent situation. This assumption is oc-
casionally inappropriate for an isolated point
9
-------
source, but clearly it does not represent the
urban situation at all well, especially in light
and variable wind conditions. Pollutant that
has been transported over only a portion of the
city dimensions is still there, and should be
reckoned with during the next hour of calcula-
tion. Similarly, pollutant that has passed
beyond the calculation limits or has risen above
the current mixing depth may often return to
contribute to later concentration patterns.
Again, the lighter and wore variable the wind,
the more likely it is that recirculation would
be import ant.
Surely, pollutants must gradually build up and
decrease with time. Even with a steady, invari-
ant wind flow and diffusion rate, a city will
not reach an equilibrium concentration immedi-
ately. In fact, it can be shown by the simplest
of approximations (e.g. Smith, 1961) that both
the equilibrium concentration and the time
it takes to reach it are proportional to
S/u, where S ia a parameter related to
the overall size of the city and u is the
transport wind speed through the volume of the
city air (the ventilation rate). For a city the
size of St. Louis, it may take twenty-four hours
or more to reach an equilibrium concentration if
the mean wind is no more than 1 m/sec, especial-
ly if there is a rather slow decay or removal
process operating on the polluting substance.
The point of the foregoing is not to suggest that
sophisticated models would necessarily do a more
10
-------
accurate job than the simple ones we are reviewing,
hut rather to point out that these models do not
make any pretense of depicting the details of the
urban pollution situation. Furthermore, modern
computer capabilities would permit introducing some
of the sophisticat ion, should anyone desire to up-
date the modeling systems. Se1f-cons is tent wind
fields and real-time, variable-trajectory transport
and diffusion representation are within the state of
the science today.
There are other uncertainties that might be
addressed, such as our lack of experimental veri-
fication of the urban diffusion parameters and the
questions about the initial vertical and horizontal
mixing of the pollutants after release, but these
considerations are secondary to the more fundamental
limitations discussed above.
11
-------
III. REVIEW OF THE DATA. 8ET
A. General Considerations
Although it is conceded that the RAPS/RAMS data are
the most appropriate for a model validation study,
the reviewer must be concerned with the fact that
the entire exercise is based on a single batch of
information. As with the rural evaluation completed
earlier, the reviewer is left with a nagging concern
that there may be little generality in whatever per-
formance has been observed.
A second basic problem is that the SO2 recorders
were incapable of measuring concent rat ions greater
than 1 ppm, a concentration level that could have
been reached and exceeded in a city like St. Louis
on time scales shorter than one hour. Thus, very
important values near the upper end of the concen-
tration range may not be reflected in the data.
& Concern* on Specific Problems
1" High Reading* at Station 104
The processed data showed an annual average of
116 ug/m3 at Station 104, a value more than a
factor of two larger than the next highest sta-
tion and grossly in excess or all predictions
for that Station. One must conclude that there
was either an important source that was unac-
counted for or that there were serious instru-
mental problems.
12
-------
2. Overall Hourly Maximum at Station 104
The highest observed one-hour maximum in the
entire data set occurred at Station 104 on
December 6, (2487 ug/m3) immediately following
an extended period of "missing data" in the
listing. The hour after this showed a concen-
tration of only 755 ug/ra3. One wonders whether
there was something odd about this first hour
after the outage, if it was an outage.
Parenthetically, this type of problem raises
once again the problems inherent in regulatory
systems geared to the extreme upper end of the
data set. Too much depends upon a very few
observed or predicted values.
3. Uncertainty About Mixing Deptha
The equilibrium concentration that would be
reached in a city is directly dependent upon the
depth of the mixed layer, and it i« clear that
the observational network did not permit ade-
quate specification of this variable. This
is an important defect in the data set.
13
-------
RESULTS OF THE COMPARISONS
Partly because the evaluation involved urban models,
four of them predicting only annual numbers, and partly
because EPA-TRC responded to the distress of the earlier
rural reviewers, the data set to be digested was merci-
fully smaller than that prepared for the earlier model
review. It was therefore considerably easier to do some
manipulation of the results to reveal other facets of
the comparisons, and it was also easier to investigate
the individual cases resulting in high predicted and
observed short-term concentrations.
A. Annual Comparison
Three of the four annual models, AQDM, CDM and TCM
performed rather similarly in this evaluation.
Averaged over all stations as shown below, they were
reasonably close to the observed value, and there is
little choice among them. The ERTAQ model, however,
ia a bit farther from the mark, overpredicting by a
modest amount .
Stat ion Measured AQDM CDM TCM ERTAQ
Average 42 42.2 39.2 37.4 50.7
The data become more interesting if one separates
them by station and the level of Che measured values
(Table A). My own immediate reaction was Co ask
what is wrong with the measured value of 116 ug/m3
that occurred at Station 104. It is about twice the
predicted values and begs for a detailed investiga-
tion. Either the measured value is strange or a
major source was misrepresented.
14
-------
TABLE A
ANNUAL COMPARISON
GROUPED BY
LEVEL OF THE MEASURED VALUES
Range of
Measured
Values
Station
Meaa .
AqDM
CDM
TCM
ERTA
104
116
62
62
53
80
>40 ug/m 3
101
55
82
83
102
101
106
52
61
46
40
58
105
43
57
46
34
57
Mean
67
66
59
57
74
As above
without
Station 104
Mean
50
67
58
58
72
108
37
42
42
42
53
103
36
50
51
44
61
35-40 ug/m3
113
36
40
33
27
43
114
35
33
32
29
44
Mean
36
41
40
36
50
12 1
30
25
18
17
30
115
28
29
44
47
50
^35 ug/m3
120
27
23
15
12
25
122
27
20
14
15
27
1 16
24
26
23
24
31
Mean
27
25
23
23
33
15
-------
Leas startling but also interesting is the situation
at Station 101, where 55 ug/ra3 was measured and all
of the predictions were much larger. This too
should be investigated, especially since any pre-
dicted value of 80+ ug/m3 has a special regulatory
s igni £ ic ance.
On a broader perspective it is evident that ERTAQ
overpredicts in comparison to the other models at
almost all concentration levels, and I would con-
sider it the least promising of the group.
It is interesting to note that among the other three
models, there is a slight tendency for overpredic-
tion ac Stations where the measured levels are
theraselven high, shifting to a slight tendency for
underpreH.ic t ion where the measured levels are low.
This same comment applies to distance from the
center of the city. One could indulge in numerous
speculations as to the reasons for this shift, but
the exercise would be beyond the scope of this
review.
B. Short-term Comparison
Although there are only two models involved in this
part of the study, the comparative exercise is far
more interesting and informative with respect to
urban diffusion and modeling problems. Tables 5-3
and 5-4 of the TRC draft report summarize most of
the salient details, and they are reproduced as part
of this review.
Table 5-3 clearly indicates that there are signif-
icant differences between the two models, as well as
16
-------
TABLE 5-3
COMPARISON OF 25 HIGHEST OBSERVED AND PREDICTED SO3 CONCENTRATION VALUES
(UNPAIRED IN TIME Q?> LOCATION) FOR THE ONE, THREE AND TWENTY-FOUR HOUR AVERAGING PERIODS
RAPS (1976)
Model
Average
Observed
Value
(ug/m 3)
Averaging Time: One Hour
TEM-8 1929
RAM
1929
Averaging Time: Three Hours
TEM-8 1351
RAM
1351
Averaging Time: Twenty-Four Hours
TEM-8 664
RAM
664
Average
Predicted
Value
(ug/m3)
3998
1622
2821
811
1312
334
Difference
of Averages*
(Obs.-Pred.)
(ug/m3)
-2069
(-2460, -1678)
307
(137, 477)
-1470
Difference
of Medians
(Obs.-Pred.)
(ug/m 3)
-1838
(-2100,-1614)
325
(224, 456)
-1223
Variance
Comparison*
(Obs./Pred.)
0.07
(0.03, 0.16)
0.47
(0.21, 1.07)
0.05
(-1757, -1183) (-1493, -1101) (0.02, 0.11)
540
(457, 623)
-648
(-792, -504)
330
(237, 423)
584
(476, 614)
-647
1.07
(0.47, 2.43)
0.61
(-750, -524) (0.27, 1.38)
282
19.32
Maximum
Frequency
Difference
1.00 (0.385)
0.76 (0.385)
1.00 (0.385)
1.00 (0.385)
0.88 (0.385)
0.96 (0.385)
(194, 357) (8.51, 43.86)
*95 percent confidence interval in parentheses
-------
TABLE 5-4
COMPARISON OF 25 HIGHEST OBSERVED AND PREDICTED SO ? CONCENTRATION VALUES
(UNPAIRED IN TIME OR LOCATION) FOR THE ONE HOUR AVERAGING PERIOD
RAPS (1976)
TEM-8A RAM
Average
Average
Difference
Average
Average
Di f ference
Observed
Predicted
of Averages*
Variance
Observed
Pred icted
of Averages
Variance
Value
Value
(Obs.-Pred.)
Comparison
Value
Value
(Obs.-Pred.)
Comparison
Data Sets
(ug/m3)
(ug/m3)
(ug/m 3)
(Obs./Pred.)
(ug/m 3)
(ug/m 3)
( ug /m 3)
(Obs./Pred.)
By Station:
Station 3
(I01)a
921
3940
-3019
0.18
921
904
17
10.50
Station 2
(103)
511
1678
-1167
0.54
511
599
-88
16>6
Station 3
(104)
1886
1721
165
0.23
1886
1133
753
1.72
Station 4
(105)
421
1798
-1377
0.11
421
884
-463
0.33
Station 5
(106)
652
1419
-767
1.78
652
603
48
16. 10
Station 6
(108)
427
1595
-1168
1.61
427
626
-199
16.64
Station 7
(113)
440
1047
-607
1.43
439
514
-75
5.28
Station 8
(il4)
452
1585
-1133
0.13
452
1053
-601
0.35
Station 9
(115)
353
2336
-1983
0.02
353
1461
-1108
0.04
Station 10
H
Am
(116)
438
1351
-913
0.10
438
497
-59
3.12
Station 11
(120)
398
1093
-695
0.29
398
426
-28
4.66
Station 12
(121)
320
1139
-819
0.02
320
544
-224
0.19
Station 13
(122)
435
1022
-587
0.65
435
451
-16
9.11
By Tine of Day
:
0000-0600
1561
3801
-2240
0.11
1561
1424
137
0.60
0600-1200
1474
2274
-800
0.21
1474
881
593
1.55
1200-1800
1467
1443
24
1.11
1467
792
675
1.72
1800-2400
1329
2772
-1443
0.21
1329
1305
24
2.21
Py Meteorological Condition:
A. Wind Speed
<2.5 a/a
1671
3998
-2327
0.10
1671
1620
51
0.68
2.5 to
5 u/a
1696
1830
-134
3.87
1696
982
714
3.79
>5 m/&
881
977
-96
3.87
881
472
409
4.88
B. Stability Group
Class A & B
657
784
-127
8.43
657
441
216
22.20
Class C
1333
1131
202
0.96
1333
539
794
15.21
Class D
1509
2463
-954
0.36
1509
688
821
6.49
Class Z & P
1774
3970
-2196
0.09
1774
1622
152
0.65
BAPS/SAMS Monitoring ID codes in parenthesis
-------
significant differences between each of the models
and the measured values. The data presented here
are summaries of the 25 highest unpaired observed
and predicted values, and they show that TEM-8A is
definitely an overpredictor while RAM is an under-
predictor. Using the three-hour and twenty-four-
hour time periods as guides because of their
regulatory significance, one would be grossly dis-
appointed in the performance of either. Perhaps the
best solution would be to average the mean results
of the two models; it would provide a fairly good
estimate of the measured meanst
I find it hard to limit ray comments strictly to
model performance, especially when Table 5-4 lends
'tself so well to suggesting what may be wrong with
the two models. First, the segregation by time of
day in the center of the table shows that the meas-
ured values are almost identical throughout the four
six-hour periods of the day, whereas both models
show marked variation from night to day. TEM-8A is
close to the mark in the afternoon but overpredicts
seriously at other times, especially at night. RAM,
on the other hand, does well at night but seriously
underpredicts during the daylight hours.
One must speculate that the observed lack of varia-
bility over the day-night grouping may well imply
that diffusion in a city does not vary nearly as
much from day to night as the modeling concepts
imply, or, equally likely, that there is such a long
lag time in the change of pollutant concentrations
in a city that no model of the type under investiga-
tion would perform satisfactorily.
19
-------
V. CASS STUDIES
The TRC draft report
situations chat resuV
highest one-hour, thr
dieted and observed d
copied these tables an
look at the meteorolo
try to glean some anuj
happening. The sets
increasing time period
ending with twenty-£c
receptors and the cas
c s complete data for those
the highest and second-
noted before. Perhaps
perhaps it is not.
1 together show the details
rvat ion of the data set. One
i is the fact that this very
of "missing data," as we have
is a real observation, but
Highest One-Hour Fred
RAM
The tables and figure
of information for the
RAM model. In this case,
of the so-called Wood R i
Sioux power plant near A!
urban situation and it mi
curately with a rural mod
nothing much happened a
,n.ued as 2 show the same type
.iicst hourly prediction by the
the receptor (9) was downwind
r refinery complex and the
ion, Illinois. This is not an
ht have been treated more ac-
l. Interestingly, either
o monitor, or the data were
?.[)
-------
TAm.:: D-1
SELECTED DAYS OF DATA . SELECTION CRITERIA
Date Criterion Hour Ending Receptor
01/15/76 RAM-H3 3 104
01/26/76 RAM-2H24 24 115
01/26/76 TEM-2H24 24 115
01/27/76 RAM-HI 5 115
08/23/76 RAM-2 H1 5 114
08/23/76 TEM-2H1 5 114
10/28/76 TEM-H1 5 101
10/28/76 TEM-H3 6 101
11/15/76 TEM-H24 24 101
11/16/76 * * *
12/06/76 OBS-H1 14 104
12/11/76 0BS-H3 18 104
12/15/76 0BS-H24 24 104
12/31/76 RAM-H24 24 101
* Several high or second-high .lues were observed and
predicted for one, three am! twenty-four-hour averaging
periods on 11/16/76.
RAM - RAM model predicted concentration
TEM - TEM-8A model predicted concentration
OBS - observed concentration
HI, H3, H24 ¦ Highest one, thr.;.> and twenty-four-hour average
concentration f^ the year.
2H1, 2H3, 2H24 ¦ Second high eat one, three and twenty-four-
hour average concentration for the year.
21
-------
bad, as indicated by the frequent missing designations
for Che day.
Second-Highest One-Hour > miction* - RAH and TEM-8A
The Cables and figure in the "3" set represent the
second-highest one-hour .diction by both RAM and TEM-
8A. With the northeast vinJ indicated at the time, this
case again represents th effect of the Wood River com-
plex, and it is not a tr 1 y urban situation. Once again
the monitor allegedly at:cced showed nothing of any
eignificance.
Highest One-Hour and Thr ¦ ¦-ilour Predictions - TEM-8A
TEM-8A predicted its highest one-hour and highest three-
hour concentrations on the same day and during the same
time interval, as shown Tables AA and 4B and Figure
4. The wind speeds were very light and variable during
these early morning hours and the calculated concentra-
tions were associated with stable diffusion and a low
mixing depth. The monitor appears to have been totally
unaware of the situation.
Bighast Obsarved Three-Ko-r Case
The highest observed three-hour case was associated with
light SSW winds and neutral stability. The several
large sources lying upwind of the receptor were no doubt
responsible for the concentrations.
Hlghast Thf-Hour Prediction - RAH
In this instance, the RAN model seems to have singled
out the sama sources and receptor as in the highest
22
-------
observed three-hour case, bur. In this instance the air
was supposedly more stable and the mixing depth,was very
shallow. The monitor die! show some S02, but the values
were modest.
Highest Observed Twenty-Four-Hour Concentration
This particular day, summarized in the "7" series, prob-
ably identifies the source or source area contributing
8o substantially to the high annual average at this
station. The winds were consistently between SW and W,
with a very steady wind speed and temperature. The
designated stabilities, ranging from 6 to 3 throughout
the period, probably do net reflect the nearly-neutral
conditions suggested by the winds and temperatures.
Highest Twenty-Four-Hour Prediction - RAM
The "8" series shows that the RAM model predicted its
three-hour maximum on a day with persistent WNW winds of
moderate speed. It is difficult to see from the map
layout just which source(s) was supposed to be responsi-
ble, and the monitor showed no such effect.
Highest Tirenty-Four-Hour Prediction - TEM-8A
The "9" series shows that TEM-8A predicted its twenty-
four-hour maximum with light NG winds. The monitor did
receive some SO2, but significantly less than Station
104.
23
-------
Second-Highest Tvnty-Four-Hour Predictions - RAM and
TKM-8A
The final set of data selected by both models again
represented an essentially rural situation, with the
Wood River complex and the Sioux power plant as the
responsible sources. In this instance, the chosen mon-
itor did show elevated SO2 concentrations, although not
as high as those predicted.
24
-------
RKF2KHNCF8
Londergan, R.J. et al. 1983. Evaluation of Urban Air Quality
Simulation Models, (Draft) TRC Env. Consultants, Proj.
1829-E81.
Smith, H.E. 1961. The Concentrations and Residence Times of
Pollutants in the Atmosphere, Proceedings, Symposium on
"Chemical Reactions in the Tower and Upper Atmosphere,"
SRI.
25
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
il
as
5
5
5
6
5
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
TABLE 1A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LODIS RAPS/RAMS STATIONS
12/6/76
HIGHEST ONE-HOUR OBSERVED
Hind Direction
Wind Speed
Temperature
Mixing Height
(Degrees)
(M/S)
(Degree K)
(Meters)
133
2 .42
273.71
132.00
130
2.38
273.71
132.00
124
2. 16
273.71
132.00
135
2.01
274.26
132.00
144
1.41
274.26
132.00
142
1.29
273.71
712.30
164
1 .57
274.26
704.63
168
2.36
274 .82
696.97
179
2 .05
275.37
689.31
187
1.70
275.93
681 .65
184
2 .07
276.48
673.99
192
1.77
277.04
666.32
213
1 . 68
277.04
658.66
328
1.69
275.93
651.00
348
4 . 75
274.26
651.00
352
5.21
273.71
651 .00
353
5.31
273.15
647.98
348
5 .40
272.04
640.50
350
6.24
271.48
633.03
348
6.12
270.93
625.55
340
5 .84
269.26
618.08
334
6. 18
268.15
610.60
334
6.58
266.48
603.13
332
6.83
265.37
595.65
-------
idi
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
39
45
38
37
45
97
96
83
88
78
46
29
32
34
38
31
-1
15
8
7
12
7
-1
-1
TABLE IB
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/M3) BY STATION
2 12/6/?6
HIGHEST ONE-HOUR OBSERVED
Stat ion
101
103
104
105
106
108
113
114
115
116
238
52
j
84
466
35
144
26
37
7
185
50
-I
78
469
26
177
12
35
7
205
38
-1
73
472
18
138
8
36
7
198
-1
-1
63
443
12
113
9
29
7
136
45
-1
57
394
13
109
12
-1
7
126
22
-1
52
291
12
88
10
22
7
140
14
-1
38
213
41
94
111
28
7
163
24
-1
23
748
110
160
140
32
7
142
32
-1
41
1268
105
335
208
9
7
120
88
-1
27
1007
192
-1
154
7
7
116
-1
-1
59
400
191
-1
195
7
7
63
58
-1
85
-1
122
-1
75
7
7
106
40
X
43
1260
176
-1
131
7
-1
144
199
|2487|
243
327
124
-1
56
29
7
77
74
755
79
290
31
48
23
32
49
-1
33
187
108
459
31
64
32
37
7
56
31
152
90
412
28
45
16
-1
7
42
26
262
6 9
320
14
27
7
29
7
36
19
176
49
289
10
19
7
24
7
23
7
331
41
254
7
-1
7
9
7
35
12
12.1
27
174
7
-1
7
62
7
7
10
118
10
74
7
-1
7
110
11
7
11
273
-1
-1
7
-1
-1
45
-1
-1
13
-1
-1
-1
8
-1
-1
-1
1
(-1 indicates
miss ing
data)
-------
4330
4320
4310
4300
4290
4280
4270
4260
4250
4240
4230
4220
7
1*
122
Wind Direction
25km
115
116
~
V
I
N
u
LEGEND
S02 POINT SOURCES
< 50 ( xlO^ kg/yr)
O 50-250
~ >250
SO2 MONITORS
X RECEPTORS (RAMS 10)
~
20km
r
lot
0 720 730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 1
Highest One-Hour Observed
(A receptor number encloaed in a rectangle denotes a maximum
observed or predicted velue.)
28
-------
d i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
i
a_
7
7
7
7
6
6
6
5
4
3
2
2
2
2
2
3
3
4
5
6
5
5
5
TABLE 2A
HOPRLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
1/27/76
HIGHEST ONE-HOUR PREDICTED - RAM
Hind Direction
Vind Speed
Temperature
Mixing Height
(Degrees)
(M/S)
(Degree K)
(Meters)
350
2.40
263 .70
159.00
348
1.62
263. 15
159.00
333
1 . 69
263. 15
159.00
328
1 . 50
262.59
159.00
314
1 .48
262.59
159.00
300
1. 92
262.04
159.00
301
1.92
262.59
159.00
295
2.06
263.70
244.33
304
1 .83
265.37
363.94
323
1 .95
266.48
483.55
318
1 . 93
267.59
603. 16
302
1 .48
268 . 7 1
722.78
293
1 . 68
269 .82
842.39
296
1 . 38
270. 37
962.00
289
. 79
270.93
962.00
203
1.21
270.93
962.00
170
1 . 19
269.82
962.00
106
1. 38
269 .26
952 . 37
120
1.81
268 . 7 1
747.87
116
2.36
268 . 7 1
630 . 10
132
2 . 96
268.71
512.32
149
3.52
268 . 71
394.55
152
4. 14
268.71
276.77
158
4.44
268.71
159.00
-------
TABLE 2B
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/M3) BY STATION
1/27/76
HIGHEST ONE-HOUR PREDICTED - RAM
Hour Station
Ending
101
103
104
105
106
108
113
114
115
Alt
120
121
122
1
22
7
14
99
56
8
7
7
7
9
7
10
10
2
14
7
42
23
162
7
7
7
7
24
7
7
3
22
7
63
28
127
7
7
7
-1
57
7
7
4
29
17
71
44
12
7
7
7
m
60
7
7
5
17
12
100
14
12
7
7
7
35
7
7
6
38
7
94
20
13
7
8
7
9
7
7
7
49
10
146
79
10
7
17
7
7
7
7
8
53
9
112
84
23
7
21
7
-1
7
7
7
9
-1
27
153
81
25
7
16
7
-i
76
7
7
10
-1
31
85
56
15
7
7
7
-i
205
7
7
11
-1
19
114
46
7
168
8
25
-i
290
7
7
10
12
-1
9
106
39
13
10
7
7
-1
188
7
7
13
-1
35
98
30
7
8
7
7
-1
152
7
7
14
12
51
25
14
7
7
7
7
-1
128
7
7
15
-1
40
46
23
7
8
7
7
-i
36
7
7
16
80
46
68
28
7
7
7
7
27
7
7
7
17
55
39
26
38
7
7
7
7
9
7
7
7
18
50
13
52
58
15
7
8
7
7
7
7
7
19
51
28
83
11
10
7
7
7
10
17
7
7
20
106
44
77
21
66
7
10
7
90
41
66
7
21
81
32
67
30
66
30
7
74
67
186
7
22
171
39
55
31
60
11
122
32
64
61
123
-1
11
23
-1
115
98
-1
-1
41
127
89
-1
-1
75
153
-1
24
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
(-1 indicates missing data
-------
Wind Direction
4300
5 4290
M
O
<3 4280
2 4270
LEGEND
SO2 POINT SOURCES
<50 (x 103 kg/yr)
O50-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 10)
25km
0 4260
4250
4240
20km
4220
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 2
Highest One-Hour Predicted - RAM
(A receptor number enclosed in a rectangle denote, a maximum
obaerved or predicted value.)
31
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i
a_
6"
5
4
3
2
2
2
3
3
4
3
3
4
5
5
5
6
6
TABLE 3A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
8/23/76
SECOND-HIGHEST ONE-HOUR PREDICTED - RAM & TEM
Wind Direction Wind Speed Teaperature Mixing Height
(Degrees ) ( M / S ) (Degree K) (Meters)
157 1.04 293.71 135.00
140 1.33 293.15 135.00
146 1.34 292.60 135.00
34 .91 292.04 135.00
62 .75 291.48 135.00
82 .76 292.04 266.78
59 1.04 294.82 504.06
56 1.12 297.04 741.33
58 1.19 299.82 978.61
88 1.78 301.49 1215.89
112 2.34 302.60 1453.17
106 3.03 303.15 1690.44
93 3.56 302.60 1927.72
90 4.07 302.04 2165.00
76 4.18 302.60 2165.00
74 4.31 302.04 2165.00
75 3.96 301.49 2165.00
75 3.19 3C0.37 2165.00
75 2.82 299.26 2047.15
78 2.57 298.15 1664.72
86 1.94 297.60 1282.29
83 1.94 297.04 899.86
83 1.50 296.49 517.43
52 1.53 295.37 135.00
-------
id i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
10
15
25
-1
16
-1
15
2 1
14
17
13
24
15
18
20
15
-1
13
- 1
12
-1
22
-1
TABLE 3B
SELECTED HODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/M*) BY STATION
8/23/76
SECOND-HIGHEST ONE-HOUR PREDICTED - RAH 8. TEM
Station
101
103
104
105
106
108
113
114
115
116
120
121
11
|
7
7
27
i
-1
7
7
7
7
42
7
-1
7
7
25
-1
-1
7
7
7
7
37
7
-1
7
7
20
-1
-1
7
7
7
7
10
28
-1
7
7
17
7
-1
7
7
7
7
7
39
-1
7
7
12
7
-1
m
7
7
7
7
7
10
-1
7
18
7
-1
7
7
7
7
10
-1
7
7
16
7
7
7
7
7
7
24
-1
21
9
49
-1
-1
6
6
6
34
30
-1
42
178
92
10
-1
-1
6
6
6
30
13
-1
40
143
101
6
-1
-1
6
6
6
234
57
19
40
143
-1
6
73
-1
6
34
141
76
-1
58
64
-1
6
-1
-1
6
6
54
-1
12
-1
21
69
- 1
12
-1
-1
6
6
-1
-1
7
-1
7
92
71
- 1
-1
6
6
6
-1
- 1
6
-1
6
70
36
- 1
-1
6
6
6
-1
-1
6
-1
6
111
29
10
-1
6
6
6
15
45
6
-1
6
103
32
6
-1
6
6
6
1.3
42
6
6
68
32
7
-1
6
6
6
9
36
6
-1
6
29
36
6
-1
6
6
6
6
34
6
6
6
44
42
-1
-1
6
6
6
6
33
18
6
6
191
32
7
-1
6
6
6
6
-1
9
6
6
152
78
-1
-1
6
6
6
6
-1
11
-1
6
202
-1
-1
-1
7
-1
-1
-1
-1
-I
-1
-1
-1
-I
7
-1
-1
-1
-1
-1
-1
(-1 indicates missing data)
-------
4330
4320 -
Wind Direction
4300
4290
$4270
5
25km
4260
4250
4240
N
A
4220
710
720
LEGEND
S02 POINT SOURCES
<50 (x 103 kg/yr )
0 50-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 10)
20km
730 740 750 760 770 700
UTM COORDINATES (EASTING )
790
800
810
FIGURE 3
Second-Highest One-Hour Predicted - RAM & TEM
(A receptor number enclo.ed in t rectangle denote, « maximum
observed or predicted value.)
34
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
TABLE 4A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
10/28/76
HIGHEST ONE-HOOR PREDICTED - TEM
HIGHEST THREE-HOUR PREDICTED - TEM
Hind Direction Hind Speed Temperature Mixing Height
(Degrees) (M/S) (Degree K) (Meters)
40 .80 273.15 116.00
46 .70 272.59 116.00
270 .58 272.59 116.00
340 .67 272.04 116.00
16 T31 271.48 116.00
135 .71 271.48 116.00
178 T70l 272.59 179.73
184 1.11 275.37 301.34
187 1.71 277.04 422.95
198 2.86 278.71 544.56
194 2.47 279.82 666.17
183 1.28 280.37 787.78
187 1.98 281 .48 909.39
196 2.20 282.04 1031.00
209 2.46 282.04 1031.00
202 2.39 281.48 1031.00
191 2.17 279.82 1031.00
174 1.82 278.71 902.93
161 2.03 277.04 770.44
168 2.40 277.04 637.95
173 2.42 276.48 505.46
177 2.50 275.93 372.98
180 2.92 275.93 240.49
188 3.33 275.37 108.00
-------
idi
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
122
7
7
7
7
7
7
7
7
7
42
138
172
136
230
158
60
38
60
41
22
33
22
58
-1
TABLE 4B
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/MJ) BY STATION
10/28/76
HIGHEST ONE-HOUR PREDICTED - TEM
HIGHEST THREE-HOUR PREDICTED - TEM
Station
101
103
104
105
106
108
113
114
115
116
120
121
9
. 2
9
-1
26
8
7
8
7
7
-1
10
11
7
18
9
22
9
7
7
7
7
-1
8
19
7
32
20
-1
10
7
-1
7
7
-1
8
11
7
26
72
23
8
7
-1
7
7
-1
7
Uil
40
50
11
8
7
-1
mm *
7
-1
7
9
7
84
8
12
7
8
-1
7
7
-1
7
10
11
86
9
22
12
17
-1
7
7
-1
7
21
26
63
8
34
9
42
-1
7
7
-1
7
17
32
37
23
229
69
37
-1
9
7
-1
8
140
62
264
52
366
121
7 1
- 1
10
7
-1
48
81
28
92
-1
167
62
159
-1
36
7
-1
112
48
-1
66
45
50
70
121
-1
88
7
-1
81
55
-1
67
48
46
82
87
- 1
94
7
-1
44
59
-1
85
55
79
98
72
-1
88
7
-1
90
74
-1
54
81
132
70
61
-1
76
7
-1
86
126
55
61
80
146
9 7
110
-1
63
7
-1
60
108
62
66
70
147
132
94
-1
49
7
-1
80
147
116
159
140
178
91
92
-1
41
31
-1
71
115
129
183
102
140
127
152
-1
45
24
-1
68
73
48
105
58
94
99
173
-1
45
7
-1
129
61
23
103
30
45
50
110
-1
31
7
-1
171
41
9
80
25
-1
54
56
-1
-1
7
-1
206
-1
22
45
18
21
50
-1
-1
-1
7
-1
-1
-1
10
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
(-1 indicates missing data)
-------
4300
Q 4290
m.
< 4270
LEGEND
S02 POINT SOURCES
<50 (Xto3 kg/yr)
0 50-250
~ >250
S02 MONITORS
X RECEPTORS {RAMS 10)
25km
P 4260
4250
4240
i20km
4220
710
730 740 750 760 770 780
UTM COORDINATES (EASTING )
FIGURE 4
Highest One-Hour Predicted - TEM
Highest Three-Hour Predicted -TEM
(A. receptor number enclosed in « rectangle dene tea a maximum
observed or predicted value.)
37
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
>i'
_aj
4
5
5
5
5
5
5
4
4
3
4
3
3
3
4
4
4
4
4
4
4
4
4
4
TABLE 5A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
12/11/76
HIGHEST THREE-HOUR OBSERVED
Wind Direction
Wind Speed
Temperature
Mixing Height
(Degrees)
(M/S)
(Degree K)
(Meters)
11
4.04
266.48
622.45
13
3.68
265.93
188.00
13
3.24
265.93
188.00
6
2.99
265 .93
188.00
10
2.59
265.93
188.00
10
2.99
265.93
188.00
8
3.17
265.93
188.00
19
2.77
265.93
256.90
2/
2 .48
266.48
349.91
26
2.32
266.48
442.93
40
1 . 56
267.04
535.95
36
.94
268.15
628.97
21
.66
269.26
721.98
296
. 72
269.82
815.00
250
.91
269.82
815.00
180
1 .81
270.37
815.00
190
1.63
270.37
815.00
192
1 .80
270.93
815.00
199
1.16
270.93
815.00
194
1 . 10
271.48
815.00
177
1 .27
271.48
815.00
191
1 .34
271.48
815.00
200
2 .32
272.04
815.00
211
2.41
272.04
815.00
-------
101
32
30
SO
59
78
94
87
89
123
124
166
165
194
-1
-1
360
300
301
371
385
415
350
258
153
22
7
7
7
7
7
11
27
44
49
73
82
18
63
22
69
69
69
68
61
66
95
84
TABLE 5B
SELECTED MODEL EVALUATION INPPT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (PG/M3) BY STATION~
12/11/76
HIGHEST THREE-HOUR OBSERVED
Stat ion
103
104
105
106
108
113
114
115
116
120
52
360
63
63
7
7
7
7
10
7
81
570
73
36
12
7
7
7
7
7
117
702
97
59
18
7
7
11
8
-1
-1
788
143
93
31
7
15
19
28
15
78
541
159
102
45
7
7
15
36
34
124
1114
198
121
36
11
7
17
42
50
83
776
179
121
26
8
7
10
63
29
108
793
160
109
38
8
27
19
47
36
117
473
276
90
111
47
69
51
36
61
116
-1
249
134
141
1C9
97
58
42
3 V
106
-1
314
151
155
159
159
51
36
55
86
757
247
221
76
200
218
45
36
64
98
1126
300
227
83
188
203
-1
35
76
139
1564
314
283
159
202
202
-1
35
67
228
1741
330
191
285
129
141
193
60
58
300
1998
325
172
279
136
198
138
157
63
333
1711
286
324
324
217
245
174
217
64
238
1476
359
-1
-1
-1
342
-1
361
61
456
-1
532
-1
401
-1
347
291
554
59
498
2099
556
-1
309
-1
334
337
-1
60
441
-1
513
-1
423
-1
311
389
500
73
494
-1
347
-1
464
-1
305
368
406
58
331
1735
227
-1
413
-1
378
383
437
58
195
1894
153
-1
292
-1
262
386
305
39
(-1 indicates missing data)
-------
4330
4320
4310
4300
O
g4290
P$
O
534200
^ 4270
S
§
O 4260
^4250
4240
4230
4220
7
Pr^TT
25km
'115
H16
LEGEND
S02 POINT SOURCES
<50 (x 103 kg/yr)
O50-250
0> 250
S02 MONITORS
X RECEPTORS (RAMS 10)
1
I
I
20 km
al_
0 720 730 740 750 7«Q 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 5
Highest Three-Hour Observed
(A. receptor number enclosed in a rectangle denotes a maximum
observed or predicted value.)
40
-------
di
~T
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i
a_
z
6
6
6
6
6
5
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
TABLE 6A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
1/15/?6 -
HIGHEST THREE-HOUR PREDICTED - RAM
Wind Direction
(Degrees )
Wind Speed
(M/S)
Temperature
(Degree K)
Mixing Height
(Meters)
213
203
198
190
191
186
176
175
179
177
182
181
186
193
203
207
232
228
222
223
230
230
233
257
2.89
2.72
2.52
2 .53
2.93
3. 14
3.32
4.36
4.52
5.22
5. 70
6 .68
6.83
7 .40
6.61
5 . 76
8.48
8 . 22
6.32
5.43
5.15
4.58
3.98
4. 15
270.37
269.82
269.82
269.82
269.26
269.82
270.37
272.04
273.71
275.37
278. 15
279.82
280.37
280.93
280.93
280.93
280.37
279.82
279.82
279.82
279.82
279.26
279.26
279.26
159.00
159.00
159.00
159.00
159.00
159.00
159.00
249.94
400.12
550.29
700.47
850.65
1000.82
1151.00
1151 .00
1151.00
1150.49
1137.85
1125.21
1112.56
1099.92
1087.28
1074.64
1061.99
-------
din
X
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
7
7
7
7
7
7
20
18
49
69
22
31
60
68
16
8
TABLE 6B
SELECTED MODEL EVALUATION IHPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/MJ) BY STATION
l/l5/?6
HIGHEST THREE-HOUR PREDICTED - RAM
St at ion
101
103
104
105
106
108
113
114
115
116
120
37
20
202
32
18
56
-1
23
40
47
7
35
21
123
27
34
94
-1
50
62
29
7
84
8
106
24
40
66
-1
44
30
33
7
187
11
121
24
120
115
-1
107
11
25
9
253
24
89
23
85
147
31
55
7
44
16
259
27
82
27
101
195
-1
73
19
99
8
241
95
92
54
159
240
-1
142
181
70
14
306
96
98
30
85
194
-1
159
82
-1
7
41
51
135
12
63
72
-1
202
7
-1
7
48
107
96
20
93
70
-1
62
7
22
7
73
28
55
17
80
68
-1
62
7
8
7
7
33
45
7
85
30
-1
35
7
7
7
33
46
40
8
151
37
-1
69
7
7
7
68
56
182
7
200
64
-1
66
7
7
7
186
7
215
43
153
132
-1
86
7
7
84
56
279
282
13
179
-1
7
20
7
7
8
52
185
106
14
7
-1
7
12
11
11
8
-1
129
163
19
7
-1
7
7
7
7
17
- J
178
145
14
7
-1
7
7
7
7
7
1
193
242
18
7
-1
7
7
30
7
9
1
78
74
18
8
-1
14
7
70
86
10
-1
69
58
14
7
119
116
18
75
277
-1
1
79
113
20
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
(-1 indicates nissing data)
-------
4330
4320
4310
4300
4290
4280
4270
4260
4250
4240
4230
4220
7
25km
115
~
Wind Direction
N
A
LEGEND
S02 POINT SOURCES
<50 (x 103 kg/yr)
050-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 10)
i
.20km
1 tml
0 720
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
610
FIGURE 0
Highest Three-Hour Predicted - .RAM
(A receptor number enclosed in a rectangle denotes a maximum
observed or predicted value.)
43
-------
idi
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i
a_
6
6
6
6
6
6
6
5
4
3
3
3
3
3
4
4
5
6
6
6
5
5
5
5
TABLE 7A
HOURLY METEOROLOGY
_
OMPOSITE FROM
25 ST. LOUIS RAPS/RAMS STATIONS
12/15/76
HIGHEST
TWEHTY-FOUR HOUR
OBSERVED
Wind Direction
Wind Speed
Temperature
Mixing Height
(Degrees)
(M/S)
(Degree K)
(Meters)
214
2 . 90
274.82
124.00
217
3.02
274.26
124.00
224
2.60
274.26
124.00
237
2.54
273.71
124.00
260
2.44
273.71
124.00
262
2.48
273.15
124.00
265
2.35
273.15
124.00
244
2.37
275.37
189.93
243
2.02
277.04
285. 10
244
2.26
278. 71
380.28
261
2.24
280.37
475.46
262
2. 13
282.04
570.64
270
2. 70
283.71
665.82
269
2.78
284.26
761.00
269
2 . 72
284.26
761.00
271
2.50
282.60
761.00
245
2.09
280.93
730.09
237
2. 18
279.26
650.94
219
2 .47
278. 71
571 .78
230
2.80
278.71
492.63
237
3.06
278.15
413.47
248
3.57
278.71
334.31
282
4.91
278. 15
255.16
303
5.34
277.04
176.00
-------
TABLE 7B
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (PG/MJ) BY STATION
12/15/?6
HIGHEST TVENTY-FOUR HOUR OBSERVED
Hour Station
Ending
101
103
104
105
106
108
113
114
115
116
120
121
122
i
-1
95
1835
62
83
93
20
38
53
-1
25
1
7
2
-1
53
-1
68
56
112
27
56
78
-1
9
- 1
7
3
75
60
1807
64
130
101
50
48
82
85
20
1
8
4
52
65
1365
53
64
112
44
58
119
80
13
- 1
12
5
43
51
1551
33
40
93
13
47
87
112
7
i
8
6
32
62
1129
42
71
87
8
7
58
113
7
1
7
51
65
-1
44
45
53
7
7
61
-1
7
1
8
55
125
1725
108
46
47
7
7
64
-1
7
1
9
77
90
1302
101
34
44
8
7
78
-1
1
10
59
79
1376
79
58
28
7
7
80
48
-1
1
11
52
72
1163
148
34
13
9
28
53
7
7
12
38
65
518
28
24
82
7
7
7
9
7
1
13
37
34
341
30
28
73
7
7
7
7
7
- 1
14
25
19
206
26
36
119
7
7
8
7
8
1
15
21
29
237
46
42
82
7
7
7
7
8
- 1
16
40
31
466
122
-1
11
-1
7
7
7
12
1
17
160
71
1367
297
-1
7
-1
7
7
7
16
- 1
18
349
271
2206
582
610
9
39
7
7
7
15
1
19
493
419
-1
497
588.
38
89
22
7
54
34
1
20
124
260
1460
100
286
110
78
41
28
120
46
1
21
72
76
2135
63
78
87
48
12
109
29
25
10
1
22
70
55
-1
54
64
68
8
7
108
60
11
7
-1
23
73
120
-1
63
-1
24
24
7
80
-1
-1
7
1
24
-1
45
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
1
(-1 indicates aissing data)
-------
4300
O
S 4290
2 4280
<4270
LEGEND
S02 POINT SOURCES
<50 (x 103 kg/yr)
0 50-250
~ >250
S02 MONITORS
X RECEPTORS {RAMS 10)
25km
4260
4250
20km
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 7
Highest Twenty-Four Hour - Observed
(A receptor nuaber enclosed in a rectangle denotes a maximum
observed or predicted value.)
46
-------
TABLE 8B
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOUiLY MEASURED SO CONCENTRATIONS (UG/MJ) BY STATION
12/31/76
HIGHEST TWENTY-FOUR HOUR PREDICTED - RAM
Hour Stat ion
Ending
101
103
104
105
106
108
113
114
115
116
120
121
122
1
8
8
127
8
9
8
8
8
117
8
8
8
-1
2
8
8
423
8
11
8
8
8
210
48
8
8
1
3
8
8
253
8
15
17
8
45
244
72
-1
8
1
4
' 8
8
146
8
17
11
8
20
166
34
8
8
1
5
33
-I
69
14
8
-1
8
8
88
8
8
8
1
6
22
10
72
22
8
3
14
8
19
20
-1
8
- 1
7
34
10
60
30
8
8
15
8
19
18
-I
8
1
8
13
8
13
73
8
8
8
8
8
8
-1
8
1
9
59
8
163
55
8
8
10
8
8
8
-1
8
" 1
10
22
9
631
44
20
124
8
71
29
35
8
8
1
11
32
72
683
49
23
-1
23
120
101
187
8
8
* 1
12
25
73
198
17
9
7
17
7
181
212
7
7
1
13
20
82
257
9
11
7
10
7
71
35
7
7
1
14
47
40
242
25
16
8
10
7
57
8
7
7
1
15
15
20
104
25
9
8
15
7
59
19
7
7
1
16
33
11
115
24
17
8
9
7
92
37
7
7
1
17
33
8
121
22
9
8
10
7
98
79
7
7
* 1
18
13
11
96
23
10
8
13
7
8
30
7
7
1
19
26
20
148
21
7
8
, 8
7
7
7
7
7
1
20
45
18
113
15
7
7
8
7
7
7
7
7
"* 1
21
52
24
187
42
10
7
8
7
7
7
7
1
22
65
23
258
22
25
9
17
8
8
8
8
8
1
23
8
-1
150
-1
39
8
9
8
8
-1
8
8
8
24
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
(-1
indicates
mi S8 ing
data)
-------
d i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i
S:
5
5
5
5
5
5
5
4
4
3
3
3
3
3
3
4
5
6
6
6
6
6
5
5
TABLE 8A
HOORLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
12/31/76
HIGHEST TWENTY-FOUR HOUR PREDICTED - RAM
Wind Direction
(Degrees )
Wind Speed
(M/S)
Teaperature
(Degree K)
Mixing Height
(Meters)
316
5.20
254.26
245.00
316
4.40
254.26
245.00
317
4 . 30
253. 70
245.00
312
3.84
254.26
245.00
303
3.66
254 .82
245.00
298
3.60
254.26
245.00
299
3.47
254.26
245.00
299
3.55
254.26
271. 78
302
3.96
254.26
317.82
306
3.02
255.37
363.86
284
2 .84
256.48
409.89
281
3.09
258. 15
455.93
284
3. 22
259. 26
501.96
300
3.23
260.37.
548.00
300
3. 20
260.93
548.00
303
3.24
260.37
548.00
304
3. 18
259.26
534.62
297
2.88
258.70
482.25
298
3.06
258. 70
429.87
299
3.17
258. 15
377 .50
295
2.71
257.59
325. 12
292
2.84
257.59
272.75
304
3. 15
257.59
220.37
305
3.60
257.59
168.00
-------
4330
4300
O
§4290
m
24280
< 4270
LEGEND
802 POINT SOURCES
<50 (xio3 kg/yr)
0 50-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 10)
25km
4230
4240
20km
4220
720
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
600
810
FOTURE8
Higheet Twnt y-Four-Hour Predicted - RAM
(A receptor nuaber encloaed in a rectangle denotes a aaxiaua
obaerved or predicted value.)
49
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
aj
6
6
6
7
7
7
6
5
4
3
2
2
2
2
2
TABLE 9A
HOURLY METEOROLOGY
COMPOSITE FROM 25 ST. LOUIS RAPS/RAMS STATIONS
11/15/76
HIGHEST TWENTY-FOUR HOUR PREDICTED - TEM
irection
Wind Speed
Temperature
Mixing Height
greei)
(M/S)
(Degree K)
(Meters)
115
.65
273. 15
96.00
87
.58
273.15
96.00
34
.86
273.15
96.00
16
1.32
272.59
.96.00
19
1.41
272.04
96.00
16
1.08
271.48
96.00
22
.85
271.48
117.31
20
.76
273.71
236.41
68
. 74
275.93
355.51
91
1.16
277.59
474.61
77
1.26
278.71
593.70
62
1.01
279.26
712.00
39
.89
279.82
831.90
115
.59
280.37
951.00
167
.63
280.37
951.00
83
.67
279.82
951.00
90
.82
278.15
954.56
76
1.22
277.04
802.59
104
1 . 36
275.93
684.82
108
1.48
274.82
567.06
114
1.45
274.26
449.29
148
1.76
274-26
331.53
163
1.68
273.71
213.76
165
1.36
273.15
96.00
-------
101
~~63
41
47
52
34
45
45
66
72
155
139
119
66
116
141
153
111
66
73
104
88
147
-1
-1
22
-1
7
7
7
7
7
7
7
30
69
60
51
33
34
31
27
28
21
13
7
8
8
-1
-1
TABLE 9B
SELECTED MODEL EVALUATION INPUT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/MJ) BY STATION
2 11/15/76
HIGHEST TWENTY-FOUR HOUR PREDICTED - TEM
Stat ion
103
104
105
106
108
113
114
115
116
120
20
314
10
52
10
33
7
J
7
26
26
386
-1
-1
13
29
7
-1
7
30
12
415
10
25
7
24
8
-1
7
28
8
323
14
42
7
14
12
-1
7
44
17
146
25
78
12
9
9
-1
7
18
22
140
31
40
10
24
17
-1
7
14
23
232
28
39
15
23
26
-1
7
7
30
404
28
75
16
43
23
-1
7
7
67
809
31
260
60
63
136
-1
7
24
29
454
127
226
58
127
144
-1
7
64
71
456
105
240
48
191
122
-1
7
122
44
376
51
203
27
296
120
-1
7
124
24
313
47
125
21
190
140
-1
7
117
45
237
43
116
15
205
60
-1
7
113
42
647
32
129
21
75
19
-1
7
-1
67
7 34
32
125
32
82
11
-1
7
88
51
401
36
179
45
92
37
-1
7
89
48
420
33
194
79
139
22
-1
9
90
46
622
32
135
62
64
44
-1
8
68
48
801
23
89
45
67
35
-1
18
56
26
580
47
86
36
62
38
-1
137
69
32
782
20
37
36
59
48
-1
231
86
-1
-1
8
-1
63
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
(-1 indicates aisaing data)
-------
4330
4300
%
2 4270
25km
O 4260
4250
4240
4220
N
A
LEGEND
802 POINT SOURCES
<50 (x.103 kg/yr)
050-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 101
20 km
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 9
Higheat Twmty-Pour-Hour Predicted - TEM
(A receptor number enclosed in a rectangle denote* a maximum
obaerved or predicted value.)
52
-------
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i
a
4
4
4
4
4
4
4
4
4
4
4
3
4
4
4
4
4
5
5
5
5
6
6
6
TABLE 10A
HOORLY METEOROLOGY
COMPOSITE FROM 25
ST. LOUIS RAPS/RAMS
STATIONS
1/26/76
SECOND-HIGHEST TWENTY-
-FOUR HOUR PREDICTED
- RAM & TEM
Wind Direction
Wind Speed
Temperature
Mixing Height
(Degrees)
(M/S)
(Degree K)
(Meters)
321
3.59
272 .04
810.27
322
3.70
272.04
805.71
321
3.33
271 .48
801.15
317
3.83
270.93
796.59
318
4. 37
270.37
792.03
319
4.74
269.82
787.47
313
3.88
269.26
782.91
316
3.74
268.71
778.35
313
3.83
268.71
773.80
311
4.21
268.71
769.24
312
3. 38
269.26
764.68
313
3.23
270.37
760.12
305
3.23
270.93
755.56
304
5.01
270.93
751.00
305
5.52
270.93
751.00
303
5.12
270.37
751.00
305
5.05
268.71
751.00
308
4.39
268.15
678.49
311
4.52
2<>7.04
591 .91
317
4.53
266.48
505.33
321
4.34
265.93
418.75
330
4.11
265.37
332.16
346
3. 18
264.82
245.58
350
2.89
264.26
159.00
-------
>di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
TABL3 10B
SELECTED MODSL EVALUATION INPDT DATA FOR RAPS DATA BASE
HOURLY MEASURED SO, CONCENTRATIONS (UG/M3) BY STATION
1/26/76
SECOND-
HIGHEST
TWENTY-FOUR
HOUR PREDICTED -
RAM
& TEM
Stat ion
104
105
106 108
113
114
115
116
45
12
7 37
7
32
160
38
96
7
7 13
7
36
204
63
76
7
7 17
7
21
239
85
48
7
7 10
7
8
46
43
34
7
7 10
7
10
80
22
50
7
7 7
7
7
66
43
61
19
7 7
7
7
86
46
35
45
8 7
7
29
38
34
59
43
7 7
7
54
31
22
32
7 7
7
96
41
43
13
36
7 103
7
-1
80
-1
11
44
7 105
7
26
159
-1
18
28
7 15
7
16
83
83
23
34
7 38
7
15
24
112
8
8
7 10
9
7
45
110
8
7
8 7
8
7
95
159
7
8
7 7
9
7
121
147
8
8
7 7
7
7
240
60
13
12
7 7
7
7
288
129
65
7
7 7
7
7
161
140
38
7
7 7
7
7
152
139
90
10
7 162
7
61
72
20
17
14
7 84
-1
-1
12
8
-1
-1
-1 -1
-1
-1
7
-1
101
27
26
23
15
18
21
17
12
-1
-1
-1
-1
-1
57
41
33
-1
-1
40
43
43
39
-1
-1
103
120
121
122
10
(-1 indicates visaing data)
-------
4330
«4280
2 4270
LEGEND
802 POINT SOURCES
<50 (x103 kg/yr)
0 50-250
~ >250
S02 MONITORS
X RECEPTORS (RAMS 10)
25km
4250
. 20 km
4220
710 720
730 740 750 760 770 780
UTM COORDINATES (EASTING )
790
800
810
FIGURE 10
Second-Higheit Twenty-Four-Hour Predicted - RAM & TEM
(A receptor number enclosed in a rectangle denote* a maximum
observed or predicted value.)
55
-------
REVIEW OF PRBAN DISPKR8I0N MODELS
J. C. Weil
Martin Marietta Corporation
Baltimore, MD
June 14, 1983
I. OVERVIEW
Six urban dispersion models for relatively inert pol-
lutants were reviewed for their scientific merits as
well as their operational performance. All models are
based on the Gaussian plume and apply to area and point,
stationary sources, the typical pollutants of concern
being sulfur dioxide and particulate matter. Two of
these models, RAM and TEM-8A, pertain to short-term con-'
centration averages (one, three, and twenty-four-hour
periods) and remaining four, AQDM, COM, ERTAQ, and TCM,
apply to annual average concentrations. The short-term
models use hourly inputs of source and meteorological
conditions and are generally run for a year's record of
data to determine the cumulative frequency distributions
of the one, three, and twenty-four-hour concentration
averages at a number of sites (receptors). The long-
term models calculate annual average concentrations
through the use of a joint frequency function, which
gives the relative frequency of occurrence of various
categories of meteorologica1 conditions - wind direc-
tion, wind speed, and stability. jhe latter models
contain jnly a vertical Gaussian distribution; the
crosswind (Gaussian) distribution is eliminated by aver-
aging it over finite wind direction sectors comparable
56
-------
to or larger Chan Che angular plume width. A nice
derivation of the formula used for the annual averages
is given by Calder (1971) (see Appendix D of CDM User's
Guide).
The Gaussian model components (plume rise formulas,
dispersion parameters, and stability classification
methods) are much the same in all models and are quite
similar to those in the rural models reviewed earlier;
typically they are 10 to 20 years out of date. Most of
the six models use the Pasqui11-Gifford (PG) dispersion
curves developed for surface releases in smooch open
countryside and all use the Turner (1964) stability
criteria. Only one model, RAM, uses dispersion curves
(Briggs urban; see Gifford, 1975) developed from tracer
releases in an urban area, St. Louis. Two models (CDM,
ERTAQ) account for the increased roughness or building
wake effects on dispersion by adopting an initial ver-
tical dispersion parameter
-------
all demonstrate the importance of convective scaling
parameters (*$, the CBL depth, and v*, the convec-
tive velocity acale) in ordering turbulence data; w*
is proportional to (Q0zj,)*'3, where Q0 is the surface
heat flux. The utility of these parameters in CBL dif-
fusion modeling has been shown through the laboratory
experiments of Willis and Deardorff (1978, 1981) and the
numerical simulations of Lamb (1979). Furthermore, Weil
and Brower (1982) have applied convective scaling con-
cepts to stability and dispersion estimates in an opera-
tional Gaussian model for tall stacks and shown that
such concepts lead to far better predictions of ground-
level SO 2 concentrations than does the EPA CRSTER model.
The latter model is based on the older PG dispersion
curves and the Turner stability criteria.
The point of the above discussion is that improved PBL
understanding within the past decade or so has indeed
made a demonstrable improvement in both the science and
performance of a simple Gaussian model for elevated
sources in relatively uncomplicated terrain. I believe
that such understanding can similarly be applied to
elevated and near surface releases in the urban environ-
ment . More strongly, I believe it must be done to im-
prove the abysmal "science" exemplified in the existing
models. Such an application needs to consider the two
major effects of the urban surface on the turbulence
structure and dispersion; the increased heat flux and
the larger roughness elements with the correspondingly
larger u*'s (friction velocity). These effects can be
incorporated within the aame framework applied to the
tall stack, rural terrain problem (Weil and Brower); a
discussion of the prospects and problems of this appli-
cation is given in Section 2.
58
-------
The performance of Che six urban dispersion models was
assessed with SO2 measurements from a 13-station moni-
toring network in St. Louis. For the annual average
models, the performance evaluation is relatively
straightforward and simple and consists of comparing
observed and predicted concentrations at each of the 13
stations. With one exception (station* 104), the models
perform fairly well with little difference between them
(see Pig. 1). I would be hard pressed to recommend one
model over another, based on performance. However, 1
should note that the correlation coefficient, r, from a
regression fit to plots of the observed vs. predicted
concentration (Fig. 1) is slightly smaller from TCM (r ¦
0.46) than for AQDM, COM, and BRTAQ (r - 0.64, 0.62,
0.67, respectively). If we delete the station 104
result in Fig. 1, we obtain r values for AQDM, CDM, and
ERTAQ of 0.95, 0.85, and 0.81, respectively; oddly, the
oldest model, AQDM developed in 1969, has the highest
correlation coefficient (although the differences in r
are probably not statistically significant).
The most disconcerting feature of the comparisons in
Figure 1, with station 104 deleted, are the low slopes
(b) of the regression equations, b £ C.05, and the pos-
itive intercepts, a. The departures of the a and b from
their ideal values (0 and I, respectively) are attrib-
uted to model formulation problems. Such departures
(and in the same direction a-s found here) appear to be
endemic to these models (see Calder, 1971).
For the short-term models, a significant difference
between TEM-8A and RAM occurs in the bias, i.e., the
average of the differences between observed and pre-
dicted concentrations. Baaed on comparisons paired in
59
-------
space and time, TEM-8A has a negative bias typically
equal to twice the observed average concentration (for
the one to twenty-four-hour averages); a negative bias
means that the model prediction is greater than the
observed concentration. This overall bias is dominated
by the especially high model overprediction for stable
conditions and is discussed further in Section 3. In
contrast, the RAM n>od»l overpredicts the one to twenty-
four-hour concentration values by or less on aver-
age. However, the standard 6evi *t i.on (SD) of the
residuals (observed-predicted concentration) is quite
large, being typically twice the average observed con-
centration (c0|>g); for TEM-8A, the SD is about five
times To put the SD of the residuals in proper
perspective, it is necessary to compare it to the ob-
served concentration fluctuation resulting from natural
variation in meteorological variables; i.e., fluctua-
tions in hourly-averaged concentrations for nominally
the same hourly-averaged wind direction, wind speed, and
stability condition.* Hanna's (1982) analysis of SO2
concentration fluctuations from the 1976 St. Louis data
show that the geometric standard deviation (GSD) of
coba/Q» where Q is the total emission rate, is 2.6.
Unfortunately, we do not have the GSD's of the residuals
*The degree to which the hourly-averaged conditions can
be held fixed is somewhat limited due to the finite
number of data points available. One common practice,
as followed by Hanna (1982), is to categorise the
concentrations by ranges ("bins") of wind direction*
wind speed, and stability; this then permits one to
accumulate a large number of data points in specific
bins.
60
-------
from the IRC model comparisons, but my guess (based on
the arithmetic SD'a) is that they are larger than 2.6
for RAM and certainly so for TEM-8A. This indicates the
potential and need for improvements in the model forau-
lat ion.
On the whole, 1 would have to judge the RAM model a
better performer than TEM-8A based largely on the bias
and SD results*, this is not to be construed as a strong
endorsement of the model, as its performance in many
ways is not good. For example, the correlation coeffi-
cient for the paired data (Table 5-12 of the TRC report)
is low and about the same for RAM as for TEM-8A (r
ranges from ~ 0.1 for one-hour averages to ~ 0.33 for
twenty-four averages). The reasons for the large dif-
ferences in bias between RAM and TEM-8A as well as some
other aspects of model performance are taken up in
Section 3.
61
-------
D18CD88IOI OF MODEL COMPOIKBTS
The specific model features discussed in Che following
are: plume rise formulas, plume penetration of elevated
stable layers, dispersion parameters (tfy and dt),
stability classification, area source treatment, and
meteorological inputs. Some of the remarks are repe-
titious of those made in an earlier review of rural
dispersion models (Weil, 1982a) but are restated for
completeness. Finally, some thoughts are given on
prospects for an updated urban dispersion model.
A. Plume Kiss
For elevated, buoyancy dominated sources, all of the
urban models use Briggs' (1971) model for computing
final rise in neutral conditions and apply the same
model to unstable conditions. However, this model
did not even consider convectively generated turbu-
lence, which is clearly most important for buoyant
plumes rising in the CBL. 1 think that Briggs*
(1975) models, "breakup" and "touchdown," are much
better choices for final rise because they incor-
porate more sound and up-to-date physics of the
neutral and convective PBL than the 1971 model.
Furthermore, they agree much better with the maximum
observed rise of power plant plumes as reported by
Weil (1979a).
There is no quarrel with the formula chosen for
final rise of buoyant plumes in stable conditions,
i.e., Briggs (1975); this formula is well-documented
by Briggs and others.
62
-------
Several of Che models (TCM, TEM-8A, RAM) compute
plume rice due to source toieatui flux when this
dominates the buoyancy effect. (The dominating
effect is simply determined by that formula, for
buoyancy or momentum, yielding the larger rise.) 1
believe this consideration is a necessity for all
models because some of the point sources in the
emission inventory indeed have a near ambient tem-
perature, i.e., little buoyancy. However, I object
to the formula chosen for neutral and convective
conditions, the one given by Briggs (1969); it is
simply outdated. Pinal rise formulas based on more
contemporary understanding of the neutral and con-
vective PBL are given by Briggs (1975) and should be
adopted. The formulas used for momentum dominated
rise in stable conditions (Briggs, 1969) appears
adequate and consistent with later formulas given by
him (Briggs, 197$, 19B0).
Actually, the state-of-the art formulas for final
rise for both buoyancy and momentum-dominated
sources are all summarised in Briggs (1980), How-
ever, one detail should be noted. The empirical
constant 4.3 in Eq. (66) of Briggs (1975) is pre-
ferred over the value 3.0 in Eq. (8.101) of Briggs
(1980) because the former was more consistent with
plume rise observations as reported by Weil (1982b);
the specific equation involved is that for final
riae of buoyant plumes due to the breakup model for
convective conditions.
The urban models only consider plume rise for point
sources; i.e., they neglect it for area sources with
the exception of an adhoc approach used in BAN. The
63
-------
neglect for area sources beara further scrutiny
becauae under very light winds, whan building vakaa
are much leaa intense, emissions from snail roof-top
sources aay indeed have a riaa coaparable to the
building height; the eaiasions are aostly from apace
heating and thue are hot. Clearly, auch rise would
aitigate the ground-level concent rationa, which ara
computed to be highest under the lightest winda (u <
2.5 ¦/a, aee table 5-4 of TEC report). In addition,
it should be noted that the TCH and TEM-8A aodala
asause the area source height to be aero. which
Magnified the concentrationa for the light wind
caaea even aore.
B. Pluae Penetration of Elevated Stable Layera
the urban aodala adopt an "all or none" criterion
for determining pluae penetration of elevated atable
layera. Either coaplete or aero penetration occura,
i.e., no partial penetration, with the criterion be-
ing h# > k*£, where ht ia the effective atack
height and k ¦ 1 for all aodala except for TEM-8A,
k 2, and TCH, which diaaiaaea the penetration
situation. If the pluae penetrate* the stable layer
(i.e., the ab< /e inequality ia aatisfied), no
ground-level concentrations are assuaed to occur; if
penetration doea not occur (the inequality ia not
aatiafied), the pluae reaaina trapped in the CEL
with reflection occurring at s * *£. The h* ia
baaed on the final riae foraula for neutral eondi*
tiona, which effectively asauaea that the atratifi-
cation below at exiata above a£ aa well, i.e.,
up to the point of final riae. Clearly, thia
aaauaption ia wrong. The aajor pitfall of thla
64
-------
criterion is that it ignores Che Gratification
change at a ¦ ££, i.e., the degree of stable
tret ificetion in the overlying air (z > t±). Any
realisti. penetration criteria must take into ac-
count the stable stratification of the elevated
layer since this is the major impedance to further
r ise.
In suaaary, I believe that the "all or none" cri-
terion is a poor approach and should not be used. A
¦uch preferred alternative is the penetration cri-
teria given by Briggs (1980); it does include the
stratification in the elevated stable layer. In
addition, it has been successfully tested against
field observations by Briggs (1930) and Weil (1980).
Furthermore, I note that Weil and Brower (1982) used
this criteria in their Gaussian model and found that
it predicted no occurrences of plume penetration of
the elevated stable layer, consistent with their
observed ground-level concentrations. In contrast,
the "all or none" criterion predicted several occur-
rences (26 out of 145 cases) of full penetration,
i.e., aero ground-level concent rat ions were pre-
dicted when in fact significantly high concentra-
tions were measured.
C. Dispersion Parameters
Three sets of dispersion curves (tfy, tfg versus x
and stability) are used in the urban models: the PG
curves in AQDH, CDM, ERTAQ, TCM, and TEM-8A, the
Gifford-Hanna curves in TCH and T8M-8A for low lev-
el, area sources only, and the Briggs urban curves
in RAH. As pointed out earlier, the PG curves were
*5
-------
devised from ground-level release* over flee, smooth
terrain and should not be applied to elevated point
source* especially under the most unstable condi-
tions, Class A. The reason is that plumes from
elevated sources (he > 0.1s£) disperse in a
rather homogeneous turbulence region within the CBL
and exhibit a vertical as well as horizontal growth
proportional to downwind distance, 6 z,
-------
tied to the earlier St. Louis diffusion experiment
(McElroy, 1969), the variation of 20 a is quite arbitrary (Calder, 1971).
It seeae to me that wind tunnel modeling could be
put to good use in determining the appropriate ini-
tial vertical as well as horizontal growth and the
subsequent rate of spread (beyond the immediate
building cavity). In summary, the urban model
authors for the most oart have misapplied and mis-
adapted the PG curves} I am not sure it is worth
pursuing these curves further.
The Gifford-Hanna (1971) curves as applied to low-
level area sources, are essentially the ASMS disper-
sion parameters (Smith, 1968), which were derived
from an elevated (~ 100 m) point source release.
Their applicability to low-level releases in cities
is not clear and to my knowledge has not been justi-
fied. Furthermore, no account is taken of initial
dispersion due to building wakes.
The Briggs' urban curves (see Gifford, 1975) were
developed from the McElroy-Pooler empirical curves
(McElroy, 1969) which were based on neutrally buoy-
ant tracer releases of a one-hour duration; the
releases were made near the ground (my guess is
within 20 a or so of the surface) in St. Louis.
These curves eapirieally account for the effects of
increased roughness and heat flux in the urban en-
vironment and are probably the aost legitimate
curves for near surface releases. However, I be-
lieve that their application to tall sources, as is
done in RAH, is incorrect and unjustified; the rea-
son is that for the aost unstable A-B conditions,
67
-------
3/2
the Briggs' curve gives « xJ' . However,
in the urb«n CBL, we should expect a near homogene-
ity of with z to exist for heights above ~0.1z£
jus*: as it does in the CBL over rural areas. This
character is expected because of the dominance of
convective over mechanical turbulence during very
unstable conditions; i.e., the boundary layer
structure is effectively independent of roughness
effects. For plumes dispersing in homogeneous tur-
bulence, we expect « x as discussed earlier.
For tall sources in the urban environment, I think
that Briggs' rural curves may be more appropriate;
these curves show <5y,
-------
stability class down by one towards the unstable
side. While qualitatively in the right direction,
the number of downward shifts is quite arbitrary; it
has not been backed up by theory or experiments.
Moreover, it seems difficult to properly adjust this
criteria to the urban environment when it is so far
off the mark for rural terrain.
For convective conditions, a much more theoretically
sound criteria for stability classification is the
ratio 2j/-L, where L is the Monin-Obukhov length.
This ratio measures the relative heights of im-
portance of convection and surface friction; typ-
ically, it exceeds 10 during daytime. Two useful
alternatives to zj/-L are w*/u* and w*/u, where
u is the mean wind speed in the CBL. The first is
related to zj/-L using the definitions of w* and
L , w*/ u * ¦ ( - z i/xL) 1/3, where k is the von Karman
constant. The second, w*/u, follows from the first
upon choosing a typical value for u*/u (~ 0.05).
Weil and Brower(19S2) adopted u/w* as the stabil-
ity parameter in their model and related it to
Briggs1 rural dispersion curves (see their report or
Weil, 1983 for details). In applying the model to
tall stacks, they found that it resulted in concen-
tration predictions in far better agreement with
observed SOj concentrations than did the EPA CRSTER
model; the latter employs the PG curves and the
Turner stability criteria.
The favorable results achieved with the Weil and
Brower framework give ua confidence that the convec-
tive scaling concepts are working properly and that
69
-------
they have a strong positive effect on model perform-
ance. The framework shows the explicit relationship
between dispersion curves and the key meteorological
variables, Qq and (in w*) and u; it would
probably be better to adopt u* instead of u since
u* incorporates roughness changes directly. We
should be able to extend this framework to the
urban setting and relate w*/u* to the appropri-
ate dispersion curves for both ground and elevated
releases. The key task is the proper parameteriza-
tion of the heat flux perturbation induced by the
urban area; for rural terrain Q0 can be assumed
proportional to the insolation. Once the heat flux
is parameterized, u* can be determined from a wind
speed measurement at one height using the similarity
wind profile and an appropriate zQ estimate.
E. Area Source Treatment
There are three points that come up regarding the
treatment of area sources in the urban models.
First, as was already discussed, the source height
should be included (TCM and TEM-8A) and plume rise
should be considered; at the minimum, the effect of
plume rise on the surface concentration in light
winds needs to be assessed.
Second, the degree of spatial resolution required
for integrating the area source distribution to
compute ground-level concentrations needs to be
resolved. The resolution ranges from the area
source grid size (typically 1 km by 1 km) in TCM and
TEM-8A (which use the Gifford-Hanna model) to small
fractions of the grid sire (CDM, RAM, ERTAQ). It
70
-------
seems Co me that a sensitivity analysis should be
done to see how important the spatial resolution is;
at issue is whether a lot of unnecessary computation
is being done.
Third, it should be made clear that the adequacy of
the narrow plume assumption used in this treatment
depends on the source grid size, downwind distance,
and Che 6y curve used (or stability class). Ef-
fectively whac this assumption means is thac one can
ignore "end effects" due to the finite size Cor side
length) of a source grid; the concentration downwind
of the grid center is the same as if the area source
were infinitely wide in Che croscwind direction.
The assumption remains valid for a receptor downwind
of the center of a grid square as long as cfy
< 0.3w, here w is the grid side length. For
the Brigga' urban curves and for w - 1 km, the nar-
row plume assumption breaks down for distances be-
yond: ~ 1.1 km (A-B curve), ~ 1.7 km (C curve), ~
2.5 km (D curve), and ~ 4.2 km (E-F curve).
F. Meteorological Inputs
The six urban models basically treat the meteorolog-
ical inputs in the same manner. For the short-term
models, there are two basic problems currently used
(same as discussed er.rlier in Weil, 1982a). The
first is the interpolation scheme for predicting the
mixed layer height zj. It is linearly interpo-
lated with time from twice-daily estimates of zj
(early morning and mid-afternoon), the estimates
being determined by the intersection of a surface-
71
-------
extrapolated adiabat with the early morning tempera-
ture profile. These estimates are then modified,
quite arbitrarily, using sj obtained from the
previous and succeeding days and the Turner stabil-
ity class. The use of such an overly simplified
approach seems unnecessary because mixed layer
height models based on an energy balance of the
lowest air layers, initial stratification, and the
integrated surface heat flux have been around for a
decade (Carson, 1973; Tennekes, 1973). I believe a
model of the latter sort should become pare of the
meteorological preprocessor package. (Recently,
Weil and Brower, 1983, adopted a slightly modified
form of Carson's model to be used for Z£ estimates
in diffusion model applications; they also included
schemes for other important CBL parameters (u, u*,
and Q0) all to be determined from routinely avail-
able data.)
The second basic problem with the short-term models
is in the treatment of light winds or cal ns, to
which the Gaussian model does not apply. For u < 1
mfa, the RAM model arbitrarily assumes u ¦ 1 m/s
whereas TEM-8A uses the input wind no matter how
small it is. I believe that both approaches are
wrong. Given the importance of light wind situa-
tions (both in convective and stable conditions), I
think that the Gaussian model should be extended to
include alongwind dispersion; alternatively, an-
other model could be formulated. Until an appropri-
ate model is developed, it is better to remove the
"calm" hou - from the modeling.
72
-------
Two additional and untouched problems that arise for
the short-term models are the parameterization of
the urban heat flux perturbation and the horizontal
inhomogeneity in it as well as Z£. Horizontal
inhomogeneity should be a problem more at night when
the urban perturbations are more significant rela-
tive to the rural background area.
For the annual average models, I wonder whether
seasonal variations in winds, stability, and zj_
have been examined to determine their effect on the
annual average concentration computed; all of the
urban models usa the annual average frequency func-
tion. Since the area source strength is highest in
winter, a strong difference in the frequency func-
tion between the winter and annual values could
significantly change the computed concentration.
C. Prospects for an Updated Urban Dispersion Model
Based on the above discussion, 1 feel that suffi-
cient technology exists to develop a much more
scientifically sound model for urban dispersion.
Such development would entail relating existing
dispersion cutves (McE1roy-Poo 1er or Briggs) for the
urban environment to w*/u*. This can be done by
coupling the statistical (Taylor, 1921) and similar-
ity theories (Yaglom, 1971) of diffusion to the
tfw, dv versus w*/u* relationship (see Weil
and Brower), Evidence exists that during strong
convection, 6W and over an urban area bear
essentially the same relationship to w* as over a
rural area (Ching et al., 1982). As stated earlier,
a key task is to parameterize the urban heat flux
73
-------
perturbation. Available data from the St. Louis
RAPS program (e.g., Godowitch et al., 1981; Ching et
al., 1982) as well as from other programs Oke (1982)
should make thii task feasible.
One important problem is the assessmenc of the
horizontal inhomogeneity in Q0, z£, roughness,
etc. on turbulence and diffusion. Here it seems
that there will be a need to come up with a simple
boundary layer height model, of the Carson flavor,
but with a spatially as well as temporally varying
heat flux. More complex numerical models as well as
observations will be invaluable.
Finally, as already noted, wind tunnels can be used
to assess the role of roughness on dispersion from
low-level sources. An informative experiment here
would consist of progressive increases in the number
of buildings (roughness elements) from 0 and 1 to a
larger and larger area of same until an asymptotic
dispersion curve is attained. Huber and Snyder
(1976) have done systematic experiments with and
without a single building; experiments with a large
number of buildings may also have been done and
should be checked.
I am optimistic that an urban model based on con-
temporary understanding of turbulence and diffusion
in the PBL will lead to predictions in much better
agreement with observations; this is based on our
experience with the tali stack, rural terrain prob-
lem .
74
-------
III. DISCUSSIOH OF MODEL PERFORMANCE
Further examination of the short-term modelo, TEM-8A and
RAM, was made to better understand their performance and
the large differences between the two. In looking at
Che highest 25 predicted and observed values (one to
twenty-four-hour averages) by monitoring station, one
observes the 3ame general results as seen using all data
paired in both time and space; typically, TEM-8A pre-
dicts high by a factor of 2 to 4 whereas RAM is rela-
tively close to the observed value on average but can
differ by factors of 2 at individual stations. In exam-
ining the highest one-hour averaged concentrations by
time of day, wind speed, and stability class, one finds
that: both models predict their highest values during
the night, in the lightest (u < 2 m/s) winds, and for
stable conditions (Class E and F). In contract, the
highest observations tend to be about the same through-
out the day. They do, however, show the same trend with
stability as predicted by the models: lowest for A-B
stability and highest for E-F stability.
The significant differences in the TEM-8A and RAM con-
centration predictions occur for all stability condi-
tions (Table 5-4 of TRC report), but are largest for D,
E, and F stability. The differences are attributed
primarily to three factors: 1) the smaller dispersion
parameters given by the PC and Gifford-Hanna curves used
in TEM-8A than by the Briggs' urban curves used in RAM
(differences are especially large for the D, E, and F
curves), 2) the mor-i severe penetration criteria used in
TEM-8A than in RAM (a plume stays trapped below
more often in TEM-8A), and 3) the use of the "as meas-
ured" wind speed in TEM-8A rather than the restricted
75
-------
speed, u > 1 m/s, in RAM (some very low winds, u ~ 0.5
m/s, result in high concentrations). The smaller dis-
persion parameters in TEM-8A lead to higher concentra-
tions for low-level area and point sources but lower
concentrations for very tall stacks. A scan of the
point source emission inventory reveals a predominance
of low to moderate point source emissions of SO2 from
relatively short stacks, typically 40 to 50 m high.
Thus, TEM-8A would tend to predict higher concentrations
for them. Quite a few of such sources are in the neigh-
borhood of monitoring stations 101-106, in the center of
St. Louis.
Examination of the highest and highest second-highest
one and three-hour averaged concentrations in Tables Cl-
C4 (TRC report) reveals that the observed values are
nearly split equally between day and night, whereas both
models predict the highest values at night as already
mentioned. The high daytime observations suggest that
elevated point sources are contributing. However, most
of the high observations (day or night) as well as the
highest twenty-four-hour averaged concentrations occur
in the winter. This occurrence would suggest the
importance of area sources (i.e., wintertime space
heating) or limiting meteorological conditions (e.g.,
smaller z^'s) for point sources in winter or perhaps
both. Further examination of this problem and with a
better model is needed to really sort out which sources,
area or point, are contributing most to the ground-level
concentrations.
As a final point, we note that RAM significantly under-
predicted concentrations at station 104 as did the an-
nual average aodels. High anomalies at this station
7ft
-------
were reported Co occur over about a one-aonth period
(Ruff, 1980). As discussed by Ruff, the anoaaliea Bay
have been due to an uncatalogued source. Thia aay also
have been cauaed by aoae upaet condition in an existing
source or to a special flow problea--building downwish.
A scan of the point source emission inventory shows that
several point sources are located within - 1 km of the
monitor (104). Typical stack heighta are AS m and two
of the sources have quite low buoyancy fluxes (F " 0.3
and 15 m VS3).
77
-------
REFERENCES
Briggs, G.A. 1969. Plume Rise, U.S. Atomic Energy Commission.
TID - 25075, 81 pp.
Brigga, G.A. 1971. Some Recent Analyses of Plume Riae Obser-
vations, pp. 1029-1032, Proceedinga of the Second Interna-
tional Clean Air Congreaa. Edited by H.M. Englund and
W.T. Berry, Academic Preaa, New York.
Brigga, G.A. 1975. Plume Riae Predictions, pp. 59-111, Lec-
tures on Air Pollution and Environmental Impact Analyses.
American Meteorological Society, Boaton, MA.
Brigga, G.A. 1980. Plume Riae and Buoyancy Effects, Chapter
8, Atmospheric Science and Power Production. U.S. De-
partment of Energy, in press.
Calder, K.L. 1971. A C1imato1ogica 1 Model for Multiple Source
Urban Air Pollution. Paper preaented at rirst Meeting of
the NATO/CCMS Panel on Modeling.
Carson, D.J. 1973. The Development of a Dry Inversion-Capped
Convectively Unstable Boundary Layer, Q.J. Roy. Meteorol.
Soc¦, Vol. 99, pp. 450-467.
Ching, J.K.S., J.F. Clarke, J.S. Irwin, and J.M. Godowitch.
1982. Review of EPA Mixed Layer Diffusion Programs and
Assessment of Future Needs, Proceedings of the Workshop on
the Parameterization of Mixed Layer Diffusion, pp. 20-23
October, 1981, Les Crucea, N.M.
Clark, J.F., J.K.S. Ching, and J.M. Godowitch. 1981. Spectral
Characteristics of Surface Boundary Layer Turbulence in an
Urban Area., pp. 167-178, Proceedings of the 5th Symposium
of Turbulence, Diffusion, and Air Pollution. American
Meteorological Society, Boston, MA.
Deardorff, J.W. 1972. Numerical Investigation of Neutral and
Unstable Planetary Boundary Layers, J. Atmos. Sci. Vol.
29, pp. 91-115.
Deardorff, J.W. and G.E. Willis. 1975. A Parameterization of
Diffusion Into the Mixed Layer, J. Appl. Meteorol., Vol.
14, pp. 1451-1458.
Gifford, F.A. 1975. Atmospheric Diapereion Models for Envi-
ronmental Pollution Applications, pp. 35-58, Lectures on
Air Pollution and Environmental Impact Analyses. American
Meteorological Society, Boston, MA.
78
-------
REFERENCES
(Cont.}
Gifford, F . A. and S.R. H«nni. 197 1. Urban Air Pollution Mod-
eling., pp. 1146-1151, Proceeding# of the Second Interna-
tional Clean Air Congreti. Edited by H.M. Englund and
W.T. Berry, Academic Press, Hew York.
Godowitch, J.M., J.K.S. Ching, and J.F. Clarke. 1981. Urban/
Rural and Temporal Variationa, PBL Turbulence Parameters
and Length Scale* Over St. Louie, MO., pp. 171-172, Pro-
ceedings of the 5th Sympos ium of Turbulence Diffusion, and
Air Pollution. American Meteorological Society, Boiton,
MA .
Hanna, S.R. 1982. Natural Variability of Obaerved Hourly SO2
and CO Concent rationa, St. Louia, Atmoa. Env iron., Vol.
16, pp. 1435-1440.
Hubtr, A.H., and W.H. Snyder. 1976. Building Wake Effect# on
Short-Stack Effluents, pp. 235-242, Proceedings 3rd
Symposium on Atmospheric Turbulence, Diffusion and Air
Quality. American Meteorological Society, Boston, MA.
Kaimal, J.S., J.C. Wyng»-;a, D. A. Haugen, O.R. Cote, Y, Izuni,
J.J. Caughey, and C.J. Readinga. 1976. Turbulence Struc-
ture in the Convective Boundary Layer, J, Atmos. Sci.
Vol. 33, pp. 2152-2169.
Lamb, R.G. 1979. The Effects of Releaae Height on Material
Dispersion in the Convective Planetary Boundary Layer,
pp. 27-33, Proceedings of 4th Symposium on Turbulence,
Diffusion and Air Pollution. Reno, NV. American
Meteorological Society, Boston, MA.
McElroy, J.L. 1969. A Comparative Study of Urban and Rural
Dispersion, J. Appl. Meteorol. Vol. 8, pp. 9-31.
Oke, T. 1982 The Energetic Basis of the Urban Heat Island,
Q.J. Roy. Meteorol. Soc. Vol. 108 pp. 1-24.
Ruff, R.E. 1980. Evaluation of the RAM Using the RAPS Data
Base, Final rept. Cont. No. 68-02-2770, Atmos. Sci.
Center., SRl Int., Menlo Park, CA. Prepared for ORD, U.S.
EPA, Research Triangle Park, NC, 83 pp.
Smith, M.E. ed. 1968. Recommended Guide for the Prediction of
the Dispersion of Airborne Effluents, The American Society
of Mechanical Engineers, Hew York, N.Y., 8b pp.
faylor, G.I. 1921. Diffusion by Continuous Movements. Proc.
London Math. Soc. Ser. 2. Vol. 20, p. 196.
79
-------
1BVKKKRCKS
(Coot.)
Tennekes, K. 1973. A Model for Che Dynamics of Che Inversion
Above a Connective Boundary Layer, J. Ataos. Sc i. , Vol.
30, pp. 558-567.
Turner, D.B. 1964. A Diffusion Model for an Urban Area, J.
Appl . Meteorol., Vol. 3, pp. 83-91.
Weil, J.C. 1979a. Assessment of Plume Rise and Dispersion
Models Using Lidar Data. Prepared by Environmental
Center, Martin Marietta Corporation, for Maryland Depart-
ment of Natural Resource*. Ref. No. PPSP-MP-24.
Weil, J.C. 1979b. Applicability of Stability Classification
Schemes and Associated Parameters to Dispersion of Tall
Stack Plumes, Maryland. Atmos. Environ., Vol. 13, pp. 819-
831 .
leil, J.C. 1980. Performance of Simple Models for Stack Plume
Dispersion During Convective Conditions. Prepared by
Environmental Center, Martin Marietta Corporation, for
Maryland Department of Natural Resources. Ref. No. PPSP-
MP-30.
Weil, J.C. 1982a. Review of Point Source Dispersion Models
for the American Meteorological Society. Prepared for the
Steering Committee of the AMS/EPA Cooperative Agreement on
Air Quality Modeling.
Weil, J.C. 1982b. Source Buoyancy Effects in Boundary Layer
Diffusion, pp. 235-246, Proceedings of the Workshop on the
Parameterization of Mixed Layer Diffusion, 20-23 October
1981, Las Cruces, NM.
Weil, J.C. 1983. Application of Advances in Planetary Boundary
Layer Understanding to Diffusion Modeling, pp. 42-56,
Proceedings of 6th Symposium on Turbulence and Diffusion,
American Meteorological Society, Boston, MA.
Weil, J.C. and R.P. Brower. 1982. The Maryland PPSP Disper-
sion Model for Tall Stacks. Prepared by Martin Marietta
Environmental Center for Maryland Department of Natural
Resources, Power Plant Siting Program. Ref. No. PPSP-MP-
36.
80
-------
(Coat.)
Weill J.C., and R.P. Brover. 1983. Estimsting Convectiv*
Boundary Layer Parameters Cor Diffusion Applications.
Prepared by Martin Marietta Environmental Center for
Maryland Department of Natural Resources, Power Plant
Siting Program. *ef. Ho, PPSP-MP-48.
Willis, G.E., and J.W. Deardorff. 1974. A Laboratory Model of
the Unstable Planetary Boundary Layer, J. Atmos. Sci.
Vol. 31, pp. 1297-1307.
Willis, G.E., and J.W. Deardorff. 1978. A Laboratory Study of
Dispersion from an Elevated Source Within a Modeled Con-
vective Planetsry Boundsry Layer, Atmos. Environ. Vol. 12,
pp. 1305-1312.
Willis, G.G. and J.W. Deardorff. 1981. A Laboratory Study of
Diapersion from a Source in the Middle of the Convectively
Mixed Layer, Atwos. Environ., Vol. 15, pp. 109-117.
Yagloa, A.M. 1971. Turbulent Diffusion in the Surface Layer
of the Atmosphere, lay. Acad_. Sci., USSR, Atmos. Oceanic
Phys. Vol. 8, pp. 333-340.
81
-------
AH OKBAH MODBLKEVIEW
J. C. Wyngaard
National Center for Atmoapheric Research
Boulder, CO
July 5, 1983
I. IHTRODUCTIOH
This review of urban models concentrates first, and most
heavily, on their scientific foundations, and assesses
the strength of these foundations in the light of our
current understanding of the planetary boundary layer
(PBL). It then reviews the performance statistics from
the test runs made with the 1976 St. Louis data base.
This is a comfortable and convenient perspective for me,
but is also, I feel, an appiopriate perspective. More
than 20 years have passed since the basic element of
these six urban models, the Gaussian plume parameteriza-
tion, was published. I believe that history will record
that as a milestone. I believe that history will also
record the sweeping advances that occurred in boundary-
layer meteorology in this same 20-plus years. I feel,
however, that history will note further that few of
these advances had been used in urban dispersion model-
ing by the time of this review.
My approach here is influenced by the hope that this
review process can serve a larger purpose than simply
assessing six urban diffusion models. I hope it can
also restimulate the technology transfer between PBL
research and diffusion modeling that seems to have
faltered in recent years.
82
-------
mm mo m icumiyic fodmpatioms or the mod»l»
A. fcaacription of tht Prban Modala
1 find that tht dlfferencea batwaan members of Chit
alx-aodal sat art aaall, avan though chair public*-
cion data* ranga froa 1969 (AQDH) Co 1980 (BRTAQ and
TCM). I do noC ainiaisa tha improvements in Cha
later models, but I find thaaa difference! to be
aaall coaparad Co the difference between any of then
and a model which would truly reflect our current
understanding. In view of thia, and in order Co
aiaplify thia discuasion, 1 will not refer in thia
aection to the deCaila of apecific models, buC
inatead to the characteristics of the "generic"
model which reflects the attributes of the group aa
a whole. In so doing, I will ignore the detailed
differences between individual models and this
generic mode I.
The model is based on Che Gaussian-p1ume parameter-
ization now associated wich Fasquill and Gifford.
The underlying concept is Chat the (ensemble aver-
age) plume downwind of a continuous point source has
a Gauaaian-1ike shape. Furthermore, turbulent dif-
fuaion is mathematically linear in the concentra-
Cion, and (in the simplest cases) mass-conserving.
These faces led Co Che simple, aleganC Paaquill-
Gifford representation of downwind concentraCion in
terma of a minimum of parameters: mean wind apaed,
pollutant source strength, and tha characteriatie
diaenaiona (laterally and vertically) of the plume,
Che ao-called "sigmas."
83
-------
Although chia formulation wn an irapretaive engi-
neering achievement, it relied on a data baae which
probably aeemed more adequate 20 yeara ago than it
doe* today. The difficulty of determining statis-
tically reliable mean concent rationa in diffusion
experiments, which has become a matter of widespread
scientific discussion only in the past several
years, probably prevented any early, definitive
assessment of the Gaussian assumption, particularly
in the vertical. Only recently, in fact, have
researchers been able to document departures from
vertical Caussianity, and to explain them in terms
of the PBL physics (Willis and Deardorff, 1976;
Lamb, 1982).
The height of the pollutant source appears directly
in the Gaussian parameterizatio If the source has
buoyancy, the resulting initial plume rise is cal-
culated (with Briggs' technique) and added to the
actual source height, yielding the "effective source
height." The model assumes that the sigmas are
independent of the effective source height.
The growth of the sigmas with downwind distance is
specified through use of stability categories based
on the 1964 Turner classification. These categories
are based on surface wind speed, cloud cover, and
solar insolation data (cloud cover, time of year,
latitude, time of day) and are tailored to routine,
operational applications. These stability catego-
ries do not use directly any information on PBL
depth, even if that were to be available.
84
-------
The model ha* "perfect reflection" of the plume at
the surface; this is done through uae of an "image
source" placed at one effective source height below
the surface. The model assumes that this reflected
plume diffuses upward from the surface in the same
way that the upper portion of the plume diffuses
from the source .
The model recognizes the existence of a top to the
PBL or "mixed layer," and uses perfect reflection
downward from this top as well. It does this by
determining the downwind distance at which the con-
centration at mixed-layer top is appreciable (say
0.1 that at plume centerline) and assuming that the
plume achieves a flat vertical profile of concen-
tration at twice this distance; in between, it
interpolates.
The model has a simple mean wind profile, with
stability-dependent speed shear, but no shear in
direction. Entrainment- or baroclinity-induced
shear is not included. Mixing height is an input
variable, and there is a simple parameterization of
pollutant decay, to be used when appropriate.
Although the model has a short-term application
mode, the writeups explain that the predictions give
averaged rather than instantaneous behavior. I saw
no mention of the ensemble average; the references
were to a time average, and the implied averaging
time ranged from several minutes to about one hour.
85
-------
B. A ContMporiry View of PBL 8tructura and Phytic!
Because of Che experimental and numerical modeling
advances of the 1960s and 70s, we know much more
about PBL structure, physics, and measurements than
we did 20 years ago. I will briefly summarize some
of these advances here, borrowing from my recent
review paper (Wyngaard, 1983).
Pe-haps the most significant advance has been in our
knowledge of the effects of stability on the PBL.
At the same time, we have come to realize that the
neutral state is quite elusive in nature; very small
surface heating or cooling rates drive the PBL into
convective or stably stratified states.
Several large field programs, together with the
pioneering large-eddy simulations of Deardorff, have
given us a fairly complete picture of the entire
convective PBL. We now have similarity scales for
its turbulence structure. The PBL depth h, which in
practice is the base of the lowest inversion, sets
the length scale of the dominant eddies, and in
conjunction with the buoyancy parameter g/t and the
surface temperature flux determines the turbulent
velocity scale w*. Simple thermodynamic models
(Driedonks 1982a,b) are known to be capable of re-
markably good operational predictions of the evolu-
tion of h during the day. Effective, simple models
of the surface energy budget, which allow realistic
predictions of the surface temperature flux, now
also exist.
86
-------
Simple "mixed layer" scaling has proven to be
quite effective for many aspects of convective ?BL
structure, and is now part of the language of PBL
research. In addition, there is now evidence that
many scaled parameters do not depend continuously on
a stability index, say h/L. Instead, it seems that
for such parameters all unstable data can collapse
to a point upon proper scaling. A principal remain-
ing uncertainty concerns the effects of entrainment-
induced turbulence on mixed-layer structure.
The advent of the acoustic sounder in the early
1970s revealed dramatically the nature of the noc-
turnal PBL. This stimulated serious experimental,
theoretical, and numerical modeling study throughout
its depth, which the sounder showed could be quite
shallow at times. Although today our understanding
of the nocturnal PBL still lacks somewhat that of
the convective PBL, there are now dynamical models
for the evolution of its structure (Brost and
Wyngaard, 1978) and for the evolution of its depth
(Nieuwstadt and Tennekes, 1981).
A by-product of the intensive PBL research over the
past 20 years has been a growing appreciation of the
meaning of the scatter in measurements. Similarity
theories, simple models, and parameterizat ions are
inevitably deterministic; however, PBL data, no
matter how carefully gathered, always have a random
component. Ue learned early that this "scatter"
could be very large in the unstable surface layer.
During the 1968 Kansas experiments, for example, it
was found that surface-layer stress could have the
87
-------
"wrong sign" locally for a good part of an hour
(Haugen, Kaimal, and Bradley, 1971).
A formalism for interpreting scatter, or more pre-
cisely the difference between finite-1 ength time
averages and the ensemble averages upon which our
theories are based, appeared in the Lumley-Panofsky
(1964) monograph on atmospheric turbulence. Its
applications to PBL data came somewhat later (e.g.,
Wyngaard, 1973). It is now generally appreciated
that large scatter is to be expected in convective
PBL data, even for averaging periods of an hour or
more. PBL researchers today recognize the challenge
of not only gathering good-quality data, but also of
gathering enough data to produce statistically reli-
able ave rages,
C. A Contemporary View of Diffusion in the PBL
Perhaps the best single reference here is the
Nieuwstadt-Van Dop ( 1982 ) collection, where recent
advances in our understanding of both unstable and
stable PBLs are applied to dispersion problems. The
most dramatic advances could be those summarized
there by Lamb for dispersion in the convective PBL.
Departing from traditional practice, he uses mi<:
-------
over Che past decade that K is not well-behaved in
the convective PBL. Lamb and Durran (1978) demon-
strated this by deducing K from calculations of
continuous point source diffusion in the convective
PBL of Deardor f f1s large-eddy simulations. They
found K to depend strongly on the source height, for
the same turbulent velocity field. Wyngaard and
Brost (1983) argue that this "geometry-dependent"
scalar diffusivity is an inherent property of the
convective PBL, and trace its origin to the non-
uniform distribution of buoyant production of
turbulent kinetic energy with height. They show,
again through large-eddy simulations, that the
diffusivity of a scalar introduced as an area source
near the top of a convective PBL is much different
from its diffusivity when introduced near the sur-
face. This complicated and essentially nonlinear
behavior of K would seem to bring into question much
of the past work with K-closures.
Since the early 1970s another mathematical approach
to PBL diffusion, "second-order" or "higher-order"
closure, has attracted much interest and also much
controversy. This approach is based on the mean
pollutant concentration equation plus a set of
equations for second moments, including one for the
pollutant flux; this set is closed through a variety
of approximate techniques, most of which have been
discussed by Wyngaard (1982). The controversy con-
cerns the underlying rigor of the approach, which
some find lacking. For example, consider the fol-
lowing quote from Liepmann (1979), who is writing
about second-order modeling applied to turbulent
flows in general:
89
-------
"Turbulent modeling is stil) on the rise
owing to rapid development of computers
coupled with the industrial need for
management of turbulent flows. I am
convinced that much of this huge effort
will be of passing interest only. Except
for rare critical appraisals such as the
1968 Stanford contest for computation of
turbulent boundary layers, much of this
work is never subjected to any kind of
critical or comparative judgment. The
only encouraging prospect is that current
progress in understanding turbulence will
restrict the freedom of such modeling and
guide these efforts toward a more reliable
discipline."
Those are discouraging words, but one should not
conclude that the prospects for all modeling
approaches are so bleak. One can find in the
Nieuwstadt-Van Dop volume other approaches which
hold considerable promise for numerical modeling of
PBL diffusion.
The importance of scatter in finite-1 ength time
averages of PBL concentration fields is now widely
recognized. Venkatram (1979a,b) has extended the
time-ensemb1e average variance arguments to the
diffusion problem, and hats emphasized the importance
of concentration fluctuations in this context.
Hanna's (1982) study of the natural variability of
averaged pollutant concentration in St. Louis dra-
matically illustrates the importance of this issue.
He finds a factor-of-two natural variability in
hourly concentrations of SO2 and CO.
Direct evidence of the scatter inherent in time-
averaged, continuous point source dispersion comes
90
-------
from Che laboratory measurements of Willis and
Deardorff (1976). They simulated seven realisations
of dispersion from a near-surface source, and aver-
aged the downstream concentration fields for times
equivalent to 20-30 minutes in the atmosphere.
Their realizations had great scatter; the average of
the seven still had scatter on the order of 50% of
the mean, even though the equivalent averaging time
in the atmosphere was about three hours.
D. A Contemporary View of the Urban Models
The last three sections make it clear that the foun-
dations of the urban models under review are not
consistent with today's knowledge. It is now also
clear, however, that there are strict limits on how
well any model can do in predicting time-averaged
pollutant concentrations. Hanna's (1982) St. Louis
RAPS study reminds us that even a perfect model
could do no better than the factor-of-two variabil-
ity he found in hourly-averaged concentrations.
Thus the urban model builders were mistaken when
they wrote that their models represent conditions
averaged over several minutes to an hour; we know
now that such models can aspire to represent condi-
tions only over far larger averaging times.
Given today's large computers, and the advances in
dynamic mesoscale modeling and large-eddy simula-
tion, it would be possible to build a fine-mesh
model capable of realistically simulating urban
meteorology, including the important details of the
91
-------
urban boundary layer. In this way one could simu-
late the important, diurnally varying urban circula-
tions which are ignored by the current generation
of models. Such a model could then be used to do
"brute force" simulations of urban air pollution
dispersion. That might be a straightforward ap-
plication of existing technology, but would also
be expensive. In view of that, and the natural
variability problem, it might well be judged not
feasible. Thus I feel we should not dismiss com-
pletely the simple approach underlying these urban
models, but instead consider it as a base on which
to build. Lamb (1982) also takes this point of
view.
The basic element of the urban models, the Pasquill-
Gifford Gaussian plume parameterization, remains a
good approach, in my opinion. It is simple, ef-
ficient, but should be modernized. The assumed
Gaussian shape in the vertical needs to be revised,
as does the model of plume reflection, both at top
and bottom. The possibility that the effective
source height should enter the model in other ways
(for example, by influencing the aigmas) should be
carefully considered. Lamb's (1982) concept of a
virtual source height, which allows him to accommo-
date within the Gaussian model framework some of the
strange behavior of plumes in the convective PBL, is
a first step along this path. The stability catego-
ries should be redone to make them consistent with
today's knowledge of both stable and unstable PBLa.
The routines for the mean wind profile and the
mixed-layer depth, particularly at night, can also
be brought up-to-date.
92
-------
Some of this updating, of course, has already been
done; more modern models than the six we are consid-
ering now exist (e.g., Weil and Brower, 1982). This
is, in my opinion, as much an engineering challenge
as a scientific one, and therefore each candidate
model improvement must be carefully assessed in
cost/oenefit terms. An important aspect of this
challenge is obtaining the input data needed for
updated models. This need not require sophisticated
turbulence measurements; indirect techniques show
great potential (Weil and Brower, 1983).
One of the important developments in air-quality
research over the past decade is the documentation
of the natural variability (inherent uncertainty)
problem. One unfortunate side effect, however, is
the emerging suspicion that field measurements of
air pollution are not as valuable for verifying
models as we once assumed; or, put another way, it
now seems that we need much more data than we ever
suspected in order to do definitive model verifica-
tions. Consequently, we should use alternative
means of generating data bases - for example,
large-eddy simulation, which Deardorff/Lamb have so
profitably used; or laboratory modeling, which has
been so skillfully done by Wi11 is/Deardorff in the
convection chamber and Snyder/Hunt in the stably
stratified towing tank of the EPA Fluid Modeling
Facility. Such data bases could provide an ex-
cellent means of testing and refining dispersion
parameterizations. They would also allow us to
begin the serious study of inherent uncertainty,
which ultimately will have to be addressed by air-
quality models.
93
-------
IIr. AS8KSSINC THE PERFORMANCE STATISTICS
A. Factor* Influencing the Statistics
I have separated the factors which influenced the
performance statistics of the urban models into
three categories: model factors, data-base factors,
ind natural variability.
Model factors include the treatment of point and
area sources within the model, the treatment of
removal processes such as surface deposition and
chemical reactions, and the treatment of transport
and dispersion. All the models seem to use similar
techniques for handling the area sources; most of
them in fact use the Gifforrt-Hanna scheme. None
allows surface deposition. Several of the model
vriteups discuss a simple parameterization of decay
(as, for example, through react ion), but only two
models were run in this mode. Perhaps one can
neglect SO2 deposition and decay in urban applica-
tions, but I saw no discussions justifying it.
Finally, as I indicated in Section 2, all models use
basically the same transport/dispers ion physics.
As data bases go, this was a good one. The meteoro-
logical data set was unusual in that it had measure-
ments of PBL depth, although evidently these data
were not used in the annual-average model runs. One
suspects that this data set would be useful in as-
sessing a new generation of urban models using sta-
bility categories based in part on PBL depth. I
found no discussion of the fidelity of the emissions
or concentration data, or any discussion of the
94
-------
extent to which the emissions were correlated with
the meteorology. The latter consideration enters
with annual-average models.
Finally, the natural variability of the SO2 concen-
tration data is important here in view of Hanna'a
(1982) finding that their hourly averages have a
factor-of-two scatter. This scatter, which could be
due entirely to natural causes, severely constrains
the inferences thst one can make from the perform-
ance statistics.
B. Annual-Average Model Perforaance
The summary of the annual-average models was given
as Table 1 of the TRC statistics. I find the per-
formance of the models remarkably good; the average
of the measurements over the 13 stations was 42
micrograms/cubic meter, the model predictions rang-
ing from 37 to 51. On a station-by-station basis,
the agreement for all four models was generally well
within a factor of two. Some might see subtle
trends in Table 1 (e.g., the models might tend to
overpredict near the city center, and underpredict
in the outskirts). However, in view of the many
factors which influence these results, I conclude
simply that all four models performed well.
Some model guidebooks discuss a procedure for "cali-
brating" the model against a data set. It is not
clear to me that any of these four annual-average
models have been so calibrated. If not, their good
performance here is surprising, since there are many
areas where the models could now be improved. One
95
-------
tends to associate better models with better predic-
tions, but judging from these performance test re-
sults the annual-average predicions could not be
radically improved. Judging from Table 1, the weak-
est aspect of the models' performance was In their
inability to reproduce the large station-to-station
variability within the inner city.
C. Short-Term Model Perfor»ance
The performance tests of the short-term models
produced an order of magnitude more data than the
annual-average models, and perhaps relatively more
insight. My perusal of the results indicates that
RAM far outperforms TEM. While RAM seems to under-
predict somewhat during the day, TEM far overpre-
dicts at night. Thus TEM overpredicts the extremes
by about a factor of two, for one, three and twenty-
four-hour averaging times.
Table F-l of the TRC report summarizes quite dra-
matically the difference in performance between RAM
and TEM. This table, for one-hour averages, shows
that TEM overpredicted concentration at every sta-
tion, in most cases by more than a factor of two.
For the city as a whole, it overpredicted at all
times of day, although much more at night; it over-
predicted for all wind speeds and for all stability
categories.
The performance of RAM seems quite encouraging, in
view of the natural variability problem. Judging
from the one-hour statistics, the scatter in the RAM
predictions is not an order of magnitude larger than
96
-------
Che minimum implied by natural variability, as seems
to be the case with TEM. The scatter seems only
a factor of two or three larger than this natural
limit. This should be small enough to give support
to chose who use models, and yet lurge enough to
give incentive to those who hope to improve models.
97
-------
REFERENCES
Brost, R.A., and J.C. Wyngaard. 1978. A Model Study of the
Stably Stratified Planetary Boundary Layer, J. Atmos.
Sci., Vol. 35, pp. 1427-1440.
Driedonks, A.G.M.. 1982a. Sensitivity Analysis of the Equa-
tion c for a Convective Mixed Layer, Bound.-Layer Meteor.,
Vol. 22, pp. 475-489.
Lriedonks, A.G.M. 1982b. Models and Observations of the
Growth of the Atmospheric Boundary Layer, Bound.-layer
Me t eor . , Vol. 23, pp. 283-306.
Hanna, S.R.. 1982. Natural Variability of Observed Hourly S02
and CO Concentrations in St. Louis. Atmos. Env. , Vol. 16,
pp. 1435-1440.
Haugen, D.A., J.C. Kaimal, and E.F. Bradley. 1971. An Ex-
perimental Study of Reynolds Stress and Heat Flux in the
Atmospheric Surface Layer. Q. J. Roy. Meteorol. Soc.,
Vol. 97, pp. 168-180.
Lamb, R.G. 1982. Diffusion in the Convective Boundary Layer,
Atmospheric Turbulence and Air Pollution Modeling, F.T.M.
Nieuw8tadt and H. Van Dop, Eds. Reidel, Dordrecht.
Lamb, R.G., and D.R. Durran. 1978. Eddy Diffusivities Derived
from a Numerical Model of the Convective Planetary Bound-
ary Layer, 11 Nuovo Cimento, 1C, pp. 1-17.
Liepmann, H.L. 1979. The Rise and Fall of Ideas in Turbu-
lence, Am. Scientist, Vol. 67, pp. 221-228.
Luraley, J.H.L., and H.A. Panofsky. 1964. The Structure of
Atmospheric Turbulence. New York, Interscience.
Nieuwstadt, F.T.M., and H. Tenneken. 1981. A Rate Equation
for the Nocturnal Boundary-Layer Height, J. Atmos. Sci.,
Vol. 38, pp. 1418-1428.
Venkatram, A. 1979a. The Expected Deviation of Observed Con-
centrations from Predicted Ensemble Means. Atmos. Env.,
Vol. 13, pp. 1547-1550.
Venkatram, A. 1979b. A Note on the Measurement.and Modeling
of Pollutant Concentrations Associated with Point Sources.
Bound.-Layer Meteor., Vol. 17, pp. 523-536.
Weil, J.C., and R.P. Brower. 1982. The Maryland PPSP Disper-
sion Model for Tall Stacks. Report PPSP-MP-36, Martin
Marietta Corp., 1450 S. Rolling Road, Baltimore, Md.
21218.
98
-------
REFERENCES
(Cont.)
Weil, J.C., and R.P. Brower. 1983. Estimating Convective
Boundary Layer Parameters for Diffusion Applications,
Report PPSP-MP-48, Martin Marietta Corp., 1450 S. Rolling
Road, Baltimore, Md. 21218.
Willis, G.E. and J.W. Deardorff. 1976. A Laboratory Model of
Diffusion into the Convective Planetary Boundary Layer,
Q. J. Roy. Meteorol. Soc., Vol. 102, pp. 427-445.
Wyngaard, J.C. 1973. On Surface-Layer Turbulence, Workshop on
Micrometeorology, D.A. Haugen, Ed. AMS, Boston.
Wyngaard, J.C. 1982. Boundary-Layer Modeling, Atmospheric
Turbulence and Air Pollution Modeling, F.T.M. Nieuwstadt
and H. Van Dop, Eds. Reidel, Dordrecht.
Wyngaard, J.C. 1983. Lectures on the Planetary Boundary
Layer. To Appear in Mesoscale Meteorology. Theories,
Observations, and Models, T. Gal-Chen and D. Lilly, Eds.
Reide 1, Dordrecht.
Wyngaard, J.C., and R.A. Brost, 1983. Top-down and Bottom-Up
Diffusion in the Convective Boundary Layer. Submitted to
J. At mos. Sc i.
99-
------- |