-------
The residual statistics suggest that CTDM(11083-E) has the best overall
performance. The CTDM(11083-E) log-normal statistics (mg and sg) are
closest to the desired value of 1.0, and the bias (ma) and noise (sa) of
the set of paired concentrations are the lowest of all the models. However,
the summary statistics show that the performance of CTDM(11083-E) is not
much better than that of CTDM(11083) and COMPLEX I. CTDM(11083-E) and
CTDM(11083) underestimate peak concentrations while COMPLEX I overestimates
them. For the three models, Sg is greater than 3.0.
COMPLEX II and Valley simulate the observations much more poorly than
do the other models. Both models substantially overestimate the peak con-
centrations, and COMPLEX II does a poor job in reproducing the distribution
of the tracer gas concentrations observed on CCB.
The measure of model resolution identifies Valley as less responsive to
changes in meteorology than the other models, but the resolution statistics
of the other four models are not significantly different. One would expect
CTDM(li083), CTDM(11083-E), and COMPLEX I to have similar resolution
measures on the basis of the scatterplots of Figure 2 discussed below, but
COMPLEX II would be expected to be inferior. This dilemma occurs because
the measure of resolution reduces to unity for a model that produces a wide
range of concentration values compared to the range in observed concentra-
tions, as does COMPLEX II. Consequently, even though all the models except
Valley have roughly the same resolution capability, the three models with
lower noise measures perform better.
Scatterplots of peak observed concentrations scaled by the emission
rate (C0/Q) versus peak modeled concentrations scaled by the emission
rate (C /Q) are presented in Figure 2. Three of the models show quali-
tatively similar patterns, while two show patterns that are distinctly
different from the rest.
The Valley model is not designed to use onsite meteorological
measurements, but uses "worst-case" meteorology instead. Therefore, model
estimates of Cp/Q only depend on the distance from the source to the
nearest terrain feature at the elevation of the release. At CCB, this leads
to a relatively narrow band of Cp/Q values that is unlike the pattern of
the other models evaluated. We see from the figure that Valley overestimates
most CQ/Q values, but it underestimates the seven largest CQ/Q values.
This indicates that the standard "worst-case" meteorological conditions
contained in Valley for screening large power plant plumes are probably not
appropriate on the scale of the SHIS #1- at CCB.
COMPLEX II is the other model with a distinctly different scatterplot
pattern. This pattern is nearly the opposite of the Valley pattern. Valley
concentration estimates cover a range much narrower than the range of obser-
vations, while COMPLEX II estimates cover a range much greater than the
observations. In both cases, model estimates appear to be poorly correlated
with the observations.
20
-------
VALLEY
zra.e-r
C0/Q ,„.,
COMPLEX!
200.9
oe.aa ise.ee see.
COMPLEX I
286.8-1
sa.ee iee.e
CTDM(11083)
.08 288.08
asa.ea 4tra.ee ^se.ea sao.ea
CTDM(11083-E)
Figure 2. Scatterplots of observed and modeled peak concentrations
(C/Q, vs/m3) for 153 hours of SHIS #1 data.
21
-------
COMPLEX I, CTDM(11083), and CTDM(11083-E) display similar patterns of
scatter in that the range of estimated and observed peak hourly concentra-
tions are nearly the same, and the visual correlation between observations
and estimates is much better than that indicated by the Valley and
COMPLEX II patterns. Among these three models, COMPLEX I is biased toward
overestimation; CTDM(11083) and CTDM(11083-E) are somewhat biased toward
underestimation of C0/Q values of less than 100.
The geometric statistics for log-normally distributed residuals were
obtained from the cumulative frequency of occurrence plots shown in Fig-
ures 3 through 7. These plots indicate that peak residuals from none of the
models are strictly log-normal, but they also show that all of the distribu-
tions are nearly log-normal if the "tails" of the distribution are ignored.
These tails are generally made up of C0/Cp values in which either Co or C
is nearly zero.
Besides estimates of mg and sg, these frequency plots may also be
used to estimate the frequency witn which peak model estimates are within a
factor of 2 (or any other factor, for that matter) of the peak observed con-
centrations. 22% of Valley estimates, 41% of COMPLEX I estimates, 21% of
COMPLEX II estimates, 28% of CTDM(11083) estimates, and 44% of CTDM(11083-E)
estimates are within a factor of 2 of the peak observed concentrations.
These figures indicate that CTDM(11083-E) and COMPLEX I ought to have better
performance statistics than the other models. However, the cumulative fre-
quency distributions and the other residual statistics show that all the
models produce a substantial amount of noise in their simulations. There is
clearly room for improving the reliability of these models intended for use
in complex terrain settings. There is a discussion of suggested improve-
ments to CTDM and their basis in Section 3.
To help understand the noise in the model calculations, the residuals
based on the peak concentrations have been plotted against Cp and several
meteorological parameters. These plots show where a model might be doing
comparatively better or worse, thereby indicating areas for improvement.
Figure 8 shows Co/Cp vs. Cp/Q for CTDM(11083), CTDM(11083-E), COMPLEX
I, and COMPLEX II. A similar figure for Valley has not been presented
because Valley model concentration estimates fall into such a narrow range.
The figure shows that many of the underestimates from CTDM(11083) and
CTDM(11083-E) occur for very low values of Cp/Q, and that the tendency for
underestimation decreases with increasing Cp/Q. The degree of overesti-
mation also decreases with increasing Cp/Q. These observations indicate
that these models perform relatively better under those conditions that
produce the higher concentration estimates.
COMPLEX I and COMPLEX II do not display this trait. These models tend
to overestimate the observed peak concentrations to nearly the same degree
for all Cp/Q, although the number of underestimates decreases with in-
creasing Cp/Q.
Scatterplots of C0/C_ vs. wind speed are given in Figure 9 for
CTDM(11083), CTDM(11083-E3, COMPLEX I and COMPLEX II. There is considerable
22
-------
oa
o
o
(9
05
«D
O
Ul
IS
\
V
.^c
09
eg
T* H,
CO
in
OS
at
8
a
s
10
a>
o>
a>
in
a»
s
. OB
m
g-
cr
05
CD
£
3
O
o
•p
•H
t— l
S-
&
x
(D
PH
U
O
q-i
o
rH tfi
X O
0> M
LO
M-l oj
(U ni
> "d
•H
+-> i-l
Cti ^
CO
a.
O
23
-------
5.M
.82-
.01 •
7
V
2.88
• 01 .1 .512 5 18 28 30 43 SB 63 70 6C 08 05 98 89 00. S 00.0 00.80
Cumulative Frequency (%)
Figure 4. Cumulative frequency plot of C0/CD for COMPLEX I applied to
SHIS #1 data (153 hours).
24
-------
1.80-
7
i
L
7
7
.01 .1 .512 5 18 28 38 48 58 68 78 88 OB SS 86 88 00.5 08.8 B8.08
Cumulative Frequency (%)
Figure 5. Cumulative frequency plot of Co/Cp for COMPLEX II applied
to SHIS #1 data (153 hours).
25
-------
268.09
Cn/CD
o p
gg-
.18-
.es-
/
/
i
1
/
/
ft
l>
/
1
I
/
/
/
/
/
1 1
/
7
r
I :
/
/
/
/
/
/
/
/
f
/
/
'
1
f
- see.aa
- sa.ee
p
2.ee
1.08
.50
.28
Figure 6.
1 2 S IB 20 30 48 50 68 78 SB 0885 08 00 00.5 08.0 00.00
Cumulative Frequency (%)
Cumulative frequency plot of C0/Cp for CTDM (11083) applied
to SHIS #1 data (153 hours).
26
-------
I
t
• 81 .1 .512 5 IB 2B 30 « S3 68 7B 88 OB 85 98 BS 00.S 90.0 04.00
Cumulative Frequency (%)
Figure 7. Cumulative frequency plot of Co/Cp for CTDM (11083-E)
applied to SHIS #1 data (153 hours).
27
-------
uT
CO
§
T™
i
Q
\5
*»* ^
*
j_i^
CO
§
O
t
*
f 4
* *
* +
* ** * *
* * *• * * *
*, " • <£*!****« **.***. **
*• *v - '
ott
X
o
o
-
•
* . '
» , *
^ *
•« '•,***
\i *£#$**'*• *•* 7** *
• • • Ul m
n N -
*j
X
o
o
I
s
I
z
s
I
"
* (T
S p
3 S,
15 a
s XQ
1 °
s
s
s
'i
$
"a
3
S
g
'$
s
'S
s
'i
3 — .
i «E
1 S-
5
1 °
§
5
i
•§
g
'a
*
i
Q-
2
O
o
•• + t
B
D
^
X
LU
COMPL
it
>
* -
* .
+
*
+
*
*
* ^
* * +
+
•*• +
* * *
* + 4-
t ** *
+ * *
* *+ + +
*++ *+ **+*+•*• + *
V. *.***** ^ :\ *
+ * * * *•'!**** + + * ^> *
**i* *+*• ,.
a a 01 in m —
oa
X
o
o
t
*
* * . *
+ i-
»»*' *»>* *(.» ,.
. »*' *.*»**
t*V^* ::*"**** * * '
* »* >?>*; „ * §• -* » ** * * * t
Q.
O
o
O
3
i
1
i
s
1
S
S CY^
•| E
X
tn
s 3
% O ^
a. a)
S ° ft
s
*=
'a co
ffi
• UD
S
o
s
a tn
o
•H
« ^J
. • cd
(D
0)
O
e
i
0
•p
1 1
s rci
0)
« fH
"a a)
.« "o
MH
o
s
i pj
0
. -H
* -^T" j_l
- « ^3 ^"^
S E rt
• i 'S
^ a >
g Q
"i
OO
S (U
-I g
1 -H
r~T-t
s
"s
s
'
t/)
PI
o
•H
nj
ij
PI
O
O
0
-------
LU
* CO
* * 00
. ' O
+ + *— *
5
, * t a
* * 0
* *
* *
* ***
++
** ***
•
•*• + *
* *
* * % * *
+ *
* * *
* + + * *
* * * *******„*
* ***
** * * * ** + * *
***+ * * * *
** * ******
.****'
-t-
* ***** * *
*
« «••»". -.
*
++++
* * + *
++++ * +
** +. / * +
. + +
* * +* +•
, * * .*
* * *****
S S S ? " "
S
—
s
.1
.8 §
" S
•H
.8 — «
E S
.8 "= ^H
W 4^: •
/ — \
g CO ^3
' ^ ^
.8
' ' .. t
CO
• 00
o
,_
I
* , ^~
'* •*• *
++ * -f
+
h + t +
+ *+ *
•*• -fr
' + ' +
* + * ' *
* + *• +
+
+ *
* + +- *
* i *
* ** »*
* * * ,* ** *
****** *
4-fc "*" +
* ** * *****
. * *» » *
*** * *
+ * +
+
•. k k k ". °\ '-.
a.
. CJ
-
S
a
s
ei
s
°
§
g
S In
£
-? 3
u
.i
8
m
8
$
8
'
CD S
i ^ d%
/ ' '" ; "' •" "i
a.
* * * v 1
» CJ
+ * +
* * * * *
» »* * * * *
****** * * *
+
+ +
+ * +•
* *
* * **
» * * *
* ** *
*****
+ + +• 4-4-
* ** *
**+ + *
***** * * *
*******
** * **** * **
* *-
** *
+ * * *
i i i i 1 1 • —
a at m ui m — ,
a.
O
= fn 0
CD S
2 CO O
• rQ O
O
fn
g 4-4 CD
d O O
Oj
S OP
• -H
P X
s Cti i— 1
* -H f-t
^ o
J « ^" ^
§
S3 .
01
£ CD
bO
S .H
I
i
^o >
" • o
29
-------
scatter in all of the plots, but some trend can be seen in the patterns.
CTDM(11083) exhibits a distinct bias toward underestimating observed peak
concentrations for wind speeds in excess of about 5 m/sec. Because Hc is
probably small (or zero) compared to the source heights when source-height
wind speeds are as great as 5 m/sec, this suggests that the Lift component
of CTDM(11083) is underestimating the amount of plume material on the
surface under the more nearly "neutral" flow conditions. When crz is
enhanced as in CTDM(11083-E), the figure shows that much of this bias at
larger wind speeds is reduced, although it is not eliminated. Therefore,
the Lift component may need revisions to account for the combined effects of
flow distortion and plume dispersion more realistically.
COMPLEX I exhibits a bias towards overestimating peak observed
concentrations at the lower wind speeds. COMPLEX II appears to exhibit the
same behavior, except that a few large underestimates also occur at light
wind speeds. The overestimates for light winds may be the result of using
PGT CTy values in COMPLEX II, and 22.5° sector-averaging in COMPLEX I
(e.g., see Figure 12 in the CTMD Second Milestone Report). At very low wind
speeds, the wind direction often underwent large variations at CCB. The
22.5° sector within COMPLEX I may underestimate the plume meander in these
conditions, and thereby consistently overestimate concentrations on the
hill. COMPLEX II would also certainly underestimate the meander, but its
narrow Gaussian plume might also nearly miss the hill at times, thereby
producing both the underestimates and the overestimates indicated in
Figure 9.
Scatterplots of CQ/Cp against other modeling parameters also
reflect the patterns just described. For example, u/N (Figure 10), the
ratio of the mean wind speed to the Brunt-Vaisala frequency, distinguishes
between the more "stable" and the more "neutral" hours, and the patterns of
model performance are similar to those discussed above for the plots against
wind speed. l-Hc/zr (Figure 11), where zr is the plume release height,
orders model performance in the near-neutral limit when 1-Hc/z- is greater
than approximately 0.5, and in the very stable limit when l-Hc/zr is less
than zero. Figure 11 indicates that CTDM(11083) and CTDM(11083-E) are most
prone to overestimate peak concentrations when Hc exceeds 0.5 zr, but is
less than 1.5 zr, and CTDM(11083) generally produces underestimates for
HC less than 0.5 zr. The bias toward overestimating peak concentrations
with COMPLEX I increases as HC increases.
Figure 12 contains scatterplots of Co/Cp against l-Hm/zr, where
Hm is the elevation of the peak observed concentration. This plot shows
that CTDM(11083) tends to underestimate peak concentrations when they occur
appreciably above the release height—yet another indicator of more nearly
"neutral" flow. Enhancing az in CTDM(11083-E) tends to lessen this
trend. The corresponding plots for COMPLEX I and II (not shown here) show
no discernible pattern.
Figure 13 contains a plot of Co/Cp vs. the product of the cross-
wind vertical and horizontal turbulence intensities for CTDM(11083) and
CTDM(11083-E). (The other models do not use these data). Large turbulence
intensity products indicate a relatively large dilution of plume material-
The figure indicates that modeled and observed peak concentrations most
30
-------
Ill
ci
00
o
5
Q
0
*
+
*
*
•f
* + +
+
+
++• ^
+• **• *
**,* ** *
* *
•*• + •*" + 4- +
****** **
* * * *** V, /* *
* *+* *„*..*.* *:****,* * *
f+* H-* ***
/ '* { * *
i — i ( 1 — i , . —
X
'i
1
s
'1
s
•s §
S
•H
1 g
E S
•; z i-:
D
en
°. t — i
C/D
• '4-1
1 0
.J J •
cd 3
•S *
2 rj
TTJ 4-J
H
0 0 0 4J
•P cd
i j_)
"Td -P
•t & fi
^ • >
+* * *
* . f
+
* * *
. ++ **^
* + * +
. + , *
* + + •«• +,
. ++ +
* * * *
,** * *
+ +*+ ++*•«- »
* +*** „*+**
* * »,*** * •**, * * ***
t * /»,**. ** *
. „ ^
O
i
g
1
X
i
t
I
i
i
*• I
1 x
3
I
S
i
I
g
s
a
i
X
a!
I
*
»
*
»
»
» ** *
** *
+ f
* + + ^
* *• + +. *
v * %r **" ** **
* * » »* » * ,.*
*+ *++* + **+ + **+
**** > *"**** *,*,** * * * *
oa
'I FH 0
-------
CO
n
ffi
co
• H ffi
+J I
tD -H
o
O
O
O C
S O
I -H
O 4J
•s,
.8
O
.S
-P
O
O -P
(D
32
-------
3.8
2.8
CTDM(11083)
* *
3.8-
2.8-
C0/CP
CTDM(11083-E)
Figure 12. Variation of observed-to-modeled ratios of SHIS #1
maximum hourly tracer concentrations with l-Hm/zr.
(Hm is the height of the observed concentration.)
33
-------
I8.B-
N
2.8-
i.a-
.«-
.3-
'P
.1 -
.8-
.1
IB.8-
3.8-
2.8-
i.e-
.5-
.3-
•*
.1 -
.8-
.e
•V
f
+ +
k'
+1?
+
++
+ •<
+
+ +
•h
+
*
*+
+
+
18
f+
+
i;
ir
rf+*
%4
>+
V*
+
+
+
+
*
10
+
+
+
+
+ +
% ++ ' +
* + +
+
•f +
•
+
+ +
+
+
+
+
CTDM(11083)
.85 .18 .IS .2
iyiz
+ +
+
+ +
t / * + •
+
+ +
++ + +
+
+
+
- , CTDM(11083-E)
.es .ie .is .21
iyiz
Figure 13. Variation of observed-to-modeled ratios of SHIS #1 peak
hourly tracer concentrations with the product of the
turbulence intensities i and iz.
34
-------
nearly agree when the plume is well-diluted. When the dilution is much
weaker, the plume is more compact, exhibiting considerably less meander.
Under these conditions, the peak modeled concentration is very sensitive to
plume path assumptions, wind direction, and postulated flow distortion/plume
dispersion effects. This sensitivity is illustrated in the figure by the
large scatter for low values of iyiz. The figure also shows that the bulk
of the CCB data falls into this category.
35
-------
SECTION 3
DEVELOPMENT OF THE COMPLEX TERRAIN
DISPERSION MODEL (CTDM)
3.1 Introduction
CTDM is a point source plume model that incorporates several concepts
about stratified flow and dispersion over an isolated hill. The emphasis
to date has been on including those phenomena that are thought to be impor-
tant in controlling the magnitude and distribution of plume concentrations
across the windward portion of CCB. However, although the formulation of
modeling concepts contained in CTDM has general inherent applicability, cer-
tain details make explicit use of the geometry of CCB to simplify the code.
Consequently, CTDM should be viewed as a research code at this time rather
than a code suitable for regulatory applications.
We expect CTDM to evolve during the CTMD project. As more data are
analyzed from SHIS #1, and as data from SHIS #2 are analyzed, the specific
formulations will change as will the range of phenomena contained in the
model. These changes will be treated as updates to CTDM rather than the
basis for new model nomenclature.
CTDM evolved from model codes described in the first two Complex
Terrain Model Development (CTMD) milestone reports.
Two preliminary models were described in the First CTMD Milestone
Report. The Impingement model corresponds to low Froude-number flows in
which a plume remains horizontal as it impinges on and flows around the
hill. The Neutral model corresponds to moderate or large Froude-number
flows in which a plume goes over the top of the hill. A number of refine-
ments were made to these two models, and these were described in the CTMD
Second Milestone Report. These refined versions of the Neutral and Impinge-
ment models are referred to as the Lift and Wrap models.
Since then the Lift and Wrap models have undergone additional refine-
ments and have been combined into one model; now CTDM performs both kinds of
computations whenever the dividing-streamline height, H^,, is non-zero.
This section of the report provides a description of how Lift and Wrap were
modified and combined to form version 11083 of CTDM. The description is
followed by a detailed study of the data from two SHIS #1 experiment days.
CTDM upgrades suggested by the analysis of these data are presented, and the
performance of this latest version of CTDM is evaluated.
36
-------
3.2 Description of CTDM(11083)
CTDM treats the flow around an isolated hill as though the flow takes
place in two discrete layers. Flow below the critical dividing streamline
(1^) is restricted to travel in horizontal planes around the hill. Flow
above HC is allowed to rise over the top while undergoing some degree of
distortion consistent with weakly stratified flow over a sphere. That is,
Lift modeling assumptions are employed for plume material which is released
or which diffuses above H,,, and Wrap modeling assumptions are employed for
plume material which is released or which diffuses below HC. Because
streamline patterns above and below Hc are markedly different, the plume
is assumed to separate into two independent segments where the Hc surface
intersects the windward face of the terrain.
The segment initially below Hc spreads around the sides of the ter-
rain. As the plume grows with distance beyond the point of separation, how-
ever, its vertical profile is no longer Gaussian (or reflected Gaussian) in
shape. This is because the material above Hc at the point of separation
is no longer available to mix below HC.
The segment initially above Hc continues to travel up and over the
hill, but the plume material that had been below HC at the point of sep-
aration has been displaced by the solid terrain surface. The vertical pro-
file of the upper plume segment therefore is no longer a simple reflected
Gaussian distribution; rather, it is a reflected segment of a Gaussian dis-
tribution profile.
3.2.1 The Wrap Component of CTDM(11083)
Because observations indicate considerable meander of the horizontal
wind during a typical hour, we assume that the hour can be divided into
several quasi-steady periods. We also assume that the average angular
spread of the plume, as, corresponding to these periods is small com-
pared to the total meander, am, during the hour. Concentrations are
assumed to occur on the hill's surface only when a portion of the plume is
directed along the stagnation streamline; therefore, the maximum concentra-
tion occurs at the stagnation point.
During any single quasi-steady period, denoted by the subscript i, the
concentration at the stagnation point A (see Figure 14) will be non-zero
only if the mean wind direction during the period lies within otg^/2 of
the stagnation streamline (9^ = 9^ ± as£/2). If this concentration is
denoted by C^(9,9d), then the average concentration, Cmax, at A over an hour
is given by
c = J c.(e, e,) p(e) de
max 'id
O
(13)
where P(6) is the hourly probability density function of wind directions.
If the wind speed, vertical plume spread, and horizontal plume spread angle
are nearly constant among quasi-steady periods during the hour, then the
only non-zero contributions to the integral in Equation 13 arise for wind
directions within ± as/2 of the stagnation wind direction. Furthermore, if
37
-------
Receptor
Source: (xr, yr, zr)
Stagnation Streamline
Figure 14. Plume dispersion in the region of horizontal flow.
38
-------
the concentration distribution within the plume during each quasi-steady
period is assumed to be uniform, and if the spread due to plume meander
during the hour is much greater than as, so that P(e) is nearly constant
within the interval ots, then
C = C. f
max i
P(9)d6 - C. P(9_) a
ids
(14)
where
C. =
i
/2Tra U a d
z o s
(15)
In Equation 15, az, UQ, andas represent averages for the relevant hour.
From the CCB experiments we do not have a continuous record of 9(t)
as a function of t; only 5-minute averages of 9 and iy are available.
To use this _information, we assume that.9 is uniformly distributed in the
interval 2 S3dQ about the 5-minute averaged value 9, and we set
-------
sources, and integrate the response function or Green's function from the
level surface to Hc. Consequently, the "vertical" distribution of plume
material lies along the surface of the hill, and reflection occurs only
along the plane z = 0.
P(8.) Hc C (d,z) -0.5((z_ - z)/0')2
„/•_,. •> d -of R z
C(d+s,z) = •- S (e
a,R
a o
/2ir
-------
H
Section
(§)
(enlarged)
View
Undistorted Plume
Receptor (XR/ yR, ZR)
Section
View (A
Source: (xr, yr, zr)
Figure 15. Geometry used to derive the effective radius of
curvature Re for the portion of the hill above Hc.
and to incorporate the wind direction probability
distribution function P(6) in the Lift model.
41
-------
By assuming that the flow is axisymraetric, the relationship between the
undistorted plume and the distorted plume depends only upon the section of
arc through the receptor (Section A, Figure 15). In Figure 15, the angle
for specifying the probability factor is
6 - arcsin (yp/s) + e0
where 6O is the mean wind direction.
(21)
The third feature of the Lift model as incorporated into CTDM(11083) is
the treatment of the evolution of the vertical distribution of plume ma-
terial. As drawn in Figure 16, plume material below Hc is removed from
the profile at the Hc impingement level, and the remaining distribution of
material above HC continues to diffuse with full reflection from the hill
surface. To obtain concentration estimates beyond the point (so) where
Hc intercepts the hill along direction 6, the distribution of material
above HC at so is treated as a set of point sources, and the total con-
centration at downwind receptors is found by integrating the response func-
tion for each point source from Hc to infinity.
„, „, r" CQ(S>2)
C(s,9,0) = J — —
H /2ir a1
dz.
(22)
As in subsection 3.2.1, the solution to this integral produces a series of
product terms of error functions and exponentials, and
(a')2 - (a (s))2 - (a (s ))*
2 2 Z O
(23)
The distribution of material at s0 is given by
-0.
• 2ir U a (s )
02 o
2 - z)/o )
r z
-0.5((z
0/0 )'
+ e.
(24)
3.2.3 Dispersion Parameters
When the PDF form of CTDM(11083) is selected, only the dispersion para-
meter for plume spread in the vertical is needed. Background information
and analysis methods employed in developing the az expression used for
CCB data are described in Section 3.3. We find that values of az cal-
culated from the expression
iz(s
N(s
)/(U02y2))
(25)
with 2y2 = 0.55, fit the CCB data quite well. N is the Brunt-Vaisala fre-
quency and sv is the virtual source distance associated with source-
42
-------
REGION 2
Figure 16. Dispersion and flow regions for stratified flow
around hills.
43
-------
induced plume spread. If the PDF form is not selected, the dispersion para-
meter for horizontal plume spread is assumed to be given by
ay = iy(s
s'v)
(26)
where s'v is the virtual source distance based on the rate of plume growth
in the horizontal.
3.2.4 CTDM(11083-E) and CTDM(ll083-E)-5
The structure of the basic version of CTDM(11083) was discussed in the
preceding sections. Additional versions of CTDM( 11083) have been developed
to test the model sensitivity to alternate hill-effect assumptions and
alternate meteorological data resolution.
CTDM(11083), as described in subsection 3.2.2, presumes that streamline
distortion and the accompanying plume distortion for flow over the top of
the hill (Lift) do not cause any increase in diffusion over that which is
expected at similar downwind distances over level terrain. The plume cen-
terline approaches closer to the hill surface, but the vertical extent of
the plume is proportionally reduced, so concentrations at the surface remain
unchanged. This model for flow over the top is most appropriate if enhanced
diffusion due to the distortion of the streamlines can be neglected, and if
the flow over the hill is no more turbulent than that away from the hill.
The effects of increased diffusion (aided by the plume's approaching
nearer the hill surface) can be approximated in CTDM by enhancing the rate
of growth of az over the hill, without explicitly bringing the centerline of
the plume nearer the surface. The effect of this approach on the vertical
distribution factor is similar to the terrain factor approach contained in
the COMPLEX models and the Neutral model evaluated in the CTMD First
Milestone Report. But instead of viewing the factor as a change in the cen-
terline height, we view the factor as a change in the relative size of the
plume. However, because az increases, dilution also increases, which
reduces the centerline concentrations in the plume. Therefore, this treat-
ment is not equivalent to the terrain factor approach, particularly as
az approaches the height of the plume centerline over the terrain.
The a --enhancement approximation is made by altering Equation 23:
(a (s))
(27)
where
T = PPG
for
ZR- zr
T = 1 - (1 - PPC) (zn - H )/(z - H )
R c r c
(28)
for
z < z
R r
44
-------
PPG is the "plume path coefficient." This definition for the factor T is
the same as that used to define terrain factors in the COMPLEX models. This
enhanced-az version of CTDM(11083) (denoted as CTDM(11083-E)) has been
run on the CCB data base with PPG set equal to 0.5 in order to evaluate the
importance of increased dispersion over the hill.
A second alternative CTDM(11083) utilizes the sequence of 5-minute
meteorological data, rather than 1-hr average meteorology supplemented by
the calculated PDF for wind directions. , In this version, the 5"-minute
horizontal distribution of plume material is assumed to be Gaussian in form,
characterized by
a = i(5-min)
Y Y
(29)
All other model formulations remain the same. Because 1-hour average con-
centrations are constructed from calculated 5-minute average concentrations,
variations in wind speed dilution, turbulence intensities, and the dividing-
streamline height are explicitly modeled in addition to the distribution of
5-minute average wind directions. Running this version (denoted by append-
ing (5)) of CTDM(11083) on the CCB data base will test the importance of
including short-term variations in Hc, u, and turbulence intensity in the
modeling of SHIS #1.
3.3 Case Study Results - Experiments 201 and 210
The model evaluations discussed in subsection 2.4 indicate that some
improvement is gained in the performance of CTDM if az is enhanced over
the hill. However, much "noise" in the residuals remains. One possible
source of noise is the meteorological data used to drive the model esti-
mates. The MDA described in subsection 2.1 represents an objective best
estimate of the meteorological conditions at source height derived from data
measured at Tower A alone. However, alternate meteorological data for model
input might be developed for some experiments if all available data are
studied. Note that Tower A was located 2.3 km to the north of CCB, approxi-
mately one to three kilometers away from the tracer release locations.
We have begun an extensive case-study of the SHIS #1 data to develop
all available information on the transport and diffusion of plume material
at CCB. In this process, the consistency among the data will be tested, and
the performance of various modeling assumptions can be better evaluated.
Because the modeled concentrations are very sensitive to the trajectory of
plume material and the rate of diffusion of that material to the surface, we
are particularly interested in evaluating (1) the correspondence between the
sequence of 5-minute average wind directions interpolated to source height
on the basis of Tower A measurements and the smoke plume trajectory as
recorded by observers' field notes and photographs; (2) the characteristics
of the plume trajectory near the hill; (3) the correspondence between the
rate of plume growth in the vertical as observed in photographs and the
measured vertical turbulence intensity on Tower A; (4)"the relationship
among turbulence intensity measurements on the hill and those at Tower A;
and (5) the form of the growth of az upwind of the hill as derived from
45
-------
a combination of lidar data and photographs. Results of these evaluations
performed on data from Experiments 201 and 210 are summarized in this sec-
tion.
3.3.1 Plume Trajectories
Key views of the oil-fog plume that are important in estimating plume
trajectories from photographs are those taken from behind the release crane
looking along the plume trajectory and those taken from atop CCB looking
back toward the release crane. For those views from behind the crane, wind
directions are most easily estimated when the plume passes over some
recognizeable portion of the hill. Any horizontal displacement of stream-
lines when the plume is close to the hill is subjectively taken into
account. For views from the hilltop, wind directions are most easily esti-
mated when the plume is seen to pass over one of the camera positions on
either peak, or when the plume trajectory "passes through" a known landmark
on the ground away from the hill (such as a turn in a road). On the basis
of these considerations, wind direction estimates from the photographs are
least accurate when the plume misses the hill to one side or the other.
Fortunately, the most useful information during these periods is that the
plume is off to the side of the hill, and is therefore having very little
impact on tracer concentrations in the sampler array.
In Figure 17, photo-derived estimates of wind direction at the source
for each available 5-minute period during Experiment 201 are compared with
the MDA wind direction estimates and also with the wind direction measure-
ments at the 10-ra level of Tower B, which is located on the south peak of
CCB. The first 2-hour segment of data displays good general agreement in
wind direction, but the period between 2000 and 2200 displays more scatter.
In particular, the various wind direction estimates would lie nearer one
another if the Tower A estimates were shifted by approximately 10 minutes
between 2020 and 2030. Because the plume was released 952 m northwest of
CCB, the release point is nearly as close to Tower A as it is to Tower B.
Furthermore, because the wind speed exceeded 3 m/sec during the interval,
events at Tower A would be swept across the source area within about 5
minutes. Therefore, the discrepancy between the Tower A data and the infor-
mation inferred from oil-fog plume photographs and measured at Tower B is
probably due to spatial as well as temporal inhomogeneity in the flow. For
modeling purposes, it is evident that Tower A wind directions do not accu-
rately depict plume transport for this period.
Figure 18 displays the time-series of 5-minute average wind directions
for Experiment 210 as estimated from photographs, as interpolated to release
height on the basis of Tower A measurements, and as measured at the 10-and
30-m levels of Tower B. These results indicate that the 10-m level of Tower
B is most representative of the oil-fog plume transport directions for the
three hours of available data from this experiment. Tower B wind direction
data from the 30-m level are generally within 5-10° of the "photo" direc-
tions for the remainder of the experiment. The wind directions inferred
from Tower A data correspond well with the "photo" directions at times
during the experiment but are off by up to 20° at other times. This rela-
tive lack of correspondence is probably the result of relying on 80-m level
wind direction data to supplement the incomplete wind data for the 40-m
46
-------
^
§
o ^
O
S
LA
CD
CD
B
o
SU
PH 0
nj oi f-i
•P 0 13
ni fi 5)
13 nj
" 0
to
0 -H
(D
13
fi
0 fti3
S oj 0
•H f-i -p
•P bO cti
O .-<
Pi -P O
O O ft
•H
P ft
O
0 B
fH 0
•H
0
r!
•H
T3
g
•H pa
<4-t 0 fH
O 13 0
c •> o
O i-l H
tn o
•H CM g
nj -P f-t
ft S Mn
IIs<
CJ -H
f-l fH O f-l
0 0 i-H 0
P ft S
C X +J O
h-1 w nj E-1
0
!H
g)
47
-------
WD (deg)
158.01
140.0-
138.8-
120.0-
110.0-
188.0'
1 2 3
Photographs (oil-fog)
Modelers' Data Archive
7 8
Time (hr)
WD (deg)
150.01
140.0-
130.0-
128.0
118.8-
100.0 liiiiiiniiiii
0 t
Photographs (oil-fog)
Tower B @ 10 m
Tower B @ 30 m
7 8
Time(hr)
Figure 18. Intel-comparison of wind direction time-series data for
Experiment 210, derived from photographs, wind measurements
at 10'and 30 m from Tower B, and interpolated wind
measurements from Tower A. .. ., '
48
-------
level on Tower A. Modeling for this experiment must therefore consider the
wind directions measured on Tower B and those inferred from the photographs.
3.3.2 Vertical Turbulence Intensity and Initial Plume Growth
Experiment 201
Values of the vertical component of turbulence intensity, iz, have
been estimated; from available photographs by several methods and compared
with iz values measured on Tower A. Photographs used in the analysis for
Experiment 201 are wallet-size enlargements from 35-mm film. Lengths
measured on each photo are converted to angles by the formula:
a = arctan (&('mm)/113 mm)
(30)
where'A is the measured length, and 113*mm is the effective focal length
for the enlargement (50 mm • 2.26). The enlargement factor of 2.26 was
estimated by comparing the size of the dominant features on the film and the
enlarged print.
These angles are converted to full-scale distances by means of the
geometry depicted in Figure 19. The., length C is the known distance between
the camera position and the release crane; A is the distance between the
release crane and any point downwind along,the wind trajectory; B is the
distance between this point and the camera position; and the sum a^ +
a 2 is. the total angle a subtended by A as viewed from the camera
position. Because this angle can be large, the small angle approximation to
Equation 30 is not .used, and all distances along the film plane are measured
from the center of the image. Hence, two measurements are needed for a.
The angle 3 is obtained from the wind directiolfand the angle' frbm the
camera position to the release crane. With these definitions, the distance
downwind of the source is
A —
C sin (ex)
sin (180-a-3)
(31)
Similarly, the thickness of the visible oil-fog plume is obtained from the
photograph as an angle,
-------
Wind Vector
\
Source Position
B
Camera Location
Centerline of
the Photograph
Evaluation
Point
Figure 19. Definition sketch for computing distance along the
wind trajectory from the source.
50
-------
zl
a /x = E/2V/J .
z '
(33)
Method 2. This approach utilizes the same principle as Method 1, but
instead of measuring the spread angle at the source, we measure the plume
thickness, h, at several distances along the plume and set the spread angle,
E, equal to the maximum value of h/x.
Due to the uncertainties introduced in using the MDA wind directions
for the respective 5-minute periods, photo-estimated directions are used to
obtain h(x); and instead of assuming a top-hat distribution to relate E to
iz, we have assumed a bi-linear "triangular" distribution because the
smoke density is not uniform across the plume. In this method, the vertical
intensity of turbulence is given by
° _ (34)
max
Method 3. Due to uncertainties in defining the edge of the plume and
in relating the az estimates to a Gaussian distribution, Method 2 was
judged to be of marginal utility. Method 3, used in place of Method 2,
applies the technique of Gifford (1980) to relate observed plume dimensions
to equivalent az estimates, assuming that the plume spread is Gaussian.
This approach provides an estimate of a_ for each observation of h when
£i
a maximum value of h is specified for each plume.
Ideally, the plume should be photographed against a uniform back-
ground. It should be uniformly illuminated, and it should be visible at
distances large enough to define hmax. Most of the nighttime photographs
from Experiment 201 were taken with a fairly uniform background, but the
lighting is not uniform and hmax is generally not clearly established.
However, most of these photographs show a slowly-growing- plume at the larger
distances downwind, so we have set hmax = h(xmax) + 1 (rounded to the
nearest whole meter) for those cases in which hmax is not clearly attained
within the field of view of the photograph.
With this assumption, each value of h measured on the photographs gives
rise to an estimate of az for an effective Gaussian distribution of
plume material. The turbulence intensity is then calculated with
i 0 = (CT /x)
z 3 z max
(35)
Equation 35 once again assumes that the vertical growth of the plume with
distance (time) is linear in the initial stages over which az/x should
attain its maximum value. Due to the assumptions required for this method,
it should provide "upper bound" estimates for iz.
Method 4. This approach makes use of the estimates of az derived
by means of the Gifford technique, but it obtains iz estimates by matching
the observed values of a,, (t) to
az(t) = awt/(l
Nt/p)
(36)
51
-------
at t
ax-
Equation 36 is the model for az that was used in the
Second Milestone Report. In that report, p was a free parameter that was
set equal to 1.5 on the basis of comparing az estimates with az values
obtained by the WPL lidar system at CCB.
to solve for p with crw = iz3*U and az(t)
approximately 0.6. Consequently,
However, when Equation 36 is inverted
az (Gifford), p is found to be
z (Gifford)
NX
o.s
'z4
x
'
0.6u
(37)
x ^ 0.5 x
max
where xmax is the,farthest downwind distance at which a value of h can be
estimated. ;
Figure 20 (top panel) shows the correspondence between izi and the
30-m iz value interpolated from Tower A data. The intensities nearly
agree during the first two experiment-hours and during the sixth hour, but
show significant differences during the fourth and fifth experiment-hours.
In these two hours, izj_ values appear to be either too small or too
large. It should be pointed out, however, that the larger values of izi
occur during periods of significant horizontal plume meander and that this
horizontal "smearing" could appear as an increase in the vertical dimension
of the plume in the photographs.
The 30-m interpolated iz value from Tower A is also compared to iz3
in Figure 20 (bottom panel). The iz3 values are moderately greater than
the Tower A values during hours 18, 19, and 23. But they agree better dur-
ing hours 21 and 22, with the notable exception of the peak at 2100. Al-
though the peak in the "photo" iz's may have an artificial component as
has been speculated above, it is unlikely that the peak should be entirely
absent. In fact, it is tempting to shift the plots in time so that the
fluctuations in the two data sets nearly coincide. Figure 21 shows the
result of shifting the "photo" iz values by -10 minutes.
Several of the peaks in hours 21 and 22 nearly line up in Figure 21.
Therefore, it seems plausible that the turbulence field near the smoke plume
lags the turbulence measurements on Tower A during this period of time. We
saw evidence for this 10-minute phase lag in the wind direction time-series
plotted in Figure 17. Plots (not shown here) of aw and iz from 2, 10
and 40 m on Tower A and from 2 and 10 m on Tower B further support the
interpretation that the wind shift and the vertical turbulence occur uni-
formly later near the hill than at Tower A. If this is correct, then the
photo iz values are roughly 2 to 3 times the Tower A iz value during
this part of Experiment 201.
The time-series of iz4 values is plotted in Figure 22 alongside those
Tower A iz values interpolated to the release height. It is seen that the
iz4 estimates tend to inflate the larger values in the iz3 time series,
but change very little otherwise. The lower part of Figure 22 compares
iz4 with the iz values measured at 10 m and 40 m on Tower A. The com-
parison suggests that "actual" iz values at plume height could have been
underestimated in the MDA during the first two hours of Experiment 201, and
by similar reasoning they could also have been underestimated during those
52
-------
11.0
io.o
9.0
8.0
7.0
6.0-
)
5.0
4.0-
3.0-
2.0-
1.0-
0.0-
—•»•— iz-| (photo)
Modelers' Data Archive
17.0
16.0
15.0-
14.0-
13.0-
12.0-
11.0-
10.0-
9.0-
8.0-
7.0- /
6.0- '
5.0-
4.0-
3.0-
2.0-
1.0-
0.0
17.0
18.0
19.0
20.0
21.0
22.0
23.0
Time (hr)
iz3 (photo)
Modelers' Data Archive
18.0
19.0
20.0
21.0
22.0
23.0
Time (hr)
Figure 20.
Intel-comparison of vertical turbulence intensity (iz)
time-series data for Experiment 201, derived from
photographs by Methods 1 and 3, and interpolated from
Tower A. (Circles denote periods of considerable
horizontal meander.)
53
-------
11.0
10.0
9.0
8.0
7.0
6.0
)
5.0
4.0-
3.0-
2.0-
1.0-
0.0-
20.0
16.0
15.0
14.0
13.0
12.0
11.0
10.0
9.0
8.0
7.0-
6.0-
5.0-
4.0-
3.0-
2.0-
1.0-
0.0
—4— iz-| (photo: shifted by-10 min)
Modelers' Data Archive
21.0
22.0
Time (hr)
iz3 (photo: shifted by-10 min)
Modelers' Data Archive
»
20.0
Figure 21.
21.0
22.0
Time (hr)
Intel-comparison of vertical turbulence intensity (iz)
time-series data for two hours of Experiment 201, derived
from photographs by Methods 1 and 3, and interpolated
from Tower A. The "photo" estimates are shifted by
- 10 minutes.
54
-------
16.0-
15.0-
14.0-
13.0-
12.0-
11.O-
.I 0.0-
9.0-
iz(%) 8.0-
7.0-
6.0-
5.0-
4.0-
3.0-
2.0-
1.0-
iz4 (photo)
iz (Modelers' Data Archive)
Q.O.
17.0
27.0-
18.0
19.0
24.0-
21.0-
18.0-
15.0-
12.0-
9.0-
6.0-
3.0-
0.0-
h— iz4 (photo)
iz (Tower A @ 10 m)
iz (Tower A @ 40 m)
20.0
21.0
22.0
23.0
Time (hr)
17.0
18.0
20.0
21.0
22.0
23.0
Time(hr)
Figure 22. Intel-comparison of vertical turbulence intensity (iz)
time-series data for Experiment 201, derived from
photographs by Method 4, and interpolated from Tower A.
(Circles denote periods of considerable horizontal meander.)
55
-------
periods of time in which the iz value at 40 m was nearly zero. However,
the interpolated tower values appear appropriate for the remainder of hours
21 and 22, with the exception of the,peak iz (photo) values.
Experiment 210
Processed lidar data are available for two of the hours in Experiment
210, and these data are used to supplement the photographic data in evaluat-
ing the representativeness of vertical turbulence intensity estimates at
plume height (MDA). Photographic and lidar coverage of experiment-hour 3 is
the most extensive for this experiment, although photo-documentation of the
initial stages of plume growth near the source is generally less extensive
than that during Experiment 201. Consequently, much of the analysis has
been performed on experiment-hour 3.
Unlike the previous use of SHIS #1 lidar data, we have looked more
carefully at the processed data to evaluate how well calculated second
moments of the backscatter distribution represent an effective az for
the oil-fog plume, and how well an estimate of the 1-hour az
by
calculated
= a
JE
(38)
agrees with the crz for a superposition of each individual distribution
during the hour. Note that CTZ is the average of the squares of the
calculated second moments of the individual distribution profiles during the
hour, and OH is the variance of the plume centroid height during
the hour. Recall that much of our previous evaluation of the proposed
a 2 growth function used in CTDM was based upon the 1-hour crze values
obtained from the lidar data using Equation 38.
In this new analysis, best-fit oz values are obtained from the pro-
cessed individual scan lidar data arrays by following these steps :
1) Data arrays are scaled from relative bscat intensity to actual
bscaf
2) Data arrays are integrated in the crosswind direction to yield a
crosswind-integrated vertical profile of bscat.
3) A Gaussian envelope is fit to the vertical profile by the method
of least square errors. The envelope is that of a reflected
Gaussian. The location of the centerline is set at the centroid
height, and the surface is set to the height of the terrain
beneath the centroid. The fitting procedure provides both the
centerline bscat value and the az for the best-fit curve.
Step 2 takes any apparent rotation or tilt of the plume in cross-section
into account.
56
-------
Best-fit cz values for extended periods are obtained from the ver-
tical profiles resulting from step 2 above in the following way:
• A vertical grid-point spacing is set up to match the greatest ver-
tical resolution available in each scanning plane.
• , An average surface elevation is specified for each scanning plane..
scat
are interpolated to the vertical grid and summed
over all scans within each plane.
• A Gaussian envelope is fitted to the resulting vertical distribu-
tion, and best-fit az and maximum bgca^ values are obtained.
The crosswind-integrated/time-integrated vertical profiles of the plume
upwind of the hill are remarkably Gaussian. An example vertical profile is
shown in Figure 23. The best-fit CTZ values for the individual means are
generally not much different from the second moments of the distribution.
The average ratio of the best-fit a.
individual scans for hour 3 are:
to the second moment for the
scan plane
ratio
211°
.90
225'
.92
236C
.95
244'
.92
251'
.84
264°
1.06
Those for hour 6 are:
scan plane
ratio
211°
1.14
225°
1.20
236°
1.00
244C
.86
251°
1.14
264°
1.23
In each case, the 236° plane had by far the most samples.
Focussing on the 236° scanning plane, the ratio of the best-fit 1-hour
az to the effective az (second moment plus centroid) is found to be
0.77 in hour 3, and 0.94 in hour 6. These results tend to suggest that the
effective az values used in the CTMD Second Milestone Report may be
overestimating the vertical size of the plume.
Photographs of the plume taken from the side of the plume trajectory
are available from two locations during experiment-hour 3. Because lidar
data are available for this hour, measurements of optical smoke plume thick-
ness from these photos can be compared with the best-fit az values ob-
tained from individual lidar scans. Although the lidar data are not
5-minute averages of the plume as are the photographs, the plume appears
very coherent during this hour, which suggests that a comparison of the two
measurement techniques may be instructive.
Estimates of plume thickness are 'obtained from the photographs using
methods similar to those discussed above. The major departure from that
methodology involves projecting the negative image of each frame onto a
sheet of graph paper to obtain the plume dimension and position data.
Lidar-az data are compared with photo-h (thickness) data only at
those points where the lighting is sufficient to fully illuminate the plume.
57
-------
4J -H
P-l P^
O X
f-i W
U
0
•H O
<4-( O
•H O
4-> O
3 00
-P O
t/1 'H
•H 4J
<» ro
CCD
Oj 0)
M O
S
> O
O
O CN
Q CO
CD
O
bO
58
-------
For photographs taken at camera position 0-11, these points are close to the
source, about 50 to 150 m downwind. For photographs taken at camera posi-
tion 0-10, these points are near the base of the hill Ov< 600 m downwind),
and also in the col area when the, plume traveled over the top of CCB and
through the illumination of the draw tower lights. By taking only those
lidar and photo .data coincident in time, and' within 50" m"d£ stance s'we' find
that.the six available ratios of h/az average 3.69 with a variance of
0.24; Due to the criteria for strong illumination, the photo-h data ought
to fully repre'sent total plume thickness, and this may be why the variance
is so small*
If we had assumed a top-hat distribution for the plume material, then
the ratio of h/az would be 3.46. If we had assumed a semi-circular dis-
tribution, h/
-------
, ^/
CO
_ — —^
, _/
I
•H
X O
__ ^s
•*.\^
C "d
0)
0 -H
•H fH
4)
0) O H
i — I i — I
3 CM g
£> O
f-l -P f-l
3 c m
P (D
S -d
r-H -H CD
Oj fH -P
(J (D OJ
•H Pnr-1
-------
Conclusion
Vertical turbulence intensity data contained in the MDA generally "fol-
low" the observed behavior of the oil-fog plume. However, there are periods
in which these data do not appear to be representative of .turbulence at the
release site. Excluding these periods, there remains an apparent underesti-
mation of iz in the archive data compared to rates of initial plume growth
obtained from photographs. Some of this may be due to the neglect pft
source-induced turbulent mixing. Results obtained after including source-
induced turbulence are discussed in subsection 3.3.4.
3.3.3 Plume Growth in the Vertical
Because CCB surface concentrations can be very sensitive to the size of
az, the form of crz(x) is an especially important topic for evaluation. In
particular, we have evaluated the form of az(x) proposed in the CTMD Second
Milestone Report by analyzing the growth of photo-derived and lidar-derived
az values with distance between the release position and the upwind base
of CCB. , , •
Experiment 201
Plume thickness values obtained from photographs are converted to esti-
mates of 0Z by using the Gifford technique and a top-hat distribution
assumption. Because processed lidar data are as yet unavailable for Experi-
ment 201, there is no check on which of these methods is more appropriate.
Also, although the analysis of lidar data from Experiment 210 showed that
az ^ 3-7 h, this result may not be appropriate for Experiment 201
because the lighting conditions are substantially different in these two
experiments. Consequently, this analysis of the vertical growth of the
plume with distance will continue to focus on both methods. Note that the
top-hat method is expected to underestimate actual crz values with in-
creasing distance downwind because the "edges" of the plume are lost. Also,
the Gifford technique is expected to overestimate actual az values with
increasing distance because the true maximum in apparent plume thickness is
not usually seen in the photograph, and because the plume brightness may
decrease with distance due to the light scattering properties of the plume
rather than due to dilution.
Vertical turbulence intensities estimated from photographs are referred
to by izi, iz3, and iz4, and these have the same meaning as in the
previous subsection. az^ values are computed from plume thickness data
by assuming that the plume material exhibits a top-hat distribution:
a . = h/2/3
zl
(39)
az2 values are computed from h data by means of the Gifford technique
(only azi and az2 are used in the subsequent analyses). Vertical
plume spread is also computed from the PGT dispersion parameter formulas for
the most appropriate stability class as contained in the MDA. These are
simply denoted by oz(PGT).
61
-------
Figure 25 displays the ratio of azi and az2 to az(PGT) versus downwind
advection time. The ratios are plotted as logs to facilitate "factor of 2"
types of comparisons. It is evident that the aZ2 estimates follow a
growth law more similar to that contained in the crz (PGT) curves than do
the az± estimates. The analyses also suggest that az(PGT) values would
match the photographic estimates better if they were based on a more refined
estimate of vertical turbulence intensity than the surface stability class.
Even so, the PGT curves tend to overestimate the rate of growth of the ver-
tical spread with time as derived by the Gifford technique.
Figure 26 contains comparisons of azl (top hat) against az values
calculated by means of Equation 36 with p set equal to 1.5. As in
Figure 25, the ratio of azi to az is plotted as a log versus advection
time. Two versions of crz (calculated) are presented: one uses iz]_»
the other uses iz3« These two estimates of iz tend to bracket those
values interpolated from Tower A data. The results of the comparison show
that the functional form assumed in Equation 36 causes the plume to grow too
rapidly with advection time, compared to the development of az obtained
with the top-hat assumption. Futhermore, iz^ tends to align az^ with the
calculated values better than iz3 because the iz^ values are also derived
from the top-hat distribution assumption.
Figure 27 indicates that Equation 36 performs better when az2 (Gifford)
is viewed as the more accurate description of the plume spread. The upper
portion of the figure compares aZ2 with crz calculated by means of Equation
36 with iz = iz3 and p set equal to 1.5. The distribution of ratios is
centered at approximately 1.0, and the rate of growth of aZ2 beyond
approximately 30 seconds travel time tends to follow that prescribed by
Equation 36.
Because Equation 36 appears to simulate the overall growth trend in
°z2 beyond t = 30 sec during Experiment 201, we have used the a22 and iz3
data to re-evaluate the value of the parameter p.
By solving Equation 36 for p,
_ Nt (40)
p , .
(i ut/a ) -1
z z
2
and a least squares fit to the scatterplot of (iz3ut/az2) -1 versus Nt
produces a value of 0.6 for p. In the notation of Pearson et al. (1981), p
s 2y , and so a value of 0.6 produces a f of 0.55. The theory of Pearson et
al. (1981) indicates that YJ a molecular exchange coefficient, could range
from 0.1 to 0.8.
Using this new value of p, we have recalculated the iz^ estimates
described previously. Recall that the matching actually takes place at the
measurement point nearest to 0.5 t^x, where t^x is the measurement
point farthest from the release crane. A re-evaluation of p was made with
o~£ and iz4 as a consistency check, and the results are displayed in
Figure 28. The upper plot contains all data, and the lower contains only
those data that were not used to match iz4 to az2- The value of p is
nearly the same in both plots; p (upper) = 0.567, p (lower) = 0.566.
62
-------
ln(£rz1/
-------
ln(crzl/iz-|uf)
z.eee
Time (sec)
Figure 26. Comparison of az (calc.) growth curves with az-j_ data
derived from Experiment 201 photographs. (az(calc.)
iz uf, f = t/(l + Nt/p)°-s).
64
-------
In (crz2/iz3uf)
2
In (crz2/iz4uf)
Time (sec)
2.886
1.758 -
t.see-
I.2S0-
1.888-
.758-
.sae-
.258-
.258-
.SB8-
-.750-
-1.803-
-1.2SB-
-I.50B-
-I.7SB-
-2.ee
p=0.6
Time (sec)
Figure 27. Comparison of az(calc.) growth curves with az2 data
derived from Experiment 201 photographs. (az(calc.)
izuf, f = t/'•
65
-------
(iz4Ut/o-z2)2-1
ia.8 12. e
(iz4Ut/crz2}2-1
14.
Nt
(iz4 fit points removed)
• B 2.8 4.8 E.B 8.8 IB.8 12.8 14.8
Nt
Figure 28. Scatterplots used in computing best-fit p for
Experiment 201.
66
-------
These analyses increase our confidence in estimating az by Equation 36.
We see that the growth trend is similar to those Gifford technique estimates
of az derived from the photographs. And we also see (Figure 22) that
the iz values inferred by matching the photo oz values to Equation 36, with
p = 0.6, at the .measurement point nearest 0.5 tmax are generally "reason-
able" compared with iz values measured on Tower A.
• Experiment 210 '"....' ,
The analysis of the growth of CTZ with distance during Experiment
210 focuses on experiment-hour 3 because the link between vertical plume
thickness as measured from the photographs and az as calculated from the
lidar data is best for this hour.
The relationship among the photo—estimates of az, the lidar estimates
of az, and the curve CTZ(X) computed with Equation 36 (p = 0.6) and the
photo-derived iz estimates are summarized in Figure 29. In this figure,
the "plus" symbols represent the photo-estimates of az; the "box"
symbols represent the lidar-az values; and the solid line is the com-
puted az(x) curve. The figure suggests the following:
• The computed az (x) curve corresponds well with most of the
data points out to 600 to 800 m from the source.
• For those periods without photo-estimates of az, simply
multiplying the interpolated iz data from Tower A by 1.37 (the
reciprocal of 0.73) places the calculated crz(x) curve in the
neighborhood, of the lidar-az points.
• When the plume passes through the col, the apparent size of the
plume is on the order of one-half the value expected in the
absence of the hill.
• Although the az(x) curve appears quite reasonable in general,
the correspondence with the other data would improve somewhat if
the iz value were increased, and if the curve grew more slowly !
than with distance to the 1/2 power beyond 200 to 300 m.
These results indicate that the form of Equation 36 is appropriate for
estimating the dispersion of elevated smoke releases during Experiments 201
and 210, at least over distances on the order of 500 m to 1000 m. Unre-
solved, however, is the question of whether the square-root distance growth
continues well beyond 1000 m at CCB. This far-field growth law is not of
critical importance for the purpose of estimating concentrations over CCB,
but it will become more important in estimating concentrations for full-
scale power plant situations. For the present, we shall use the formulation
of Equation 36, and re-interpret our estimates of iz by including a
measure of source-generated turbulence in, the analysis .
3.3.4 az Tuning Parameters
The preceding discussions of the CTZ analyses assumed that iz and
p could be viewed as free parameters in fitting observed estimates of az
67
-------
2 S
o
§ f
i
o o
o >
•P (D
pj id
0) I
•H cd
H T3
CD -H
CD \o
p o d
CD
t/> 10 c
CD N
^
6 cu
O > ^
fH -H +
4-1 ^ rH
-------
69
-------
70
-------
to the other variables in Equation 36. But Equation 36 does not explicitly
contain the effect of source-induced turbulence on the development of a
with distance. This omission could be responsible in part for discovering
that the photo-derived iz estimates are generally greater than iz values
contained in the MDA.
When the source-induced turbulence is included as a "virtual source
distance" (xv), Equation 36 is rewritten as
a (x) = i (x + x )/(! + N (x + x )/u 2Y* )°'5
^ , Z V V
(41)
so that there are three free parameters. xv is not treated as a parameter
in itself, however. Instead, the initial mixing is characterized by an ini-
f~ 1 a 1 rr T.T!-» o >• d
tial crzo, where
azo = az(x = 0)
(42)
Fixing the value of azo therefore allows one to calculate xv by means
of Equation 41.
A Monte Carlo parameter optimizing scheme is used to select the
combination of azo, iz, and y that provides the best fit of Equation 41 with
the estimated az data. When this is applied to each 5-minute period of
Experiment 210-hour 3, y is found to vary from 0.16 to 0.76, with an
average value of 0.4. The range in y reflects the range in the shape of
oz(x) contained in Figure 29. The smaller values of y apply to a (x)
curves that level out more at large x, and the larger values apply to az(x)
curves with a greater rate of growth at large x. Due to the uncertainty in
the az estimates in Figure 29 and the possible influence of the hill on
the plume near the upwind base of the hill, no clear trend in y could be
identified in the data to construct a model for y. Therefore, we choose
to fix y at a representative median value. On the basis of the analyses
presented in subsection 3.3.3, and the generally good agreement between the
data in Figure 29 in which crz(calc.) is based on p = 0.6, y is chosen to be
0.525 (and p is now 0.55).
°zo and
Fixing the value of y forces the optimizer to adjust the values of
iz in fitting Equation 41 to the data. Using the data for
Experiment 210-hour 3 once again, azo varies from 0.3 m to 1.7 m. Eight
of the twelve values are less than 0.75, and these average 0.48 m. The
best-fit value of iz is not very sensitive to the value of CTZO when azo
is in the 0.3-m to 1.7-m range. For example, if 0.5 is substituted for
1.21, the best-fit iz value changes from 0.014 (1.4%) to 0.016.
Therefore, most of the variability in azo whenazo is appreciably different
from 0.5 during this experiment-hour goes toward improving the goodness-of-
fit. A representative value for azo appears to be 0.5 m.
All of the experiment-hours of Experiment 210 produce similar ranges of
y and azo. When y is fixed at 0.525, and azo is fixed at 0.5 m, the
best-fit values of iz generally follow the measured i_ data. Differences
£i
71
-------
between the best-fit and measured iz data are found to result from the
following factors:
• waves - the plume had pronounced waves in it at times, and best-
fit iz's were approximately 1.3 times the measured iz's. How-
ever, this difference is reduced considerably if we assume that
the photographs show a triangular distribution rather than a top
hat.
• horizontal spread - during some periods of considerable horizontal
meandering or spreading of the smoke plume, the vertical size of
the plume is overestimated.
However, there are periods in which the photographic data are genuinely dis-
similar to the Tower A turbulence intensity estimates at the plume height.
These occur when the tower iz values are less than 0.008, or when the
photos show periods of enhanced vertical spread. Figure 30 contains the
time-series plots of the interpolated iz data from the MDA and the best-
fit iz data from the preceding analysis. Those periods in which an undis-
puted discrepancy exists between the two iz values are indicated by
circling the best-fit iz data points.
The analysis of plume spread data from Experiment 210 indicates that:
• the oz formula with y = 0.5 and azo *> 0.5 is appropriate for
moderate wind speed (3-8 m/sec) experiments such as 210, when the
oil-fog is generated by the TIFA thermofogger; and
• turbulence intensities interpolated to the release height from
Tower A data are generally appropriate, although the lower values
(less than 0.008) appear consistently too small.
The same analysis has been applied to the data for Experiment 201. In
this case, we have taken the "az-£* estimates of plume spread derived
from the photographs with the Gifford technique. Preliminary tests of fit-
ting several randomly selected 5-minute time periods with y, azo, and
iz as free parameters produced several estimates of y between 0.3 and
0.6. On the basis of the overall fit of 0.57 (subsection 3.3.3) for Experi-
ment 201 and on the results found in analyzing Experiment 210 data, y was
set at 0.55; azo was set at 0.5; and best-fit values of iz were
obtained.
Figure 31 shows these estimates of iz compared to the iz data
contained in the MDA. Between 2020 and 2145, the archive wind speed, Brunt-
Vaisala frequency, and iz data are shifted in an attempt to account for
the apparent lag between the Tower A data and the plume and Tower B data.
Data for 2020 are persisted for 10 minutes, and the remaining data in the
period are shifted in time so that the 5-minute period ending at 2035 is
filled with data from 2025, and so on. The photo-derived az data are
fit using Equation 41 with the shifted u and N data, and the resulting best-
fit i2 data are compared against the shifted MDA iz data in Figure 31.
72
-------
t . _ _\
CO
(1)
+•»
(0
E
Q.
CD
D,<
O Q
si
I I
•— L.
~ 3
o to
~ CO
r- CD
<* E
E E
ir
iL
O tr
co £.
i—(
•H -H 43
erf m ft X
P erf .-i
S X fn CD
O 4^ bo !>
O o -H
0 P P
O erf
P nj
in CD
X CD 43
P -P P
w S *
f3 -H O
CD P .fn
(D
f-i
P
S
•H
CD P P
O -H erf
f3
•H O
P S
I/) CD
erf erf fn
O p tsi-H
•H erf b CJ
P Q •—'
f-i 0)
CD - 43
>
S O CD
•H r-f g
fH CD erf -H
erf 43 cj f-i
fl, +j >—r
-------
i
-a .
»- CD •!->
C -M
•H -H
0) <+H
O ,0
ctj
CD fi
CD
^ S -
cd S -H
-P
_
bo
ni
w B fn
PI -H H oi
(D +-> O i— I
4-> 10 'W
^3 +J 0)
CD -P cti >
O -H TJ -H
C S •(->
(D CN cti
r-H t> CD
rQ -H
•P
r
O CD <+H
> O
CD
cd ed
O 4-> r
•H cd o
•P Q
CD -
-------
Aside from the major isolated peaks in the best-fit iz time-series,
the best-fit estimates lie reasonably close to the MDA iz values. This
boosts our confidence in modeling az at CCB with Equation 41 and the
turbulence data interpolated to the release height. However, this analysis
has also indicated that the Tower A data are not always representative of
conditions at the release. The smoke plume undergoes considerable vertical
growth at times when the archive data would indicate more modest growth and
when low values of turbulence intensity are recorded, the smoke-plume growth
does not reflect this.
An assessment of how well the calculated crz data fit the photo-
derived oz data can be obtained by plotting the photo-derived a
values against the calculated az values. Figure 32 shows the result of
doing this for Experiments 201 and 210. The fit is uniformly best for the
smaller values of az because these are closer to the release point, and
are therefore controlled most by the best-fit iz value. Beyond this
region, the shape of the growth curve is more strongly influenced by y, N,
and u. Here the fit is not so good, but most of the calculated values lie
well within a factor of two of the photo-derived values. These figures,
taken with Figures 30 and 31, indicate that y = 0.525 and azo = 0.5
are reasonable representative values for Experiments 201 and 210.
3.3.5 Turbulence Over GCB
Experiment 201
Five towers were operated on CCB during SHIS #1. Tower B (a 30-m
tower) was placed on the south peak, and Towers C, D, E, and F (10-m towers)
were placed near the 70-m height contour on the northeast, southeast, south-
west, and northwest sides of CCB, respectively. Vertical turbulence data on
CCB for Experiment 201 are available at the 2-m level of Towers C, E, and F,
at the 2-m and 10-m level on Tower D, and at the 2-m, 10-m, and 30-m level
of Tower B. The 2-m and 10-m level data from these towers, the turbulence
data from Tower A at the 2-m, 10-m and 40-m levels, and the Tower A turbu-
lence data interpolated to the smoke-release height (30 m) are plotted in
Figure 33.
The data interpolated to smoke height are taken from the MDA. Other
Tower A data and the Tower B data are taken from the full data base, and a
prop-response correction factor is applied (as discussed in CTMD Second
Milestone Report). Data from Towers C, D, E, and F are taken from the full
data base, adjusted for prop response, and then adjusted for the inclination
of the mean 5-minute wind to the horizontal. The latter adjustment is an
approximate method to obtain the turbulence intensity in the direction
perpendicular to the mean streamline of the flow:
+ v
w
_2
U
W
(43)
This formula is approximate in that the'full expression for
-------
Sigma-z Observed (m)
EXPERIMENT 201
25.a 38.0
Sigma-z Calculated (m)
Sigma-z Observed (m)
EXPERIMENT 210
.8 2.8
e.e e.e IB. a 12.
.e 2e.e
Sigma-z Calculated (m)
Figure 32. Comparison of photo-derived crz data and calculated CTZ
utilizing best-fit iz data in Equation 41 with
7= 0.525 and crzo = 0.5.
76
-------
*l
r""
"S"
05
tq o
CM
r\
Q -P
S
* 0)
U S
o o
E-> M-l
e <
o
h f-l,
<4H (D
CTi O
•P E-i
nj
•U S
O
tslf-l
• H <+H
^ Cti
e +j
cd cd
E
O
g>
'<»
.8
1 E E
(/) CM 01
< O U-
I
u>
«n
V
> .
r>-
r i
t- O
rH CtJ
PH &
t/1 O
0) O
•H
10 U
I
« a
6 o
•H
•H
tL,
77
-------
. <*>
«M
oa
.
ra
O O
S <
cd o
E E
Q UJ
£
5
E "g
o E
(D O
• H
f-l PQ
-------
79
-------
are u'w1 - 0, and u'w1 = auaw. Because we expect u'w' to be non-zero, we
have chosen the latter limit. This implicitly assumes that a portion of the
measured crw arises from horizontal fluctuations being turned upslope
(downslope) by the hill surface. Note that w is corrected for prop response
on the basis of the wind tunnel data presented in the CTMD Second Milestone
Report.
Tower F is the truest "windward" tower on CCB for Experiment 201. The
vertical turbulence intensity, iz, at 2 m varies between 0.01 and 0.03
(1.0 and 3.0%), with one peak as great at 0.06. The Tower A iz data
interpolated to 30 m varies between 0.01 and 0.06. In general, however, the
turbulence intensity at 2 m on the windward face of CCB is considerably less
than that at 30 m on Tower A, and it is substantially less than that at 2 m
on Tower A, which averages about 0.06. Consequently, these data do not show
an increase in vertical turbulence intensity on the upwind slope.
Of further interest is a peak in the Tower F data record at 2100 MST.
This is about 10 to 15 minutes after a peak in the Tower A data. This is
further evidence of the turbulence shift seen in comparing the plume photo-
graphs with Tower A turbulence data, as reported in subsection 3.3.2.
The Tower E (at a windward location for much of Experiment 201) ver-
tical turbulence data at 2 m are similar in magnitude to the Tower F data.
However, these data exhibit a greater mean value and less variability.
Still, the turbulence intensity and
-------
In conclusion, the vertical turbulence data over CCB during Experi-
ment 201 indicate little, if any, increases in plume diffusion along the
upwind face of CCB. Increased mixing could occur near the surface of the
crest of CCB, especially during those periods when the vertical mixing away
from the hill was relatively weak, and increased mixing would certainly
occur in the lee close to the surface.
Experiment 210 '
The oil-fog plume was released on the southeast side of CCB for the
duration of Experiment 210. Therefore, Tower D is primarily a windward
tower; Tower F is a leeward tower; and Towers C and E are along the side of
the hill. Vertical turbulence data from these towers and from Tower B atop
the south peak of CCB are compared with data from Tower A, estimated for the
plume release height, in Figure 34.
The vertical turbulence intensity (iz) from the 2-m height on Tower D
is generally slightly greater than that at the release height on Tower A,
although there are periods when the fluctuations are out of phase. The
intensity at D is approximately 2% (+1%) after 0200. Data from 10 m on
Tower D are similar.
In the wake (Tower F), there is a greater difference between the 2-m
and 10-m data, and also between these data and the turbulence intensity
interpolated to release height with Tower A data. Vertical turbulence
intensity at 2 m is 2 to 3 times that at 10 m after 0200, and both are
several times the turbulence intensity at Tower A after 0200. During the
first hour, however, the Tower A turbulence exceeds that at Tower F.
Over the top of the south peak of CCB, the turbulence at 2 m exceeds
that at Tower A nearly all the time, and the turbulence at 10 m, although
less that that at 2 m, also frequently exceeds that at Tower A during the
shortened 3-hour record. This suggests that the estimates of turbulence
intensity from Tower D may be underestimated by setting u1 w1 = CTUOW.
Consequently, the turbulence data from Experiment 210 indicate that turbu-
lence levels may increase over much of the hill, causing increased plume
diffusion.
3.3.6 Implications for Modeling
The preceding analyses show that some degree of modeling error could
arise in simulating Experiments 201 and 210 due to using inappropriate
meteorological data as model input. To test the significance of this pos-
sibility, CTDM(11083), CTDM(ll083)-5, CTDM(11083-E), and CTDM(11083-E)-5
have been run on a new version of the data for the SF6 release height (at
or near the oil-fog smoke release height).
The following changes have been made to the MDA in preparing a modified
MDA for Experiment 201:
• 1705-1800 - All data are unchanged from the MDA.
81
-------
82
-------
83
-------
84
-------
1805-1900 - Best-fit iz data are substituted when available, and
a lower bound of 0.020 (2.0%) is placed on the MDA
values, reflecting the lack of any best-fit iz data
near or below 0.020.
2005-2020 - Photo wind directions are substituted when available,
and iz is set equal to 0.035 as a representative
lower bound.
2025-2030 - Photo wind directions and best-fit i_ values are
£1
substituted, and all other meteorology is persisted
from 2020.
2035-2200 - Photo wind directions and best-fit iz values are
substituted when available, and all other meteorology
is shifted by 10 minutes (e.g. 2025 data are used at
2035). iz is restricted to be no smaller than 0.020.
2205-2300 - Best-fit iz values are substituted when available,
and all other meteorology is unchanged except that
iz is restricted to be no smaller than 0.020.
Note that whenever the best-fit iz exceeds 0.080, its value is
reduced by 25%. This is done because the photo-derived az values at
these times are thought to be overestimates, either due to horizontal
meander or the approximations required to apply Gifford's technique to these
data.
The following changes have been made to the MDA in preparing a modified
MDA for Experiment 210:
• 0000-0020 - All data are unchanged from the MDA.
• 0025-0100 - Photo wind directions are substituted.
0305-0800 -
0105-0300 - Photo wind directions and best-fit iz data are
substituted when available; the missing wind direc-
tion data are filled-in with data from 10 m on Tower
B, and iz from the MDA are restricted to be no
smaller than 0.015, as reflected in the lowest best-
fit estimated for data from Experiment 210.
Same as for the period 0105-0300 except that the
average of the archive wind direction and the 30-m
Tower B wind direction are used to fill gaps in the
photo wind directions.
Experiment 201, Hour 18
The peak observed concentration is 458 ppt, and this occurred on the
windward face of the hill at release height (30 m). The wind speed varied
from between 6.0 and 7.4 m/sec during the hour, and the dividing-streamline
height was no greater than 2 m (essentially zero).
35
-------
Because the meteorology is not altered for this hour, the modeled esti-
mates are unchanged. CTDM(11083) estimates a peak concentration of 178 ppt,
and CTDM(11083-E) estimates 169 ppt. Both estimates place the peak value at
the location of the observed peak concentration. The 5-minute version of
CTDM(11083-E) estimates a peak concentration of 173 ppt, located high on the
windward face.
The observed concentration is underestimated because the plume size is
"too large," causing too great a dilution. A second calculation made on the
basis of films taken from the peak of CCB and aerial photographs fares some-
what better. The size of the instantaneous core of the plume is estimated
from the aerial photos; iy is estimated by assuming a top-hat distribu-
tion; the plume core is assumed to be circular; both ay and az are
assumed to grow linearly with distance; the center of the plume is allowed
to touch the surface; and the frequency of this impact at a receptor near
the camera location on the north peak is estimated from the film. The
resulting estimate lies between 380 ppt and 440 ppt, while the observed con-
centration is 465 ppt. This exercise indicates that even the best represen-
tative meteorological data measured at Tower A can lead to substantial
errors in estimated concentrations.
Experiment 201, Hour 19
The peak observed concentration is 401 ppt, located near the
windward peak of CCB. Peak modeled concentrations are:
• 34 ppt - CTDM(11083) with MDA, located just leeward of the crest
• 94 ppt - CTDM(11083) with modified MDA, located near the windward
peak
• 94 ppt - CTDM(ll083)-5 with modified MDA, located near the
windward peak
• 482 ppt - CTDM(11083-E) with MDA, located on the windward face
near the release elevation
• 605 ppt - CTDM(11083-E) with modified MDA, located on the
windward face near the release elevation
• 389 ppt - CTDM(ll083-E)-5 with modified MDA, located on the
windward face near the release elevation.
Using the best-fit iz data, which are generally greater than the MDA
values, leads to greater concentrations. In the case of CTDM(11083), this
improves both the location and magnitude of the peak value. For
CTDM(11083-E), however, this change causes increased overestimation. When
5-minute averages of the meteorological data are used, CTDM(11083-E)-5 per-
forms much better in estimating the peak concentration magnitude, although
the placement is too low on the windward face. Note that the second highest
observed concentration (363 ppt) is also located near the release elevation
on the windward face of the hill, so the model estimate of the location of
peak concentration is not as bad as it might first appear.
86
-------
Experiment 201, Hour 21
The peak observed concentration is 246 ppt, located slightly above the
release elevation. Peak modeled concentrations are:
• 900 ppt - CTDM(11083) with MDA
• 723 ppt - CTDM(11083) with modified MDA
• 689 ppt - CTDM(11083)-5 with modified MDA
• 797 ppt - CTDM(11083-E) with MDA
• 569 ppt - CTDM(11083-E) with modified MDA
• 851 ppt - CTDM(ll083-E)-5 with modified MDA.
All of these estimates are located midway up the windward face of the hill.
From the distribution of observed concentrations and the oil-fog plume
photographs, it appears that the plume preferentially curved around one side
of the hill. The band of greatest observed concentrations lies along
approximately the same elevation as the peak modeled concentration. The
hourly mean Hc is 23 m, or 7 m less than the release height. This sug-
gests that the plume is undergoing a greater horizontal deflection than what
is allowed by the Lift module.
The modifications made to the MDA produce a significant improvement in
the model estimates, especially in CTDM(11083-E). Perhaps an altered ver-
sion of the model which shifts more of the plume to the side of the hill
will reduce the size of the peak modeled concentration as well as improve
its location.
Experiment 201, Hour 22
The peak observed concentration is 437 ppt, located slightly above the
release elevation (the same receptor as last hour). Peak modeled concentra-
tions are:
• 1595 ppt - CTDM(11083) with MDA
« 1064 ppt - CTDM(11083) with modified MDA
e 1144 ppt - CTDM(ll083)-5 with modified MDA
• 1081 ppt - CTDM(11083-E) with MDA
• 605 ppt - CTDM(11083-E) with modified MDA
« 963 ppt - GTDM(ll083-E)-5 with modified MDA.
All of the modeled peak concentrations are located near the top of the wind-
ward face. The distribution of concentrations and the meteorology are quite
similar to that from the preceding hour. The dividing-streamline height is
now computed to be 26 m, or only 4 m below the release height.
The modified MDA again produces a substantial improvement in the peak
estimated concentrations. As with hour 21 however, more horizontal deflec-
tion seems to be needed. Also, the shift in Hc closer to the release
elevation has increased both the observed and the modeled concentrations.
This sensitivity to Hc implies that the resolution in calculating and
incorporating HC could account for much of the remaining discrepancy in
the peak concentrations.
87
-------
Experiment 201, Hour 23
The peak observed concentration is 304 ppt, located at the windward
base of the hill. Peak modeled concentrations are:
• 923 ppt - CTDM(11083) with MDA
• 695 ppt - CTDM(11083) with modified MDA
• 430 ppt - CTDM(11083)-5 with modified MDA, located 10 m
above the observed peak
• 405 ppt - CTDM(11083-E) with MDA
• 315 ppt - CTDM(11083-E) with modified MDA
• 433 ppt - CTDM(11083-E)-5 with modified MDA.
All of these estimates, except that for GTDM (11083)-5, occur on the same
face of the hill as the peak observed concentration, but at an elevation of
approximately half the hill height. The mean dividing-streamline height is
now I m above the release height, but the range in H,, over the hour
extends from 24 m to 40 m. Even with such a range in an important modeling
parameter, the inability of the model to produce the peak observed concen-
tration well below Hc is probably real. It may be that the Wrap module
needs to be re-evaluated, or that additional dynamics (e.g., some type of
upwind vortex flow) need to be introduced to bring more material to the sur-
face at the base of the hill.
Experiment 210, Hour 1
The peak observed concentration is 28 ppt, and few others are appreci-
ably greater than zero. The peak occurs at the upwind base of the hill, but
the second highest concentration (26 ppt) is found low, off on the side of
the hill ("side" of the hill as opposed to either the leeward or windward
faces of the hill). The peak modeled concentrations are:
163 ppt - CTDM(11083) with MDA
175 ppt - CTDM(11083) with modified MDA
145 ppt - CTDM(ll083)-5 with modified MDA
56 ppt - CTDM(11083-E) with MDA
59 ppt - CTDM(11083-E) with modified MDA
59 ppt - CTDM(ll083-E)-5 with modified MDA.
All of these occur on the side of the hill above Hc, which exceeds the
30-m release height by 12 m. It appears that the Wrap module does not pro-
ject enough material along the stagnation streamline, and that the Lift
module allows too much of the material away from the stagnation streamline
trajectory to impact the hill. Changes to the MDA only modify the wind
directions, and these appear to have a small impact on the modeled concen-
trations.
Experiment 210, Hour 2
The peak observed concentration is 395 ppt, located well below the
release height on the south side of the hill. With a release height of 57 m
and an Hc of 35 m, az must be large and the frequency of winds along
88
-------
the stagnation streamline must be high for the Wrap Module to produce so
large a concentration. Peak modeled concentrations are:
• 45 ppt - CTDM(11083) with MDA, located at the center of
the hill
• 226 ppt - CTDM(11083) with modified MDA, located high on the lee
side
• 146 ppt - CTDM(ll083)-5 with modified MDA, located near 35 m on
the windward face
• 162 ppt - CTDM(11083-E.) with MDA, located high on the windward face
• 386 ppt - CTDM(11083-E) with modified MDA, located on the south
peak
• 311 ppt - CTDM(ll083-E)-5 with modified MDA, located near the
release height on the windward face.
The new wind directions and vertical turbulence data make a substantial dif-
ference in model performance. Although the location of the peak concentra-
tion is not close to the location of the peak observed, CTDM(11083-E) comes
close to the observed magnitude. With 5-minute input data, CTDM(11083-E)-5
underestimates the peak value by a larger margin, but the location is im-
proved considerably.
Experiment 210, Hour 3
The peak observed concentration is 297 ppt, located at the base of the
hill in the lee. The peak modeled concentrations are:
*
9
0 ppt - CTDM(11083) with MDA
2 ppt - CTDM(11083) with modified MDA
42 ppt - CTDM(H083)-5 with modified MDA, located low on
the lee side
• 16 ppt - CTDMU1083-E) with MDA, located midway up on the
side of CCB
• 148 ppt - CTDM(11083-E) with modified MDA, located high on
the lee side
• 166 ppt - CTDM(11083-E)-5 with modified MDA, located at the
crest.
Once again, the modifications to the input data are extremely important in
modeling this hour, but equally important is the "enhancement" in
CTDM(11083-E). Together, the observed concentration on the crest of the
hill is virtually matched, and the concentrations high on the lee side are
approached. What seems to be missing in CTDM(11083-E) is the observed
streamline depression in the lee caused by lee waves.
Experiment 210, Hour 4
The peak observed concentration is 132 ppt, located high on the lee
side of the hill. Peak modeled concentrations are:
• 1 ppt - CTDM(11083) with MDA
• 0 ppt - CTDM(11083) with modified MDA
« 3 ppt - CTDM(ll083)-5 with modified MDA
89
-------
• 74 ppt - CTDM(11083-E) with MDA, located high on the lee side of
CCB
• 21 ppt - CTDM(11083-E) with modified MDA, located high on the lee
side of CCB
• 37 ppt - CTDM(11083-E)-5 with modified MDA, located high on the
windward face.
It is not at all clear from these results that the modified input data are
any more representative. CTDM(11083-E) seems to work better with the MDA.
However, a modified enhancement algorithm that would push the peak enhance-
ment leeward of the crest might improve the results of CTDM(11083-E)-5.
Because photo-derived wind directions are available for less than half of
the hour, and because no photo-derived i? data are available, the modifi-
cations made to the MDA are largely persistence-guided guesses.
Experiment 210, Hour 6
The peak observed concentration is 134 ppt, located midway-up on the
lee side of the hill. Two samplers on this part of the hill actually
recorded 134 ppt, but the next highest concentration is 67 ppt located on
the south peak. Peak modeled concentrations are:
• 0 ppt - CTDM(11083) with MDA, and CTDM(11083), CTDM(11083)-5 with
modified MDA
• 10 ppt - CTDM(11083-E) with MDA, located half-way up on the lee
side
• 7 ppt - CTDM(11083-E) with modified MDA, located at the same
receptor
• 15 ppt - CTDM(ll083-E)-5 with modified MDA, located at the same
receptor.
Little can be said about why the modeling is so inaccurate at this time.
More study is needed to understand the evolution of the observed concentra-
tions.
Experiment 210, Hour 7
The peak observed concentration is 212 ppt, located just leeward of the
south peak. Peak modeled concentrations are:
• 69 ppt - CTDM(11083) with MDA, located at the center of CCB
• 18 ppt - CTDM(11083) with modified MDA, located at the center of
CCB
• 45 ppt - CTDM(ll083)-5 with modified MDA, located just leeward of
the crest
• 166 ppt - CTDM(11083-E) with MDA, located high on the windward face
• 151 ppt - CTDM(11083-E) with modified MDA, located high on the
windward face
• 195 ppt - CTDM(ll083-E)-5 with modified MDA, located high on the
windward face.
The revised input data have a relatively minor impact on the modeled
concentrations compared to the enhancement of CTDM(11083-E). It appears
90
-------
from the distribution of concentrations that CTDM(11083-E) could do better
with a modified enhancement factor algorithm that would shift the primary
impact area to the lee side.
Experiment 210, Hour 8
The peak observed concentration is 384 ppt, located near the base of
the hill in the lee. The peak modeled concentrations are:
« 12 ppt - CTDM(11083) with MDA, located near the observed peak
• 5 ppt - CTDM(11083) with modified MDA, located near the observed
peak
• 98 ppt - CTDM(11083)-5 with modified MDA, located near the
observed peak
• 114 ppt - CTDM(11083-E) with MDA, located low on the side of the
hill
• 112 ppt - CTDM(11083-E) with modified MDA, located near the
observed peak
* 232 ppt - CTDM(11083-E)-5 with modified MDA, located low on the
side of the hill.
The modified input data have the greatest impact on the performance of
CTDM(ll083-E)-5. The peak is now within a factor of 2 of the peak observed
concentration, but the location is shifting away from the observed location
in the lee. It is difficult to judge what types of model modifications
might improve these results because no photo-derived estimates of iz or
wind direction are available. Modifications made to the MDA consist of
averaging the MDA wind directions with those from the 30-m level of Tower B.
This is based on the correspondence of these data with photo—derived wind
directions during previous hours.
Summary
Some mechanism for increasing concentrations estimated by Lift over the
hill appears to be necessary. The CTZ enhancement algorithm in CTDM(11083-E)
apparently improves the model performance, but a more carefully structured
algorithm is needed to better resolve when and where an "enhancement" should
occur. For the present, CTDM(11083-E) can be used to evaluate the impact of
modifying the input data for Experiments 201 and 210.
The ratio of the peak observed concentration to the peak modeled con-
centration (CTDM(11083-E)) is plotted in Figure 35 to compare the two sets
of input data. The results of using the actual 5-minute sequence of data
(CTDM(ll083-E)-5) are also included to evaluate the importance of simulating
the coupled variability in wind speed, Hc, and turbulence intensity. Out
of this set of 12 experiment-hours, 2 hours show no significant change (one
of these had no modifications to the MDA), and 1 hour shows a worse model
performance when the best of either the 1-hour or the 5-minute version of
CTDM(11083-E) with the modified MDA data is compared to the 1-hour version
of CTDM(11083-E) with the MDA. Therefore, the overall model performance
statistics will probably improve when the detailed evaluation of the mete-
orological information from all available sources is complete.
91
-------
.
•f
1
55
*9
uT 21uj
CO CO CO
§co oo
o o
555
a a a
T"
i
3
-i—r~
u5 V
o
•H O
O -H CS
O "^ ^
>- cd fn
CM CD <;
rQ Ctj H
Q (D
(!) - X
fn PJ
3 fH
t/)
-------
Model performance is not consistently improved by simulating each
5-minute period instead of simulating hourly concentrations with hourly-
averaged meteorological data. At times, the 5-minute simulation improves
the distribution of concentrations even though the peak estimate is de-
graded This'behavior may reflect an inadequacy in the model formulation,
and a subsequent improvement' in- the model could lead to even better model
performance'.1 ..:.>• .
Comparing mode-led- and observed concentrations with the two sets-of
input data has indicated areas in which CTDM should be changed to better
reflect the observed 'plume behaviour. Chief among these is the use of a
terrain-driven mechanism similar to that contained in CTDM(11083-E). Other
areas for model improvement are summarized in Table 3. Because most of the
hours studied in Experiments 201,-arid 210 are primarily "Lift" hours, we have
focused bur attention on Lift-related improvements. These are described in
the. next-subsection.- •-• • : .; .,. '
3.4 CTDM Upgrades: CTDM (1408 3)' (
The case-study results for Experiments 201 and 210 reported above
indicate that:
!•.... ,
• Equation 41 should be used to estimate az, with y ^ 0.5,
and a
zo
0. 5 m.
• Turbulence may increase within 10 m of the hill surface.
• The transition from Wrap,to Lift at Hc is too "severe," espe-
cially when the mean wind direction is not directed near the stag-
nation streamline.
• The effects of terrain on the path, distortion, and diffusion of
the plume over the hill must be explicitly formulated within Lift.
CTDM(11083) already incorporates the first finding, but CTDM must be reform-
ulated to incorporate the remaining three. At this stage, we will treat the
increase in near-surface turbulence as an implicit factor in the plume dis-
tortion and diffusion formulation.
This section contains a description of the new formulations for the
Lift module, an evaluation of the model performance characteristics for
Experiments 201 and 210, and an overall evaluation of the relative model
performance when CTDM(14083) is applied to the MDA.
3.4.1 Lift/Wrap Transition Upgrade
CTDM(11083) treats all flow beneath Hc as two-dimensional, horizontal
flow and all flow above Hc as "neutral" flow. Because the Lift module
deforms the horizontal distribution of plume material above H as if the
flow were part of axial flow over a sphere, material just above HC travels
up and over the hill while that just below is advected round the side. This
produces a zone of zero concentrations between these two plume segments.
93
-------
TABLE 3.
Experiment-Hour
SUMMARY OF AREAS FOR IMPROVING CTDM (11083)
BASED ON ANALYSES OF EXPERIMENTS 201 and 210*
Area for Model Improvement
201-18
201-19
201-21
201-22
201-23
210-1
210-2
210-3
210-4
210-6
210-7
210-8
Non-Gaussian formulation
Unclear
More horizontal deflection in Lift near H
c
More horizontal deflection in Lift near HC
Wrap assumptions near base of hill
More horizontal deflection in Lift, less in
Wrap near H
More horizontal deflection in Lift, less in
Wrap near H
C
Better terrain effect adjustment in Lift with
lee wave and wake sensitivity
Better terrain effect adjustment in Lift with
lee wave and wake sensitivity
Unclear
Better terrain effect adjustment in Lift with
lee wave and wake sensitivity
Unclear
*The need for a terrain effect in Lift similar to that in
CTDM(11083-E) is seen in all hours.
94
-------
Figure 36 is a schematic illustration of how this abrupt transition
between the Lift and Wrap modules differs from what one would expect on the
basis of tow-tank studies (e.g., Snyder and Hunt 1983). When the bulk of
the plume material is near or below Hc, we would expect a significant
horizontal deflection of much of the material above HC. This deflection
would be greatest for plume segments to the side of the stagnation stream-
line. Material close to the stagnation streamline, but above Hc, would
experience relatively little horizontal deflection. The structure of
CTDM(11083) allows little horizontal deflection for all plume material above
Hc because the effective radius of curvature controls the deformation.
Stratification is not accounted for.
This departure of the modeled plume distribution from that which is
expected not only produces a qualitatively incorrect distribution of surface
concentrations, it also can lead to quantitative overestimates of the peak
concentration. Below Hc, the plume is physically deflected to the side of
the hill so that concentrations on the hill are solely determined by the
probability that plume material lies along the stagnation streamline. If
the plume spends little time near the stagnation streamline, the concentra-
tions on the surface are considerably less than those under the "center" of
the plume. Above HC, CTDM(11083) shifts the plume very little in the
horizontal so that the "center" of the plume rides over some portion of the
hill. For plumes close to Hc in height but away from the stagnation
streamline, the "center" of the plume would be expected to deflect mostly
around the side of the hill instead, like material below Hc. Hence,
CTDM(11083) would overestimate peak expected concentrations.
The effective radius .of curvature can be altered in the Lift module to
increase the horizontal displacement of streamlines with increasing strat-
ification. Recall that an offset D was defined to be the difference between
the effective radius of curvature, and the height of the hill above HC.
If D is nearly zero, the horizontal shift of plume material away from the
stagnation streamline would approach that of the Wrap module, but material
close to the stagnation streamline would still rise along the surface over
the hilltop. The parameter D therefore provides a convenient mechanism
within Lift to smooth the transition between the two module components when
the release height is close to Hc.
Figure 37 illustrates the geometry. The example shown in the figure
illustrates how the method shifts the plume material farther from the hill
center as D1 becomes much less than D. Vertical profiles of plume material
are weighted by the fraction of time that the wind blows along the effective
direction to the receptor, rather than the actual direction to the receptor
from the source. Therefore the PDF for small angles from the stagnation
wind direction controls the magnitude of concentrations over most of the
hill.
This geometric method for shifting plume material is just a convenient
mechanism to obtain the desired results. The central assumption for
actually controlling the degree of redistribution is the parameter D1.
Because we wish to retain the basic structure of CTDM for the time being, D'
has been designed to respond to the release height of the plume relative to
Hc. If the plume is released below Hc, then we set D1 equal to some
95
-------
O
CO
CO
00
o
S B
O ^
•H O
•P <4-l
n!
(D
4-1 I
o to
•P O Q
tn t-<
3 -H CJ
cd ,0
t>o
96
-------
0)
O
si
• H
4)
^
•P
Oj
rH
3
O
Tj f>
CD cd
t/5
3 r-i
Oj
X-H
^ M
•P 0
0 4J
s S
o e
+->
o o
c c
o o
10
-H
C
O
97
-------
minimum value so that all plume material above H,, at the impingement
circle is distributed over the hill surface by streamlines close to the
stagnation streamline, regardless of how large crz is. If the plume is
released above Hc, D1 is set to some value between D and the minimum value
of D1, depending on the ratio of the release height to Hc. Because we
have no theory or physical modeling results as yet to formulate an expres-
sion for D', we have chosen a simple weighting expression:
D1 = D1 . + [1 - (H /z ) ]D
mm c r
for
z >
r
(44)
This expression has the desired limits when zr is equal to Hc and much
greater than Hc, and it shifts to the neutral flow limit more rapidly than
a linear interpolation formula.
3.4.2 Terrain Effects Upgrade
CTDM(11083-E) incorporates a measure of the effects of terrain by
increasing the size of az over the hill. This enhancement is viewed as
a combination of streamline distortion, diffusion, and increased turbulence
over the hill. The enhancement factor itself was taken to be the inverse of
the "terrain correction" factor used in the COMPLEX models. A better
formulation is needed to relate the enhancement to the individual effects of
streamtube contraction, speed changes, increased mixing, and lee wave
effects.
The upgrade to the method of simulating various terrain effects is
included in the Lift module. As discussed in Section 3.1, the plume is
split where the dividing-streamline of the flow meets the hill (so). If
the hill surface were flat beyond so, the surface concentration would be
2Q(s ,z) P(9)
(45)
Ho
c
where
a-2 - ar2(s) - 0,,2(sri)
«5 £t fj \J
Q(so,z) is the vertical profile of mass flux at s0:
Q(s0,z) = C0(s0,z)U0 .
The vertical profile of concentration is given by Equation 24.
(46)
(47)
If the terrain is not flat beyond so, the flow adjusts to the ter-
rain, which changes the closeness of approach of streamlines to the hill
surface. The terrain also changes the speed of the flow, the turbulence in
the flow, and the trajectory of the plume over the hill surface. Assuming
that the trajectory adjustments are already accounted for as in subsection
3.4.1, denote the net effect of plume height changes between so and s in
98
-------
Equation 45 by T^z-H,,). Denote the net effect of travel time, stream-
tube contraction, and turbulence changes by Taaz', and denote the
net effect of flow speed changes, by TUUQ. The diffusion of each
elemental plume segment' in the vertical profile at so is modeled as a
point source released at a net height of T^(z-Hc), into a flow of speed
Tuuo, with a diffusion rate characterized by Taaz' over the dis-
tance between so and s:
C(s,9,0) = 2
_., •> . C (s ,z)u
p(e) / oo o
s H /2iF T a ' I U
c a z u o
Substituting for C0(s0,z),
C(s,9,0) =
dz
(48)
2P(e)
s
Q l
2fra U T T a
CO
.
/Th(z-Hc)\»
\T /2a 'I-
e \a z>
Iz -z
1 r
l/Ta
_e\ z
zo o a u z H
z +z
r
+ e
ZOi
dz
(49)
^ /
If the flow behaved as purely two-dimensional potential flow, then
and Tu would be inversely related: TU = l/Th. Because the flow is
unlikely to behave as purely two-dimensional potential flow, set
TU " a/Th
T 5 Th/Ta
Then the solution to Equation 49 is
C(s3e,0) =
p(e)
s
1
1
4- a
QT
/lirU ctff "
0 Z
/T(z +H )\ 2
r c \
/To " )
V z /
/T(z -H )\2
1 r c
l/2a " 1 z -H a '
r \ Z /(T+arf * C Z \
Lc (1+crf /2a a" }
zo z
z +H o '
n - P r c z ^
a '
zo z
where
(50)
(51)
a I2 = a 2(s) -
z z
zo
a "2 = a 2(s) - (l-T2)a 2
Z Z ZO
(52)
99
-------
2 —
(s ) .
zo z o
Equation 51 takes on a more familiar form when HC is set to zero:
2
Tz
rf « n n) - 2P(6)
C(s,e,0) -- —
(53)
o z
The terrain modifications enter through two factors: a and T. T appears
only in combination with trz". From Equation 52,
(54)
+ a 2
zo
so that the factor T is seen to modify the rate of growth of az between
s0 and s, much like, the az enhancement factor of CTDM(11083-E). If
o is set to 1.0 as in the two-dimensional flow limit, then the terrain
effect is explicitly a relative increase in dispersion, presumably related
to the combined effects of streamtube distortion (enhanced concentration
gradients) and increased turbulence. If a is set to T, then the net
"centerline" concentration in the plume remains unaltered by the terrain, so
that the terrain effect is explicitly a displacement of the plume toward the
surface.
The dependence of a and T on the parameters controlling the flow is
not known for certain. Approximate expressions for the vertical deflection
of streamlines in uniform weakly stratified flows (Frjj>l, but Frj
-------
is reached, we assume that T remains at this value due to a combination of
streamline adjustments (Smith 1980) and wake turbulence.
is defined to be ,
H-H
(55)
where Li/2^Hc^ ^s the half-length of the hill midway between HC and •>
the top of the hill, H. FrH is the Froude number based on the height of
the hill, calculated for the flow above HC. The distance from the
impingement point (or the base of the hill if HC = 0) to the greatest ter-
rain elevation along the wind trajectory from the source to the effective
receptor is denoted as LQ. The distance from the impingement point to the
point where T first attains its minimum value for this wind direction is
-2.3 Fr
L = L /(I - e
o
(56)
When Fr-r is much greater than 1,
L = L,
'o'
When FrL is equal to 0.3,
L = 2L . The coefficient of FrL contained in Equation 56 is chosen to
place the minimum value of T at the leeward base of the hill (H = Hc) when
attains its minimum expected value (FrL = 0.3). For receptors at
distances between s0 and L, T is obtained by linear interpolation:
1 - T
T = T . +
mm
mm
(s + L - s)
(57)
sol
s 1 so
+ L
where s is the distance to the receptor, and so is the distance to the
impingement point (or the base of the hill if Hc = 0). T is set equal to
1.0 for receptors with s less than so, and T is held at
tors beyond so + L.
mm
for recep-
The value of T^.^ in Equation 57 should be different for a plume seg-
ment that passes toward the side of a hill than for a plume segment that
passes directly over the peak of a hill. To include this effect, let
equal To over the peak of a hill, and
mm
min
- T0) (y/a(Hc))2
(58)
if
a(Hc)
where y is the cross-wind position of the receptor in a coordinate system
with x aligned with the mean wind direction, and a(Hc) is the approximate
radius of the hill at an elevation equal to HC.
Because T/ct is a measure of the amount of dilution experienced by the
modeled plume as T becomes different from 1.0, we would expect a to be
approximately equal to 1.0 when the plume is very close to the surface com-
pared to 0Z, and we would expect a to be approximately equal to T when
101
-------
the plume is far from the surface compared to az. When the plume is
close to the surface (i.e., close to Hc), much of the decrease in T will
probably result from increased diffusion because the flow distortion places
much of the plume within a region of increased turbulence. When the plume
is well above the surface, the convergence of streamlines towards the
surface will place a good deal more of the "tail" of the plume close to the
surface and within the zone of increased turbulence, but the majority of the
plume will remain above and will not be subject to a marked increase in
mixing.
These concepts have been tested by running the new version of the model
(CTDM(14083)) on the data for Experiments 201 and 210, with a set equal to
T. CTDM(14083) incorporates Equations. 44 and 51, and TQ has been set to
0.5. By using a. similar enhancement scale, the new formulation of the Lift
model can be compared with CTDM(11083-E). Because a appears outside the
exponential functions, a comparison of the observed and modeled concentra-
tion patterns will indicate how a should vary. This comparison has been
made with the use of the modified MDA.
The ratio of the net release height to the size of az at the
impingement point ((zr - Hc)/azo) is found to order the behavior of a, and
our expectation of how a should vary is confirmed :
• a -\. T when (zr - HC)/CTZO = 3.14, 3.91, 5.17, 8.50
• a *> I when (zr - Hc)/azo = -0.20, 0.33, 1.00, 2.00
Note that four of the hours are not pertinent to this evaluation of a.
The comparison also indicates that some plume dilution is needed immediately
on the windward face of the hill when (zr - Hc)/azo is less than 0(1.0),
that Equation 57 works well for T when (zr - Hc)/azo is greater than 0(1.0),
and that TQ = 0.5 is generally of the right magnitude when (zr - Hc)/azo
is less than 5.0. For (zr - Hc)/azo much greater than 5.0, T should be
controlled exclusively by the streamline depression. If we take the minimum
approach distance from potential flow theory for flow over a sphere, then
the lower bound for T0 should be approximately 0.3.
These findings have been incorporated into CTDM(14083). When
(zr - Hc)/crzo is less than 1.0, T has been set equal to TQ at all
receptors and a has been set equal to 1.0 to simulate dilution of plumes
very close to Hc as they are swept over the hill in a turbulent boundary
layer. Whenever (zr - HC)/CTZO is greater than 3.0, a has been set
equal to T to simulate the importance of plume displacement relative to
dilution. Linear interpolation is used for a when (zr - Hc)/azo
is between 2.0 and 3.0. And To varies linearly between 0.5 and 0.3 for
(zr - Hc)/azo between 5.0 and 10.0. TQ is 0.5 below 5.0, and 0.3
above 10.0.
3.4.3 CTDM(14083) Performance Evaluation
CTDM(14083) was applied to the 12 hours from Experiments 201 and 210
for which modified input data were developed. The results are displayed in
102
-------
Figure 38 in terms of the ratio of the peak observed concentration to the
peak modeled concentration, unpaired in space.
CTDM(14083) is compared with CTDM(11083-E) with both 1-hour and
5-minute data in the figure. When the best results for either the 1-hour or
5-minute version of the two models are compared, version 14083'does better
than in version 11083-E in 3 of the,hours; it does about as well as version
11083-E in 5 of the hours; and does worse in the remaining 4 hours. The
performance of CTDM(14083) tends to be worse than that of CTDM(11083-E) for
those hours when the peak concentration occurs on the windward face of the
hill not far above Hc. In these cases, the enhancement structure in ver-
sion 11083-E begins to increase the dilution of the plume dramatically (see
Equation 27), while the structure in version 14083 requires a significant
growth in az with distance before the enhancement can produce a similar
effect (see Equations 52 and 54). Several hours in Experiment 201 appear to
require this dilution. This deficiency in the new version of the Lift
module emphasizes the need to consider the effect of the hill on the growth
of oz upwind of the point where Hc "intersects" the hill. The tow-
tank experiments of Snyder and Hunt (1983) document significant vertical
motion in impinging streamlines below Hc, and observers during SHIS #1
frequently reported a more "diffuse" plume when the oil-fog plume was
directed toward the hill, below H,,. For releases well above H-., CTDM
(14083) generally does better in locating the peak concentration correctly,
even though the magnitude may not be estimated better.
CTDM(14083) has also been tested with all 153 hours of data in the
MDA. The results are listed in Table 4 along with those from CTDM
(11083-E). These results are similar to those contained in Figure 38 in
that no great improvement is evident in the performance of the new version.
The bias has moved marginally closer to the ideal, and the noise has
increased marginally. A slight improvement in the resolution measure is
also noted. These results indicate that a significant improvement in CTDM
will require a careful evaluation of the structure of a and T, the inclu-
sion of a framework for modifying flow and turbulence assumptions in the
Wrap module, and detailed evaluation of the data contained in the MDA on a
case-hour by case—hour basis.
103
-------
104
-------
w
Jfi
H
CO
0 O
H H
CO CO
M &
H
H ftS
CO I
en
w oo
0 0
5 c?
3
0 g
M O
s
Q
iJ E?
n
O '"»
S C""!
00
fj-j ^^
0
-------
SECTION 4
THE HOGBACK RIDGE EXPERIMENT
4.1 Geographic and Meteorological Setting
The Hogback Ridge (HER) was selected as the site for the second Small
Hill Impaction Study (SHIS #2). HBR is located in the northwestern corner
of New Mexico, about 15 miles west of Farmington. It lies on the semiarid
Colorado Plateau near the western slopes of the San Juan Mountains
(Figure 39). Three rivers—La Plata, Animas, and Chaco—drain into the San
Juan River near Farmington. The terrain features of the area include
occasional isolated ridges (e.g., HBR), isolated promontories (e.g.,
Shiprock), and low mesas and plateaus. The area is characterized by a
sparse vegetative cover of desert shrubs and grasses.
SHIS #2 was conducted in the environs of an approximately 1.5-km long
section of the ridge north of the San Juan River (Figure 40). Here, the
Hogback is oriented NNE-SSW and rises about 85 m above its base elevation.
HBR extends from just north of the SHIS #2 experimental area to about 8
miles south of the San Juan River. The ridge splits where the San Juan and
Chaco Rivers flow westward and forms separate "hogbacks." The two southern
sections, separated by the Chaco River, are shown in Figure 41 as viewed
with a telephoto lens from the SHIS #2 site; the plume from the Four Corners
plant is passing over the ridge. Waughan Arroyo is located just east
(upwind) of the experimental section. Farther east a series of irregular
mesas, arroyos, and surface coal mines extend all the way to Farmington.
Figure 42 presents a view of the experimental section of HBR from the east,
and Figure 43 shows the area east of the section, as viewed from the top of
HBR.
Because of substantial reserves of coal and the adequate water supply
from the San Juan River, two major electric generating facilities operate in
the region. The San Juan Power Plant is located just east of the experi-
mental area, and the Four Corners Power Plant is located south of the San
Juan River. Public Service Company of New Mexico (PNM) and Arizona Public
Service Company (APS), the respective operators of the two generating
stations, have sponsored several meteorological and air quality measurement
programs. The National Weather Service operates a weather station at the
Farmington (FMN) Airport. Consequently, there was a wealth of meteorologi-
cal data available during the site selection and experimental design phases.
106
-------
^r- 'jA«
'- "?i,!*?'* &-• - " "'^?OT^?^y^ v
jie** . - '"" .>> %& _ r^ %,^v -^
asiL
S^\""
TV-'- -N(i '• • -*•» !r*rru: P S^*^* %'^/&*WR*;
&•:
0)
t>0
•a
o
(D
•6
(S)
=tt=
en
4-1
O
f-i
oj
O
•H
M
CD
a
3.
•H
107
-------
Figure 40. The SHIS #2 site. Contour interval 20 m.
108
-------
§
c/}
0)
o
X!
3
o
CD
•H
•H
a.
109
-------
0
-------
•a
O
33
CD
•P
S
O
(D
1
4->
O
(D
i
O
•M
•H
to
0)
f-l
bO
•H
BH
111
-------
The meteorological measurements taken by PNM, APS and the National
Weather Service provide detailed information on the dispersion climatology
of the HER region. Most of the available information has been analyzed and
summarized recently by Moore et al. (1981). Figure 44 shows two wind roses
derived from FMN data. The first rose shows the distribution of winds
during 1976; the second shows the distribution of Oct-Dec winds for
stability "E" conditions. Figure 45 shows October hourly FMN wind roses for
four nighttime hours—2200, 0000, 0200, and 0400. Evidently, the wind
typically "turns around" (Crow 1975) from westerly to easterly during the
night as the drainage from the San Juan Mountains is established.
Drainage periods occur often during the summer and fall months and
produce stable easterly flows toward HBR at night. The drainage and
turnaround days are characterized by light winds all day. Cooling in the
evening and night results in strong surface inversions. Dense air flows
from the mountains down the river valleys and the Chaco wash. Stable air
flows toward HBR along the San Juan River valley, around Pinon Mesa from the
northeast and occasionally from the southeast. Figure 46 (taken from Moore
et al. 1981) illustrates a typical summer morning downslope drainage flow
situation. Figure 47 shows representative early morning upper air soundings
during easterly drainage situations. The deep stable layer, which is
uncoupled from the synoptic westerly winds aloft, continues well into the
morning until convective turbulence destroys the inversion, and the flow
near the ground is again coupled with the synoptic winds aloft.
In summary, the available meteorological data suggested that frequent
stable easterly winds occur at night during the fall months. This was the
principal reason for selecting HBR as the site for SHIS #2. Other reasons
included: (1) HBR is the dominant terrain feature in the area of interest;
(2) the area is easily accessible and has electric power available; (3) PNM
provided records of meteorological data taken on and near HBR and was
willing to provide the data in real-time during the experiment; and (4) the
Bureau of Land Management, which manages the HBR and the area to its east,
granted permission to use the ridge.
4.2 Preliminary Flow Visualization Experiment
A preliminary flow visualization study was conducted at a section of
HBR just south of the power transmission lines (see Figure 40) during the
period June 6 - June 11, 1982. Scientists from ERT, EPA/Terrain Effects
Branch (TEB), EPA/Fluid Modeling Facility (FMF), NOAA/Wave Propagation Lab
(WPL) and the Los Alamos Scientific Lab (LASL) participated.
Meteorological data were obtained using two tethersondes, pibals and
fixed meteorological sensors on three towers (two 10-m towers on the Hogback
and a 200-ft tower east of the arroyo) operated by PNM. Smoke was released
using smoke candles and canisters suspended from a 600-ft^ blimp at a
location approximately 150 m east-southeast of the road along the base of
ridge and 300 m south of the major transmission lines crossing the ridge.
112
-------
5.3
7.0
19.7
11.3
T=Q28-°
Figure 44(a). Climatological annual wind rose for Farmington, N.M.
(2927 valid data points, 3-hour NWS data for 1976).
2.8
2.0 2-0 2.8
12.2[
28.4
7.2
2.8
2.4
Percent Occurrence
0.0 8.0 16.0 24.0 32.0
,1-3 8-12,
4'7 13-18 19-24 Over24 (mph)
Figure 44(b). October-December stability E Farmington wind rose.
113
-------
Percent Occurrence
0.0 8.0 16.0 24.0 32.0
^"—^ 4'7 13-18 19-24 Over 24 (mph)
Faimington. New Mexico — Airport (FAA) October (1977-19811 Hourly Rose - Hour (LSI) = 22
Farmington. Now Mexico — Airport (FAA) October (1977-1981) Hourly Rose - Hour (LST) = 24
313,5
Formmgton. New Mexico — Airport (FAA| October |1977-1981) Hourly Rose - Hour (LST| = 02
Farmington, New Mexico — Airport (FAA) October (1977-1981) Hourly Rose - Hour (LST) = 04
Figure 45. October hourly wind roses for four nighttime
hours: 2200, 0000, 0200, and 0400 MST.
114
-------
090
i i *i i I i i i i i i i i i
710
720 730
ERSTING
740
750
020
Figure 46. Summer morning downslope drainage flow. (Taken from Moore et al,
1981.) The squares (ffl) indicate the San Juan and Four Corners
power plants. The SHIS #2 site is just west of San Juan at UTM
coordinates 722 E, 4074 N.
115
-------
111 Tl-t H H H H+H UlU U HiU IU III1111111111
I I I III I II I II I I I M
in o
uj 2^
a:
^ a
h- 36
tr *
a a: o
UJ
o
CM
CO CO <*• CM O 03 ID
«-1 4-1
O rt
O If)
O
-H
•H
to -d
c o
•H O
3 cti
o o
Ul -H
bO
H O
•H rH
nj O
^i
!-H O
CD CD
4)
^
M
•H
PH
116
-------
WPL operated a tetherson.de near the blimp to document meteorological
conditions near the source. LASL operated a tethersonde east of Waughan
Arroyo to document conditions upwind of the source. In addition, EPA
released pibals and neutrally buoyant ballons.
The flow visualization experiments were documented with 35 mm cameras
and 8 mm movie cameras. The meteorological data and the photographs were
analyzed to aid in the SHIS #2 experimental design. Specifically, the flow
visualization experiment helped in the following areas:
« selection of the section of HBR for SHIS #2,
• location of the roads for the release crane,
• location of the tracer gas sampler locations,
9 location of the meteorological towers and other meteorological
sensing devices, and
• location of the lidar.
In addition, photographs and tethersonde data were analyzed to test the
critical dividing-streamline height (Hc) concept for the Hogback. Table 5
summarizes three experiments during which plumes released above H went up
and over the ridge, and plumes below the HC stagnated and tended to remain
horizontal. The release height is denoted by zr.
4.3 Fluid Modeling in Support of the SHIS #2 Experimental Design
During September 1982, a series of wind tunnel and towing tank flow
visualization experiments were conducted at the EPA FMF to provide input to
the SHIS #2 experimental design. The studies were designed to investigate:
* plume height above the surface over the hill crest and at the
upwind edge of the hill,
• apparent size of any plume deformation upwind of the hill,
• lee wave importance and structure, and
• sensitivity of the plume trajectory to "wind angle."
This information was then used to (l) guide the design of the smoke and
tracer gas release protocols at HBR, and (2) help select sampler and camera
locations.
Two models of the Hogback were used—one for the tow-tank and a second
for the wind tunnel. Each has a scaled radius equivalent to 800 m, and the
northernmost point is the prominent peak just north of the high tension
power lines near PNM station 105. The ends of the models were extended to
the walls of the test chambers. Peak elevations above the base plate did
not exceed 14 cm (84 m, scaled). Each model was covered with approximately
3 mm size gravel.
Two tests were made in the wind tunnel. One test was done with the
ridge perpendicular to the flow, the other with the ridge rotated by 30°.
Because the flow was very turbulent (boundary layer about l~m thick) with
the 'trip fence in place, the fence was removed to gain a better measure of
plume centerline height.
117
-------
TABLE 5. PRELIMINARY VISUALIZATION EXPERIMENT
1.
11 June
z.
1982
. = 40
EXAMPLES OF Hc ANALYSES
and 60 m at 0539MDT
Hc = 37 m at 0456 (WPL #21 ascent)
10 m at 0528 (WPL #21 descent)
6 m at 0503 (LASL #35 ascent)
Observer comment: Smoke largely made it over from both release
heights.
Photos: Narrow plumes definitely above HC.
11 June 1982
zr = 60 m at 0620
Hc = 8 m at 0557 (WPL #22 ascent)
39 m at 0626 (WPL #22 descent)
1 m at 0603 (LASL #36 descent)
Photos: Narrow plume above Hc
10 June 1982
zr - 15 and 35 m at 0650
Hc = 30 m at 0635 (WPL #17 ascent)
19 m at 0650 (WPL #17 descent)
Observer comment: The 15-m level is very diffuse and the 35-ra
level maintains some integrity and it has some definite direction
to it. The 15-m level release is going right down to the ground
within a few tens of meters from the source. Smoke from the 35-m
release seems to be approaching the hill slope but getting up and
over. The 15-m smoke release is involved in all the hillocks at
the bottom of the Hogback but seems to be getting up and over the
ridge. It does not appear to be blocked.
Photos: Hr is between 15 and 35 m.
118
-------
In the first test, smoke was released 10 cm above the floor of the
tunnel. The centerline appeared to be about 9-11 cm above the surface of
the model near the base of the hill, and about 6 cm above the surface at the
crest (l cm = 6 m) . Therefore, the plume maintained its initial height
upwind of the hill and approached the crest surface at nearly half its
initial height. This is approximately consistent with what would be
expected in potential flow. The plume was forced to speed up over the crest
of the ridge, but no region of deformation could be seen upwind of the hill.
In the second test, the flow angle produced little additional effect on
streamline height but did produce a small plume path deflection. Although
the deflection was difficult to quantify, it was estimated to be about 8 cm
(about 1/2 the visible plume width). Because the smoke was released about
800 m from the crest (scaled), this deflection (^50 m, scaled) was less
than 4°.
Eight individual tests (Table 6) were run in the tow-tank. The first
three were run with Froude numbers of 0.8, 1.0, and 2.0 and with a wind
direction of 118° (wind flow perpendicular to the ridge). In the cases of
Frjj =0.8 and 1.0, tow speeds were set to produce Hc values of 0.2 H and
0.0 H. Consequently, bulk Froude numbers were probably less than 1.0 in
both tows, due to the mixed layer near the surface of the model.
Dye streaks were released at heights of 2, 6, 10, 14, 18, 22, 25, 30,
35, 40, and 45 cm at the upstream edge of the baseplate. These heights are
referenced to the water surface. The dye streamer released at 2 cm was
mixed in a stagnation region upstream of the ridge during the first tow.
Tows 4 and 5 had Frjj =1.0 and 2.0, respectively, based on the linear
density profile above the surface mixed layer (no mixed layer was present
during the first tow at Frji = 1.0). Dye release heights were switched to
2, 10, 18, 25, 35, and 45 cm. Dye released at 10 cm Ov60 m, scaled) rose
to 15 cm above the surface at the upstream edge of the Hogback face in both
tows and dropped back to 9 cm and 10 cm (Fr = 1.0, 2.0) at the crest except
near the end of the fourth tow (Figure 48) when the dye passed over the
crest at an elevation of about 3 cm. Table 7 summarizes the measured
heights of the'10-cm streamer for five experiments.
In experiments 6 and 7 the model was rotated 10° and retowed at
Frjj = 1.0, and 2.0. Streamline deflections at 2 cm were small in both
tows. A rough estimate puts it at 2-3°. A final experiment 8 was done with
= » (neutral conditions) .
The FMF flow visualization experiments suggested that during
unstratified conditions the streamline patterns are similar to those
expected from potential flow theory. The plume should approach HBR at its
initial elevation upwind of the ridge, and flow over the crest at nearly
half its height. During the weakly stratified simulated conditions, the
tow-tank experiments suggested a rise in streamlines near the upwind base of
the hill and a fall over the crest to near or slightly lower than its
119
-------
TABLE 6. FMF TOW-TANK EXPERIMENTS
Tow No.
HOG 1
HOG 2
HOG 3
HOG 4
HOG 5
HOG 6
HOG 7
HOG 8
Direction
118C
118'
118'
118°
118'
108C
108°
108°
0.8
1.0
2.0
1.0
2.0
1.0
2.0
Remarks
Slight shifting at lower levels
toward south, but all streamers
over ridge top. Strong
contraction of streamers observed
on lee side. Slow motion below
hill top.
More pronounced shift of lower
level streamers to south, but for
shorter time period. Wider
spreading on lee side. Rotor
observed at 4-5 hill heights on
lee side. Lower streamers upwind.
Straight narrow plumes. Less
contraction in lee.
Dye released at 2, 10, 18, 25,
35, 45 cm. Stack visibility
good. Fairly wide diffusion in
lower levels. Blockage noted
upstream.
Plumes somewhat wider. Distance
from crest to 1st max. in lee
wave ^ 4. 9 m.
Lower plumes twist or deviate to
south upstream consistently,
straighten out downstream.
Slight S-shape in trajectory.
More plume meandering. Wider
dispersion.
120
-------
TABLE 7. MEASURED HEIGHTS OF/THE 10-cm HIGH'STREAMER AT THE
UPWIND BASE AND CREST OF THE HER MODEL
Height (cm) at
Upstream Base
HOG 4 FrH =1.0
14
15
15
15
15
14
12
Height (cm) at
Crest
8.5
9
9
8
7
5
3
HOG 5 FrH= 2.0
15
15
15
15
15
9.5
10
11
10
9.5
HOG 6 FrH =1.0
12
14
14
14
15
8.
9
9
8
8
HOG 8 FrH =
Wind Tunnel
13
10
121
-------
a. Side view
b. Top view
"Flow"—i
Figure 48. FMF tow tank flow visualization experiment - Hogback Ridge
(wind direction = 118°, Fr = 1.0).
122
-------
Initial height.* Low-level releases near or below Hc are expected to
experience considerable mixing. Wind direction variability will not be
significantly amplified.
4.4 Experimental Design
The SHIS #2 at the Hogback was designed to obtain meteorological, flow
visualization, and tracer gas emission and concentration data in the
vicinity of a nearly two-dimensional ridge in order to enlarge the modeling
data base and to provide a good basis for testing, evaluating, and refining
the modeling concepts developed from the CCB data base and the various FMF
experiments. The experimental methods of the SHIS #2 were similar to those
used and tested at CCB. The experiment was conducted at the Hogback from
October 5 - October 29, 1982 and included:
• Releases of two tracer gases (SFg and Freon 13B1 (CF3Br)) and
oil-fog, using a mobile 150-ft crane and a tower as source
platforms;
• Fixed meteorological measurements:
- a 150-m tower instrumented at ten levels,
- a 30-m tower instrumented at five levels,
- a 10-m tower instrumented at three levels,
- a 60-m tower instrumented at two levels,**
two monostatic acoustic sounders,
a doppler acoustic sounder, and
- three optical crosswind anemometers;
• Two tethersondes:
- one operated at source elevation to document meteorological
conditions representative of the source, and
one operated to measure vertical profiles of meteorological
parameters upwind of the ridge;
• Ground-level tracer gas concentrations;
• Lidar measurements; and
• Photographs and videotapes.
*The initial rise (e.g., from 10 cm to 15 cm) near the base of the HER
model was not expected. Initial analyses of the SHIS #2 photographs
suggest the occurrence of this phenomenon in the field, although the
results are currently obfuscated by the thermal plume rise of the
oil-fog plume.
**Tower P was operated by PNM. PNM maintained temperature and
temperature difference sensors on the tower. ERT installed two
cup-and-vane sensors for SHIS #2. In addition, PNM operated a
network of instruments in the area, including two 10-m towers on the
Hogback. The PNM data were displayed in real-time at the ERT
command post.
123
-------
After three smoke visualization experiments, 11 combined tracer and
flow visualization experiments were performed. The three initial
experiments were conducted to (1) understand the autumn weather conditions
at HBR (the preliminary experiment was conducted in June), (2) gain
experience working at the HBR site, and (3) finalize the release and
sampling protocols. During the 11 tracer and simultaneous flow
visualization experiments, SFg and CF3Br were released, sampled and
analyzed for concentrations on the ridge. The meteorological data were
archived and displayed in real-time by a system of onsite minicomputers.
The real-time information on ambient meteorological conditions and the
flexibility of releasing oil-fog and tracer gases at a wide variety of
heights and locations allowed the field managers real-time control of the
experiment in the selection of source positions to obtain useful information
for model development purposes. The real-time meteorological feed-back was
supplemented by near real-time lidar observations and an approximately
48-hour turnaround time on the photographs. Figure 49 illustrates the
layout of the SHIS #2.
The SHIS #2 participants and their principal responsibilities were*:
• ERT
- overall CTMD project management
field management and control (selection of experiment times,
release locations, heights, etc.)
the 150-m and 60-m meteorological towers
- site logistics (power, roads, security, weather forecasts,
etc.)
- quality assurance
• NOAA ARLFRO
- tracer gas releases, sampling and analysis
- oil-fog generator
- the 10-m and 30-m towers
telemetry and meteorological data archive and display system
• NOAA WPL
- lidar
— two sonic anemometers and data acquisition system at the
150-m tower
- two monostatic acoustic sounders
- one doppler acoustic sounder
- three optical anemometers
- one tethersonde
• Morrison-Knudsen Company (under subcontract to ARLFRO)
- oil-fog plume photographs (five positions)
- video tapes of the plume
*The details of the SHIS #2 are presented in the "Work Plan for the
Small Hill Impaction Study No. 2," ERT Document P-B348-620,
September 1982.
124
-------
LEGEND
A 500' Tower A
A Tower B
O Tower C
1 Tracer Release Pt. No. R-80
2 Tracer Release Pt. No. 203
• 3 Tracer Release Pt. No. 21 5
* 4 Tracer Release Pt. No. 21 6
5 Tracer Release Pt. No. 111
Figure 49.
SHIS #2 field experiment layout.
125
-------
• NOAA ATDL (under agreement with ARLFRO)
- one tethersonde
video tapes of the plume during the first three experiments
• TRC Environmental Consultants (under subcontract to ERT)
- external quality assurance audit
4.4.1 Meteorological Measurements
Meteorological Towers
Four meteorological towers were operated during the SHIS #2. ERT
installed and operated the 150-m and 60-m towers. ARLFRO installed and
operated the 10-m and 30-m towers. Tables 8A through D list the instru-
ments, heights, and direct and derived measures* for the four towers.
Table 9 provides definitions of the measures. Approximately 242 5-minute
and 71 1-hour measures were archived in real-time. ARLFRO was responsible
for the data collection and display system (see Section 4.4.2). The 150-m
tower was also outfitted with a tracer gas and smoke release platform and
winch assembly to allow releases to an elevation of approximately 75 m.
NOAA WPL provided two sonic anemometer systems that measured u, v, w,
and temperature at a high frequency (20 Hz). The sonic systems were located
at the 40-m and 5-m levels of the 150-m tower. WPL also provided a separate
data logging system to archive the sonic data.
The purpose of Tower A was to characterize the approach flow and the
meteorological conditions at the release heights. Tower B (30 m) was
located west of the main road toward the base of the Hogback to document
changes in flows and turbulence intensities in this region of highly dis-
torted flow. Tower C (10 m) was located at the windward side of the top
of.the Hogback to document the wind, temperature, and turbulence field below
10 m over the hill crest. The 60-m Tower P was operated by PNM. ERT in-
stalled two wind sensors on it to help document the approach flow well
upwind of HBR. PNM also operated other meteorological instruments in the
area (Figure 50), data from which were available in real-time during SHIS #2
as half-hour averages.
Tethersondes
Two tethersondes were operated to measure temperature, horizontal wind
speed and direction, pressure, and humidity. One sonde was operated by ATDL
to characterize meteorological conditions within a few meters of the source.
A second sonde was operated by WPL to obtain vertical profiles east of the
ridge. The WPL sonde was operated near Tower B until October 15 and near
the doppler sounder through the end of the experiment. Data were recorded
*Measure is used here to indicate a direct measurement (e.g., the
u-coraponent of the wind) or a calculated parameter (e.g., turbulence
intensity).
126
-------
en
X X X X X X X X X
5 S
Crf CTl
H .-i
co \o
2: r-i
CQ
„
erf
cd
XXX
S I
I O
O H
00
s
CQ
,-^
g
^-^
u
00
•H
HI
33
en
CO
0.
o
Ca
s-^
n
(0
P.
3
O
"3 2
H <
0)
0)
VJ
3
CO
W
0)
S
co en H
H
1=1
OT CO M
en o
S vD
1-H i-H
<; ca
— CO
W *-s > 0 ° H
X - x" Pn H
O a co CQ EH H Q
X
X
%
CO
sT
CO
X
X
a
CO
g
CO
XXX
H H
3 co
127
-------
co
co
CO vO
D-,
CD
CD
00
CO /-v
E- J
2: to
E- co
CO \0
o
OS
CD
CJ
CO
CO
128
-------
TABLE 9. DEFINITIONS OF THE MEASURES
U, V, W Vector averaged wind components - props
UX, VX Vector averaged wind components - cups
S, D Scalar mean wind speed and direction - props
SX, DX Scalar mean wind speed and direction - cups
TF Fast Thermistor Temperature
T Slow RTD Temperature
DT, TC Slow RTD AT, and calculated T (T(2m)* + DT)
SU, SV, SW Turbulence Scales - sigma-u, sigraa-v, sigma-w
SD, SDX Sigma-theta (props), sigma-theta (cups)
UW, WT
ST
U'W', W'T'
Standard Deviation of TF
m/s
m/s
m/s, degree
ra/s, degree
°C
°C
°C
m/s
degree
2
(m/s), m/s °C
°C
* Or 1 m for the 30-m tower and 9.1 m for the 60-m tower.
129
-------
Figure 50. PNM air quality and meteorological monitoring sites,
130
-------
on digital cassette tape for processing. Real-time paper copies were
available to help guide the experiment.
Monostatic Acoustic Sounders
Two monostatic acoustic sounders provided data from which the vertical
profile of turbulence and stability of the planetary boundary layer above
the sounder were inferred. The two sounders were operated around the clock
and the data were recorded on facsimile charts. This information was useful
for experimental guidance by showing the development in time of the vertical
structure of the lower atmosphere, particularly the arrival of air masses
with different characteristics of temperature profile, shear, and turbulence.
One monostatic sounder was operated near Tower B upwind of the Hogback.
A second sounder was operated in the vicinity of the 150-m tower, about
800 m east of the ridge.
Doppler Acoustic Sounder
The doppler acoustic sounder measured the vertical profile of the
horizontal wind velocity in the planetary boundary layer above the sounder.
Vertical resolution was 15 m with adequate signals to a height of typically
1 km. The doppler acoustic sounder was installed about 2 km east of the
targeted segment of the ridge in the vicinity of the Shiprock Substation
(Figure 49).
Optical Crosswind Anemometers
An optical crosswind anemometer measures the path—averaged wind across
the line-of-sight path between a transmitter and receiver. Three were used
in the SHIS #2 (Figure 52).
4.4.2 Meteorological Data System (MDS)
NOAA ARLFRO provided a real-time Meteorological Data System (MDS) to
acquire, process, display, and store data. Figure 51 depicts the component
structure of the system. Operating continuously during each experiment, the
MDS sampled the 86 meteorological sensor inputs, calculated the derived
measures, and displayed selected parameters and profiles. Data from the
10-m, 30-ra and 60-m towers were transmitted by ARLFRO radio links. The
150-m tower data were transmitted by shielded signal cable.
Additional meteorological data were available from the PNM stations in
the area. The stations of particular relevance are located on top of the
Hogback; station 103 is about 5 km SW of the experiment area, and station
105 is just to the north of the experiment area. The PNM data were available
as half-hour averages, which were intercepted from the PNM RF communications
links at the 60-m tower and telemetered by ARLFRO's radio link to the
command post near the 150-m tower, where they were displayed on a line
printer at each half-hour scan.
131
-------
CO
ft
S o
W 0
1 g
10 10
JT3
c
•s **
5 n
g- ^X
ra -5
2 §
S «
D j=
CJ to
Z 0>
i 1
•g I"
*S i- •
to o O
c w
E o 15
••"S
w
CD
D>
2
§
0
c
o
1
O
O)
11
0 8
2 g
1? 3
S W
1 1
S S
to iH.
' o
* & 8
S 2 I
M w t;
S o 2
i s "
•la"
E 2 .=
If) W
UJ
•?
In
1
1
Q.
Q.
0
U.
CM
^r
*£
«£
^•^
C
jg
U)
5
CO
fx
N.
9
5
0)
M
00
z .2 c
Q. \ co CM r
1 DC =™
1 1 S .H
J ' » ^ g1
1 1H
£ * 2
ca .a c
>. s f
JS <= .2
.2 « S
/^\"\\ sx "O — j §•
/ CQ \ \\ W ro ' ' ' to
2 1 U 5 % § s
O I II "^ .. a> o
, 00 / // != co " •£
\3/ J/ .g 5^3
,s 1 1
O
— 1- O 'a.
O5 CO 0
S»
^Q
IO -o
X *-
* CO
m I
sand indices
tter diagram
•^ o
to w
(0 O
i? s
03 'IT
"5. S
•o -o
-5 §
^ »
0) W
« o
li 0.
1
^
" \
Q. \
0 \
° 1
X
to
G
O
•H
•H
•H
Cti
Cd
4-J
cd
bO
O
O
M
O
tu
S
bO
132
-------
4.4.3 Tracer Gas Release, Sampling and Analysis
Tracer Release System
Tracer releases were made from the 150-m tower and from locations on
the three roads east of HBR. The roads were culverted, graded and graveled
to support the 150-ft crane that lifted the oil-fog generator and the tracer
gas tubes. Two tracer gases, SFg and CF3Br, were released at different
heights from the tower or the boom of the mobile crane. The SFg release
was co-located with the oil-fog release and dispersed from a common nozzle.
Occasionally, the SFg and oil-fog were released from the tower while the
CF-jBr was released from the crane at a different location.
The SFg and CF-^Br tracer gases were stored in individual compressed
gas cylinders at ground level. Piping carried each gas through a linear
mass flow meter (LFM) system to the point of discharge into the atmosphere.
A time history of each tracer release was used to describe the rate and
quantity of release of tracer. The LFM measured and displayed the rate of
gaseous tracer discharge via real-time digital display, the total amount of
gas discharged via a digital counter, and the analog output voltage directly
proportional to the flow rate. The voltage was logged and monitored on a
strip chart recorder. Pre- and post—test release weights of gas tracer
cylinders were measured by certified scales. Beginning and ending times of
tracer release and the time and character of any deviations from the design
rate-of-release were logged.
Tracer Sampling
Tracer samples were collected in 2-liter Tedlar bags at about 110
locations on the ridge. ARLFRO operated 125 samplers during each approxi-
mately 8-hour experiment. Twenty samplers were used to get 10-minute
averages at five locations (4 samplers x 12 bags = 48 10-minute samples for
each of the five locations). The remaining 105 were used to get 1-hour
samples. Two of the 1-hour samplers were operated on the 30-m tower and one
on the 10-m tower.
The bag samples were collected by means of modified EMI AQSIII or
similar type of air sampler. Each sampler used 12 separate pumps, bags,
and external tubes to draw in ambient air to fill the individual 2-liter
Tedlar bags. The system was battery powered and electronically programmed
in function and timing. Time was set and maintained by a crystal-controlled
digital clock accurate to within 1 minute per month. Beginning and ending
sampling times for the individual (sequential) whole-air samples were
controlled by this clock. The actual local time (MDT) for beginning of the
sampling sequence (in each unit) was preprogrammed during servicing by
sampling team technicians within about 20 hours prior to the start of each
experiment.
The sampler locations (Figure 52) were selected by ERT scientists on
the basis of the June preliminary flow visualization experiments, the FMF
wind tunnel and tow-tank simulations, and the meteorological data collected
by PNM. Tables 10 and 11 summarize the characteristics of the sampling
grid. The sampler locations were selected to provide the widest horizontal
133
-------
tfl
PL.
CD
O
O •
O r—*
i-H CD
in cd
CD S
CD v—'
CJ
ni C_3
=tt=
cn oa
134
-------
TABLE 10. PRIMARY SAMPLER LOCATIONS
A. Four Primary Sampler Rows on Windward Face of HER
• #1 is ^10 m above the base of Tower B
16 locations centered (approx) on Tower B (60-series
stakes)
80-m horizontal spacing covering 1200 m
• #2 is ^25 m above #1
17 locations centered (approx) on Tower B (60-series
stakes)
40-m horizontal spacing out to 200 m, 'v-SO-m spacing
to approx 460 m'from center; total range is 920 m.
• #3 is 1-25 m above #2
- 21 locations centered (approx) on Tower B (60-series
stakes)
40-m horizontal spacing covering 800 m
• #4 follows "crest"; mean height is •v-lO m above #3
22 locations centered (approx) on Tower B (60-series
stakes)
- nominal 40-m spacing covers 860 m
B. Three Secondary Rows on "Windward" Face of HER
• #1 lies among hillocks at base of HER
- 5 samplers centered on Tower B
- middle is at Tower B; adjacent 2 are atop hillocks near
the road; 2 "ends" are in low areas adjacent to these.
- row covers 280 m
• #2 lies between primary rows #1 and #2 (-vl3 m above #1)
- 4 samplers centered (approx) on Tower B
- 100-m spacing covers 300 m
• #3 lies between primary rows #2 and #3 (>v.l3 m above #2)
4 samplers centered (approx) on Tower B
- 100-m spacing covers 300 m
G. Two Lee-Side Rows
- 6 samplers in each row
centered near Tower B (60-series stakes)
- 100-m spacing along and between rows
- covers 500 m along each row
- covers out to 280 m beyond crest
135
-------
TABLE 11. ADDITIONAL SAMPLERS
A. Co-located Samplers: two locations along center of grid: one on
primary row #1, and one on primary row #3 (10 m and 60 m above
the base of Tower B, respectively).
B. 10-minute Samplers: five locations: one at grid center along
primary row #2; two along primary row #1, 80 m to either side of
center; two along primary row #3, 280 m apart, centered 20 m
south of grid center.
C. Background Samplers: one on the east side of Waughan Arroyo; one
near the doppler acoustic radar near the substation on the high
ground east of the east arroyo.
D.
"Edge" Samplers: one to the north of the grid on the road up to
the top of HER beyond the lidar; one to the south where the San
Juan River flows through a gap in the ridge.
E. Elevated Samplers: one on Tower C at ^ 8m; two on Tower B at
* 14 and 28 m.
136
-------
coverage near the base of HER and the highest resolution between half way up
the ridge and the top. Three primary rows followed the height contours, and
the fourth row defined its crest. Samplers were deployed less densely in
the lee of the crest.
Tracer Analysis System
The Tedlar bag samples were analyzed for SFg and CFgBr by electron-
capture gas chromatography. Eight automated adaptations of 1972 Lovelock
prototype gas chromatographs were used to set up a gas lab (Figure 53) in
Farmington. The functional steps of the analysis are shown in Figure 54.
The laboratory analysis process had a sample bag check-in, an assignment to
a particular GG and subsequent analysis, and the output of an area propor-
tional to gas tracer concentration. These areas, along with measurements of
pressure, temperature, and physical constants, permitted calculations of
tracer concentrations. The calibrated response of individual GCs was checked
by injections and analyses of "known standard" reference mixtures of the
tracer gases instead of the unknown sample bag mixtures.
4.4.4 Flow Visualization
Oil-Fog Releases
The generation and release of oil-fog for lidar, visual, and photo-
graphic observations were accomplished by the injection of corvus oil into
the exhaust of a small turbine engine. The rate of oil injection, and the
ensuing fog density, were coordinated by the field operations director.
Fifty-five gallon oil drums and pumps provided oil to the generator. The
rate of pumping was monitored by a liquid rate—of—flow meter and was
adjusted by means of a valve. The operator recorded the start and end
times, the indicated oil flow rates, and any alterations.
Lidar Sampling
The plume-mapping lidar (or laser radar) scanned vertically through the
plume at several distances downwind from the source to produce three-
dimensional data on the fraction of light backscattered to the lidar by
particles within the plume. Plume position, dispersion, shape, and proximity
to terrain will be calculated as a function of time from processed data.
The lidar transmitter is a frequency-doubled Nd:YAG* laser emitting pulses
in the green portion of the spectrum at rates up to 10 Hz. Vertical scans
through the plume can be completed at a typical rate of two per minute. The
slower ruby laser used at SHIS #1 was installed as a back-up in case of
unanticipated failure of the Nd:YAG laser.
*Nd:YAG denotes a laser using an Yttrium-Aluminum-Garnet crystal doped
with Neodimium.
137
-------
GC
1
GC
8
GC
3
GC
4
Graphics \
Scope J
I
SP4000
Integrator
P/E
7/16
c±>
5x5 MB
Hard Disk
CON
CRT
1
9T
800/1600
cpi
9T
800/1 600
cpi
Plotter
Devices available but not online during data acquisition
Figure 53. Gas chromatography data acquisition system.
138
-------
Samples
of
Standard
Mixtures
Computer
Storage and
Calculation
Audit By
Independent
Standard
Mixtures
Figure 54. Tracer gas analysis system.
139
-------
The lidar recorded data on 9-track digital magnetic tape for later
processing. Some real-time information was available on plume behavior for
experimental guidance. The lidar was located about 2 km NNE of the tracer
release area (Figure 49).
Photography Program
Morrison-Knudsen Company, Inc., as subcontractor to NOAA ARLFRO, was
responsible for all photographic work for SHIS #2. The work consisted of
continuous video tape of each experiment and photographs taken every five
minutes from each of five positions. The photographs were taken during all
periods of smoke release, regardless of tracer release or other project
considerations. Morrison-Knudsen was responsible only for photography and
video taping; ARLFRO operated the oil-fog generator, and ERT provided a.
60-inch carbon arc lamp to illuminate the plume. The arc lamp was located
just north of the doppler acoustic sounder (Figure 49), east of the main
arroyo.
Three photography positions were fixed and two were roving. One of the
fixed positions was the 120-m level of the 150-m tower. The other two fixed
photographers were positioned to the north and south of the release east of
the base of the ridge. The two roving photographers chose positions that
allowed photography of the plume near the ridge top or of other meteorologi-
cal phenomena not in view from the fixed positions.
The photographic archive consists of 11 binders of color slides covering
experiments 1 through 15, one binder of supplemental slides, and 24 two-hour
video tapes. During periods of smoke release, and in the absence of other
technical difficulties, 60 photographs were taken each hour (five photo-
graphers each taking 12 photographs per hour). The photography program is
summarized as follows:
• five 35-mm cameras with data backs to document time of photo, etc.
• color film (Kodak Ektachrome ASA 200)
• one videocamera equipped with Starlight-scope
• locations:
120-m level of the 150-m tower
north of the release area east of the ridge base
- south of the release area east of the ridge base
- two roving photographers to get best pictures of plume
dynamics
• five-minute exposure (12 photos/hour/camera) at night; shorter
exposures in sunlight but same frequency
• turnaround - slides within 48 hours
• lighting - carbon arc lamp revolving once every 30 seconds.
4.5 Preliminary Field Study Results
4.5.1 Summary of HBR Data Base
The three flow visualization and il combined flow visualization and
tracer experiments at HBR have produced an extensive data base that covers a
wide variety of dispersion conditions and concentration patterns for
140
-------
modeling purposes. SFg was released (Table 12) during 91 experimental
hours and CF^Br during 88 hours. The tracers and oil-fog were released at
seven different locations during the course of SHIS #2 at elevations ranging
from 5 m to 75 m. The release locations (Figure 49) varied from 333 m
(R-80) to 800 m (Tower A) from the crest of the ridge.
Table 13 summarizes the number of hours with tracer releases and tracer
sampling for each SHIS #2 day. The experiments typically started around
midnight and lasted until after dawn. Approximately 80 1-hour average SFg
and CF3Br concentrations were produced every hour (Table 13). More than
7300 1-hour average surface-level concentrations of SFg and of CFgBr
were sampled and analyzed during SHIS #2.
Three samplers were suspended above the ground from the shorter
meteorological towers; two hung at approximately 14 and 29 m on the 30-m
tower at the base of the ridge, and one at about 7.5 m on the 10-m tower on
the crest. These samplers yielded more than 230 additional 1-hour concen-
trations of either tracer gas. At each of five surface sampling sites, four
sampler boxes were set out to gather 10-min samples on which to base assess-
ment of concentration variability during periods less than 1-hour long.
Approximately 1600 10-min samples of each tracer gas were analyzed.
ARLFRO has prepared the SHIS #2 tracer gas data base. A tape of
concentration data and isopleth maps (hard copies and slides) have been
given to the EPA Project Officer. After he witheld a randomly selected
subset of the data base for independent model testing (see Section 5.1), a
tape and the maps of the remaining tracer data were given to ERT. This
"learning" tracer gas data base has been used for the analyses presented in
the remainder of this report.
As discussed in Section 4.4, the tower meteorological data were archived
in real-time via the MDS. During the first few experiments of SHIS #2, it
became apparent that the meteorological data contained "noise." Subsequent
investigations during and after the experiment have suggested four types of
noises: (l) large "hits" or "spikes" which produced apparent full-scale
voltages from the instruments, (2) "cross-talk" or "skips" between MDS
channels such that signals from one channel were picked up by another
channel, (3) a high frequency, apparently random low-voltage oscillation
that was attributed to 60-cycle AC, and (4) an apparent 2- to 4-sec noise.
Figure 55 shows a plot of raw, unedited, 1-sec Tower A AT data (in
counts) taken between 2200 and 2300 on October 22, 1982. The excursions to
0 and 4095 are evident. It is also evident that signals from one channel
are being picked up by another channel. This channel-skipping appears as
vertical lines within the band of traces on the plot, the lines being
composed in this case of skips from channels 72 to 73, 73 to 74, etc. in
general, although some scans, like those near minute 31, involve only
channels 77 through 79 or some other subset of channels. Skipping is
particularly frequent in minutes 53 to 55 of this hour.
141
-------
TABLE 12. NUMBER OF SHIS #2 EXPERIMENTAL HOURS WITH
TRACER GAS RELEASES AND CONCENTRATIONS
Tracer Releases
Experiment
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Date
5
6
7
11
12
12-13
14
14-15
20
22
22-23
24
25
25-26
28-29
Oil-Fog
4
5
6
9
7
8
7
9
5
10
9
9
1*
10
11
SF6
-
-
-
7
7
8
7
9
5
10
9
9
-
9
11
CF31
-
-
-
6
7
8
7
9
5
10
9
9
-
8
10
Tracer Concentrations
SF,. CF^Br
12
12
7
7
9
7
9
5
11
10
10
6
7
9
7
9
5
11
10
10
12
12
TOTAL**
110
91
88
99
98
*Experiment 13 was terminated due to unfavorable weather.
**There are more hours of concentration data than there are of
tracer release because samples obtained after the tracer releases
were terminated were analyzed.
142
-------
TABLE 13. SUMMARY OF TRACER CONCENTRATION DATA
SF
Experiment Date One-Hour
4
5
6
7
8
9
10
11
12
14
11
12
12-13
14
14-15
20
22
22-23
24
25-26
15 28-29
Subtotals
523
497
695
527
659
393
842
798
765
936
930
7565
One-Hour
Tower
13
21
18
18
23
10
23
28
23
26
28
231
10-min
108
103
120
95
189
62
165
215
134
214
209
1614
One-Hour
487
496
696
526 -_
666
396
839
795
764
919
918
7502
One -Hour
Tower 10-min
13
21
18
20
23
10
23
28
23
26
28
233
101
103
119
82
186
62
168
211
135
214
209
1590
Total SV, 9410
Total CF3Br 9325
GRAND TOTAL 18,735
143
-------
m
N
8
o
LO
rH
a
o
S
o
o
o
I
LTl
LO
CD
• H
PL.
sjunoo
144
-------
The "hits" and "skips" can be characterized as follows:
• When a hit strikes a channel (between 1 and 16) on a multiplexer/
A-D board, it usually drives the counts to 0 (0.000 Vdc) or 4095
(5.120 Vdc). This hit will then appear on all subsequent channels
(up to and including channel 16) on that board for that scan,
although sometimes a hit that first appears as a 0 will be a 4095
in subsequent channels.
• There are two subclasses of hits, whose characteristics are suffi-
ciently different that they may be manifestations of different
phenomena.
- The hit may not drive the counts to either limit but to some
number like 19 or 23, which recurs in subsequent channels and
may happen more than once in a few minutes. This sort of hit
might result from dropped bits in the A-D conversion.
- The hit may show a gradual change from a "good" value on one
channel to bad values on subsequent channels, finally settling
on a limit value (0 or 4095) after a few channels in the scan.
• A skip seems to put the datum from channel n + 1 (occasionally n +
2) into channel n. Once a channel skip occurs, all subsequent
channels also skip to the next channel on that board. Channel 16
appears not to skip to channel 1 but to give correct values.
Due to these patterns of hits and skips, channel 1 yields the "cleanest"
data on each board; channel 16 has the most hits, and channel 15 the most
skips. As much as 30% of the data on channel 15 may be affected by these
errors for some periods.
Both ARLFRO and ERT staff tried to alleviate the noise problems during
non-experiment hours at the Hogback. Between experiments 8 and 9, the six
16-channel multiplexer boards were reconfigured so that none served instru-
ment outputs from more than one tower, and additional grounding of the
components of the control data collection system was done. These changes
seemed to remove most of the problem with the 60-HZ noise and the 2- to
4-sec noise, which may in fact have been different ways of looking at the
same phenomenon, but did not eliminate the big spikes and channel cross-talk
or skipping.
• ARLFRO and ERT scientists continue working on the hit and channel-
skipping phenomena. Preliminary methods have been developed and implemented
to filter the hits and skips. The effects of the 60-HZ noise are impossi-
ble to remove from the data. A procedure for isolating the 2- to 4-sec
noise has been developed and tried on several periods of data, but no method
has been developed for removing this noise from the data. The latter two
types of noise have little effect on the average meteorological data but may
degrade the turbulence statistics and covariances.
A first attempt to remove the hits and skips has been made by ARLFRO
staff using data collected during the 17 hours selected for initial modeling
and analysis (see Section 5.1). ARLFRO filtered the hits and skips, applied
145
-------
non-cosine response correction factors developed in 1978 by Dr. John F.
Clarke* to the prop data, and converted the counts to engineering units.
They have not yet investigated the 60-HZ and 2 to 4-sec noises, nor have
they "recovered" any of the cross-channel data. Substantial work remains to
refine and complete the tower meteorological data base.
Of the remaining SHIS #2 measurements, the WPL and ATDL tethersonde
data have been processed and delivered. The WPL data include tables and
plots of profiles of pressure, relative humidity, mixing ratio, wind speed,
wind direction and potential temperature up to altitudes of about 375 m.
The ATDL data include tabulations and plots of 15-minute averages and
standard deviations of wind speed, wind direction, and the vector wind
components, u and v, from the tethersonde flown near the smoke source. The
WPL monostatic acoustic sounder data have also been delivered as photographs
of the facsimile output from the sounder located near Tower A. The wind
speed data from the three optical anemometers have been received from WPL as
10-min averages displayed in both graphical and tabular forms. The archive
of photographs is also complete. No doppler acoustic sounder nor lidar data
are yet available. Preliminary 20-minute average sonic anemometer data,
excepting heat and momentum fluxes, are available from the real-time output
for the 17 hours selected for preliminary analysis (see Section 5). WPL is
currently reducing the sonic data.
In summary, the following data are currently (i.e., as of 31 May 83)
available from the HER experiment:
SFg and CF3Br concentrations and hourly average emissions data;
Preliminary, edited tower meteorological data for the 17 hours;
Unedited tower meteorological data;
Tethersonde data;
Monostatic acoustic sounder records;
Optical anemometer data;
Preliminary sonic anemometer data (taken from the real-time output
in the field); and
Complete photographic archive.
*These corrections were developed from wind tunnel experiments with
modified R.M. Young wind component transmitters rather than with the
Cliraatronics instruments used at CCB and HER. The Young instruments
have a vertical offset of several inches between the U and V arms,
which reduces the interference of one with the other, whereas the
Climatronics systems do not. Furthermore, the Young instruments were
modified by a plumb-bob leveling device that altered the flow
uniquely. Non-cosine corrections were developed for the Climatronics
instruments by ERT for application to the CCB data-base (Greene and
Heisler 1982) but were based on less extensive experimentation than
Clarke's.
146
-------
4.5.2 Overview of SHIS #2 Results
The flow visualization and tracer experiments at the Hogback have
produced an extensive (still evolving) data base. The real-time experimental
selection of release locations and heights based on the MDS output produced
concentration patterns for a wide variety of dispersion conditions. These
data will be used to evaluate and refine the modeling approaches developed
with the CCB data base. Initial analysis of the 17-hour data base suggests
that the concept of a critical dividing-streamline height is appropriate for
stable flows toward HER. Releases below HC dispersed in the "blocked"
flow and produced peak ground-level concentrations (GLC) on the windward
face of the ridge. Releases above Hc dispersed in a flow that went over
the ridge and produced peak GLC's near the ridge crest and on its lee side.
Figure 56 shows isopleths of hourly CF^Br concentrations (x/Q)
measured during experiment 5 (October 12, 1982) between 0500-0600. A band
of concentrations above 50 psec/m was observed from the ridge base to near
its crest. The highest x/Q of 387 psec/m was measured about 15 m uphill
(sampler 105, see Figure 52) from the intersection of Hogback Highway and
Tower Road. The tracer gas was released at location 203 at a height of
15 m. The calculated Hc was 47.9 m. Note that the CF3Br dispersed up
and over HBR while producing the maximum GLC at a ridge elevation roughly
equal to the release height. This seems to be different from the CCB
results, which show that releases below Hc tend to disperse along the
sides of the hill rather than up and over. An initial analysis of the HBR
data suggests that tracer gases released below Hc are often eventually
transported up and over.
Figure 57 shows CF-jBr concentration (x/Q) measured during experiment
15 (October 29, 1982) between 0200 and 0300. The tracer gas was released
from the 40-m level of Tower A, above the calculated Hc of 24.4 m. Only a
few non-zero GLC's were measured, and these occurred on the lee side of the
ridge.
Roughly eight concentration patterns were observed during SHIS #2:
1. Tracer gas distributed front and back of HBR
2. Primarily front side
3. Primarily lee side
4. Along upwind base
5. Near top
6. Near top and on lee
7. Spots
8. Nothing
Figure 56 exemplifies pattern 2 and Figure 57 exemplifies pattern 3 (or 7).
Other examples are given in subsection 4.5.4.
147
-------
LO
o
CQ
to
pt,
X
fi.
143
-------
"*""* "t"
,**
%>-
*
t
>
•>»
i-..
4
f
.«**
_ i
0
t
t
i
t
•
00
I
"•* •
I
O i! M 11
o — • CM ro
l\
1
\l
CO
O
_J
t uj
CO U,
CO
149
-------
4.5.3 Comparison with SHIS #1 Results
The essentially two-dimensional character of the Hogback and the
three-dimensional character of CCB dictate the behavior of the flows that
affect dispersion of tracer gases released upwind of these terrain features.
Despite these very basic differences, an analysis of the CCB data base and
an initial analysis of the 17-hour HER data base show the importance of the
critical dividing streamline in discriminating between types of flows and
subsequent dispersion conditions. At CCB, releases below Hc tended to
remain horizontal and travel around the sides of the hill. At HER, releases
below H,., tended to stagnate, producing high GLC on the windward side, but
then were often* eventually transported up and over the ridge. At both CCB
and HBR, releases above Hc flowed up and over.
At this time only the 17-hour HBR data base has been analyzed and
modeled, so any results are preliminary and not conclusive. However, it is
interesting to compare some of the results from HBR with those from the CCB
experiments. Table 14 summarizes the five highest x/Q and the associated
meteorological conditions observed at HBR and CCB. Generally higher x/Q
were measured at the ridge. The higher concentrations at HBR were measured
at elevations at least as high as the release height, while at CCB the
higher concentrations were frequently measured at elevations lower than the
release height.
Figure 58 shows a plot of maximum x/Q vs. l-Hc/zr for the entire
CCB data base and for the 17-hour HBR data base. The higher x/Q at CCB
tended to occur when zr«sHc, while at HBR the higher x/Q were measured when
zr^c' Since the ridge is basically two-dimensional, the higher values
could be explained by the highly variable and generally stagnant flow below
Hc and, if the average wind were directed toward the ridge, the tracer gas
would be transported directly to a sampler near the release elevation.
Flows below Hc at CCB tended to remain essentially horizontal with some
uphill transport and generally preferred one side of the hill. Figure 59
shows a concentration (ppt) map for CCB experiment 206, hour 8 (the hour
with the highest x/Q of 164). The pattern of concentrations is fairly
horizontal and is elongated northward along the north side of CCB. Compare
this pattern with the patterns in Figure 56 and Figure 77 (in the next
subsection). At HBR there seems to be eventual uphill transport for flows
even below Hc.
The higher concentrations at CCB occurred when the transport wind
direction was close to the stagnation streamline. Figure 60 shows a plot of
the CCB x/Q vs. A9/iv, where A6/iv is the ratio of the wind angle between
the actual wind direction and the stagnation wind direction divided by the
crosswind turbulence intensity. A large value implies that the tracer gas
plume would have little chance of "hitting" the hill. The data show that
the peak GLC values occurred when A6/iy^l. Small GLC values occurred when
A6/iy>5.0.
*Not all releases below Hc were transported over the ridge. If the
winds near the release had a substantial component parallel to the
ridge, then the tracer stayed below the crest.
150
-------
VJ C
(U O
1-) .1-1
a, u
g cd
w a)
w
C* CO CM CO O
CM CM oo m CM
in in o o o
CM CM CO CO
CO
cd
erf
M
o
CM 00
fc
CD
CO
,0
o
Cf
X
w
PQ
0
Cf
X
H
CO
Cd
ffl
Cj>
r— 1
PC
s
v_x o 'n m o
H CM i-l i-l CM
N
^-v
CO
e
o
co
3.
r~- co r-^ r~-
r-t OO 00 ^O
o
CM
^j.
0
-
-------
o
+ SF6
* CFgBr
1-Hc/zr
a. 17-hour HER data base.
138.8
120.0-
118,0-
108,0-
n"
^ Oi.O-
a so.o-
•54
SB.8-
<0,0
30.0-
20.0-
10.0-
_£*_
-.SB
1-Hc/Zr
b. CCB data base.
Figure 58. Maximum hourly x/Q
152
vs. l-Hc/zr.
-------
• » r = 500 m
N
Figure 59. Observed SFg concentrations (ppt) for CCB Experiment 206,
hour 8 (0700-0800 MST). Source: r = 595.9 m, 6= 123.6°,
net height = 29.5 m, Q = .062 g/s.
153
-------
§
to
g
w u
+ *
* *
*•*
'V
•*-•*).
.i
.8
rt
CQ
CJ
CJ
bO
154
-------
Large concentrations were observed at HBR whenever the average wind
direction approached the ridge, even for releases below HC.
In summary, there are several similarities and differences among the
results from the two experiments:
• Observed peak x/Q values were greater at HBR than at CCB;
0 Hc discriminates between the flow regimes;
• At CCB the peak GLC values occurred when z^ HC ; at HBR the
peak GLC values occurred when zr
-------
00
rn
CD
CD LO
VO
156
-------
oJ
CD
fn
oj
•p
CD
00
^
cd
•P
3 E-
O Q
w S
CQ
33
o •>
fn OJ
M-l 00
CD i-H
CD 4->
•M S
3 CD
C S
6
i
CD
H
C>0
•H
PH
157
-------
-------
Figure 64 shows vertical profiles of hourly average wind direction*,
wind speed, and temperature derived from the Tower A measurements. On the
basis of this information and measurement from the photographs, the oil-fog
plume rose to about 60-80 m and was transported toward HER in the ESE winds.
Time-series of 5-minute average winds measured on Towers A, B, and C, and
5-minute winds estimated from the photographs are plotted in Figures 65 and
66.
A map of one-hour SFg concentrations is given in Figure 67. Figure 68
gives the along-hill (approximately crosswind) distribution of hourly SFg
concentrations for nine sampling "rows." Figure 69 identifies the row
plotted by each symbol in Figure 68 and Table 15 defines the rows by their
sampler numbers. There is a wide area of relatively high (greater than
5.0 usec/m ) concentrations over most of the ridge. The peak values
were measured near the ridge crest and on its lee side.
Figure 70 gives a map of hourly CF3Br concentrations, and Figure 71
shows the along-hill distribution. The peak values occurred south of Desert
Drive at an elevation of 5470 ft (row 4), an altitude about 10 m higher than
the release elevation. Note that the wind direction at 40 m in Figure 64 is
consistent with the observed concentration pattern.
Experiment 10 (0300-0500)
,.<
The morning of October 22, 1982 also experienced an easterly drainage
wind. The drainage became. well established just after midnight, and by 0300
ESE winds were directed toward HBR at about 2 m/sec. The oil-fog, SFg,
and CF3Br were released from Tower A — the oil-fog and SFg at 50 m, the
at 30 m.
Figures 72 and 73, photographs taken from Tower A and from the ridge
crest south of the target area, show a coherent plume going up and over HBR
at 0330 MDT. The second 5-minute exposure (Figure 73) shows some plume
material diffusing to ground-level near the ridge crest. The profiles of
hourly meteorological variables are given in Figure 74. Time-series of
5-minute wind directions are given in Figures 75 and 76.
According to the photographs and the measured wind directions, the
oil-fog (and SFg) plume dispersed steadily up and over HBR south of the
light cross. The plume rose to an altitude above HC (50.1 m) as it was
*The profile plots of wind direction, wind speed, and temperatures
were produced in the following manner. The 5-minute temperatures and-
the prop wind speeds and directions were used to compute 1-hour mean
temperatures, scalar wind speeds, and unit vector u [sin WD] and unit
vector v [cos WD] wind direction components. This was done for each
instrument level of Tower A [2,5,10,20,30,40,60,80,100,150 m]. A
"spline under tension" method was used to interpolate the meteorologi-
cal variables for every 5 m between instrument levels on Tower A.
The splined u and v unit vector components were used to derive the
unit vector wind direction.
159
-------
M O r-
O
CM X
o o
o o
rH K5
o o
T
-5-
5 S
-8
•C CD CN
6 4-> 00
TJ a •-!
2 §
o fi
CD
w g
CD -H
i-H fH
•H ^
O I
160
-------
Tower A
60m
80m
100m
Photo-Derived (oil-fog)
tine (Hour)
WD (degree)
270
Tower C
,2m
- 5m
10m
Photo-Derived ;(oil-fog)
time (Hour)
Figure 65. Time-series of 5-minute wind directions from
Tower A and Tower C (Experiment 4, 10/11/82,
0300-0400 MDT).
161
-------
8
CO G>
CD tD G»
ffl
•^ ia is is is is io
N- * — flO tO PJ
_ W M (M — — ^
£ E
o o
t- CO
I
o
03 CO 03
09 G)
00
LO
-p
-------
*»*
ess
X
UJ
os> o in
{D*-«
fO
CSJ
o
•H
ed
S
CD
O
O
O
fl,
CO
f-l
I
U,
CO
CM
oc
Cs.!
163
-------
CO E
"~ X
to
B
C
o
•H
•P
cd
C
r •»• m
L§
h r-
M-l
^ O
S. 10
-------
CO
o
cc
1
cc
g
1
DC
00
1
cc
Q
OT
O
CC
*
o
tn
^
-------
Row
7
8
TABLE 15. HOGBACK SAMPLERS BY ROW
Samplers
1-5, 702, 703
101-106, 711-714, 109, 809, 721-724, 112-118
107, 108, 110, 111
201-205, 207, 208, 210, 731-734, 211, 213, 214, 216-220
206, 209, 212, 215
301-307, 741-744, 308-313, 751-754, 314-319, 810
401-417, 419-422, 701
501-506
601-606
166
-------
t-(
•—I
o
I
•H
o
•H
4->
oJ
fn
•p
C
CD
O
a
o
o
fin O
U O
XO
3 o
o to
33 o
o
0>
is.,
167
-------
I
to
ft
O O
CQ O
K50
Xo
ro CM —
S
1
O <5> CD O)
Q> CD CD u> *• m 0
168
-------
CM
OO
CN
CM
•P
I
•H
tu
CD
I
6
o
CD
fn
10
O
•
C P
• M Q
s s
I
co o
> to
•H tO
PI-I O
CD
f-i
a
•H
169
-------
CQ
OO
0 T-l
P! B
I
-------
ID O r-
II II II
o T ~
Xu.li.
1
ctf Q
4-J 5*.
ctf
*"Ci ^^
O
<—I "=t
cd o
o i
•H O
t>00
O tO
t—( O
o
^ «
o CNI
•?
17-1
-------
278
248-
218-
188-
150-
128
00
68-
38
8'
-30-
-BB-
Tower A
60m
80m
100m
—•»— Photo (oil-fog)
-00-
WD (degree)
278-1
time (Hour)
248
210
188-
158
I28-|
08
ea-
se
e-
-38-
-68-
Tower C
2m
5m
10m
—+— Photo (oil-fog)
tin* (Hour)
Figure 75. Time-series of 5-minute wind directions from
Tower A and Tower C (Experiment 10, 10/22/82,
0300-0400 MDT).
172
-------
o> '. _
C P S £r'*"r
1 g 1 = '
S 4 N
-------
transported toward the ridge. The resulting pattern of hourly average
concentrations is shown in Figure 77. Distributions of along-hill concen-
trations are given in Figure 78. The highest concentrations were measured
in the lee and near the crest (one sampler). Relatively low concentrations
were measured on the windward face.
The winds below Hc were quite variable (see Figure 76) and the
resulting CFgBr concentration pattern (Figure 79) reflects the variability.
During the beginning of the hour, the winds were from the SSE, producing the
CF3Br concentration peak along the north edge of the sampler grid. A
secondary peak was produced by the easterly winds that occurred during the
second half-hour. The along-hill distributions of Freon concentrations are
given in Figure 80. Two values above 5.0 ysec/m were measured—one
near the crest (row 7) and one at the base of HER.
The easterly drainage wind persisted throughout the next hour. The
SFg and oil-fog release was changed to 70 m; the CF3Br release stayed at
30 m. Figure 81 is a photograph of the oil-fog plume taken from the crest
south of the target area at 0425 MDT. It shows more lifting of the plume as
it goes up and over the ridge crest. Figure 82 shows the distribution of
hourly ground-level SFg concentrations. The highest values were measured
on the lee of HER.
The vertical profiles of hourly winds and temperatures are given in
Figure 83. The winds at 70 m remained from the ESE, while there was a layer
of more southerly winds centered around 30 m. The time-series of 5-minute
wind directions are given in Figure 84. Again, the CF3Br concentration
pattern (Figure 85) reflects the variability in wind directions—a primary
peak near the north end of the target area produced by the SSE winds and a
secondary peak at the end of Desert Drive.
174
-------
CN
CN
CD
CD
o
•H
S-i
•p
s
(U
o
o
Pi O
co o
i— i i
^ o
3 o
O K5
X o
CD
1-1
•H
fin
175
-------
coin
I
§ §
<\J ^- — ^ ^ _
g ^
£
i
i i
CO
00
-------
CM
OO
CM
CN
e
-------
I S
g ^
.--•"*
''''
8
g
f.ia
ffl
& i)
"S E"
g
15 S"
§ m
•P
nJ
t-t
C E-
O Q
oa o
XO
o •>
00
0)
o o
IX)
-------
CD
cd
•p
CD
o
II
w
LO
Di CM
CQ -5t
ffi O
a -
o cj
!H oo
-------
00
w
o
C/D O
33 o
rr
180
-------
in o to
ii ii n
s
u>
s
<\f
1 §
8 S -5
s
m
is
(M
s
g g
-5-
m
Is- O
2
« 3
.|
H
Q
O
T—I LO
cri o
u i
•H O
boo
o n-
i-H O
o
O CM
CD 00
(D CM
g CM
XO
U)
1
O O
^3 •-<
4H 4->
O C
O
w 6
•H
4-1
O
m
H O
CD H
s i
~ 5
bO
-§
-a
§
•181
-------
Tower A (props)
20m
30m
40m
tine (Hour)
WD
-------
CM
OO
CM
CM
•P
rt
CD
•H
f-t
CTS
(D
O
O
O
pa
fin O
U O
3 O
O •<*
DC o
LO
oo
bO
•H
O-,
183
-------
SECTION 5
PRELIMINARY MODELING OF THE HER DATA BASE
The SHIS #2 at the Hogback has enlarged the data base with which to
evaluate and refine the models developed with the SHIS #1 data base and to
extend them to a two-dimensional ridge setting. This section describes the
first evaluation of existing models with the HBR data base. CTDM( 11083) and
GTDM(11083-E), modified for the geometry and topography of HBR, ": and COMPLEX I
and COMPLEX II were run on an initial 17-hour data base that was selected to
represent a variety of dispersion conditions. Model calculations were
compared to observations, and the performance of the three models was
evaluated and compared to the SHIS #1 results.
5.1 Initial 17-Hour Data Base
5.1.1 Selection of Case Hours for Model Evaluations
The 91 hours of SF5 release and 88 hours of CF3Br release were examined
to pick hours for model evaluation. The first step was the selection of a
set of the tracer data to be withheld by the EPA Project Officer for
independent model validation purposes. The selection procedure involved
tossing (actually spinning) three coins. All experiment hours for which the
unedited wind direction at source height was between 60° and 170° were
candidates; there were 66 such hours. If the first coin toss for each hour
yielded three of a kind (prob = 1/4), a second toss was made to determine
whether SFg, Freon, or both would be withheld. If the first toss was not
three of a kind (prob = 3/4), the next hour was investigated. If the second
toss again yielded three of a kind, both tracers were withheld (prob ?= 1/4 x
1/4 - 1/16). Two heads and one tail on the second toss meant SFg was
withheld; one head and two tails meant Freon was withheld. The probability
of either event is 1/4 x 3/8 = 3/32.
The results- of this exercise were that SFg was withheld for six hours,
Freon for eight, and both for three, so that 20 tracer hours (Table 16) have
been selected. There were 86 tosses of the three coins; 20 (23%) were three
of a kind vs. 25% expected; 27 (31%) were 2H,1T vs. 37.5% expected; and 39
(45%) were 1H,2T vs. 37.5% expected.
For the remaining 159 hours of tracer data, a pattern was assigned to
each hourly SFg and CF3Br concentration map based on the eight patterns
discussed in subsection 4.5.2. After examining the hourly concentration
patterns, photographs, log book comments, and unedited meteorological data
for each experiment hour, a set of 42 hours was selected. A priority was
then assigned to each of the 42 hours (6-GOOD, 21-MEDIUM, 15-LOW) by
184
-------
TABLE 16. SHIS #2 TRACER DATA WITHHELD FOR MODEL VALIDATION PURPOSES
Exp . Hour
No. MDT at
WD* and WS**
SF& Release Ht.
SF, to be Withheld
4
5
9
10
12
15
Freon to
5
10
10
11
11
12
15
15
Both to
6
6
9
06-07
00-01
06-07
00-01
05-06
08-09
112
67
62
Var
101
102
0.7
3.9
4.6
0.6
3.0
1.8
Release Ht.
SFg Freon
30
40
40
30
75
50
20
25
20
20
50
40
H ***
60
26
31
62
40
-
Location Fr(H,J
216
203
203
203
.A
A
•"!•«••—
2.25
1.83
1.90
Not yet
2.68
—
Fr(2m)
'0.54
0.70
0.46
available
0.89
-
be Withheld
06-07
01-02
09-10
00-01
04-05
08-09
05-06
09-10
168
87
116
106
100
111
80
125
1.4
1.0
1.9
0.6
2.3
3.0
1.1
1.4
30
30
70
40
50
50
50
50
25
20
30
20
25
40
40
40
76
36
11
50
36
16
-
-
203
,203
A
203
A
A
A
. A
0.40
0.86
1.70
2.23
2.01
1.51
-
-
0.26
0.06
1.77
0.52
0.78
2.29
-
-
be Withheld
00-01
05-06
08-09
114
88
112
2.2
2.0
2.5
40
35
40
30
25
20
22
35
19
111
203
203
0.95
2.32
1.90
0.32
1.07
1.05
*Unedited wind direction estimated for SFg release height.
**Unedited wind speed (m/s).
***Calculated from the unedited meteorological data.
185
-------
analyzing concentration patterns and other information to select a variety
of situations. The final selection of 17 hours, presented in Table 17, was
based on obtaining a good distribution of concentration patterns, meteoro-
logical situations, release locations, and release heights for the modeling.
Table 18 summarizes various statistics based on the preliminary meteorolog-
ical data for the 17 hours.
5.1.2 Construction of Model Inputs
Tracer Data
The release crane position, the time and duration of tracer releases,
the emission rates of each gas, and the height of the tracer release above
the local surface were recorded in release logs maintained by ARLFRO during
the experiment at HER. The survey data were used to determine the locations
of the release point, which were expressed in the hill coordinate system, a
polar grid (r,6,z) centered at Tower A. The base elevation used in this
system is 1600 m.
Plume Rise
The oil-fog generator produced enough heat flux to cause a thermally-
induced rise in the oil-fog and co-located SFg plumes. A first attempt
was made to calculate the effective release height of the SFg plume for
model input by using a method devised by Halitsky (1961) for single-camera
measurements of smoke plumes.
Analyses of the wind direction time series and wind speed and tempera-
ture profiles were also used to guide the estimate of the plume rise.
Wind Data
A "spline under tension" method, as discussed in Section 2.1, was used
to interpolate wind speeds and directions between instrument levels on
Tower A. The 5-minute prop wind speeds and directions from Tower A were
first broken into wind components, and the components were interpolated to
obtain the horizontal wind speed and direction at plume height for each
5-rainute period.
For those 5-minute periods in which photographs taken from the 150-m
tower were available, wind directions estimated from the photo images were
substituted for the splined wind direction data. These photo-derived wind
directions could only be estimated for the SFg release because only the
SFg was emitted with the oil-fog.
A scalar average of the 5-minute interpolated wind speeds was used to
form the 1-hour mean wind speed. A 1-hour vector average of unit vectors
along each 5-minute wind direction was used to form the 1-hour mean wind
direction.
186
-------
TABLE 17. SHIS #2 HOURS.SELECTED FOR INITIAL MODELING
Exp.
4
4
5
5
6
8
8
8
10
10
11
11
12
12
14
15
15
Date /Time
11/0300-0400
11/0400-0500
12/0100-0200
12/0500-0600
13/0100-0200
15/0100-0200
15/0200-0300
15/0600-0700
22/0300-0400
22/0400-0500
23/0200-0300
23/0500-0600
24/0200-0300
24/0600-0700
26/0300-0400
29/0100-0200
29/0200-0300
Release
Location
216
216
203
203
111
215
:215
215
A
A
203
A
A
A'
203
R80/A
R80/A
MSTQ.'
30
40
30 ,
20
30
10
10
30
50
70
40.
50
75
75
40
20
20
: Pattern
IE
IS
6C
IE
6S
1C
5C ,
IN
6N
3C
5/4C
6S
7
7
IS
IE
IS
z (CF B
20 -
30
15
15
2-0
5
. '5
30
30
30
20 ..
25
40
25
20
40
40
r) Pattern
2S
2S
2C
2C
. 1C/N
2E
4C/S
IN
' 4C
2N
1E/S
IS
IN
8
2E
4C/7
3/7
187
-------
TABLE 18. 17-HOUR DATA BASE SUMMARY STATISTICS
1. Number by release location
216/215
10
2. Number
< 10m
4
3 . Number
<. 20m
2
4. Number
(>
5 . Number
17 Hours
#
%
Entire Data
#
%
by
by
by
S
120
9
203 Hi A R80
local release height
11-20 21-30 31-40 41-50 > 50
98733
height of source above the ground
21-30 31-40 41-50 51-60 > 61
2 8 6 2 14
wind direction
E N
°) (120-45) (<45)
19 6
by concentration pattern
1
13
38
Set
67
41
234567
723242
21 6 9 6 12 6
52 3 11 4 16 7
32 2 7 2 10 4
188
-------
Critical Dividing-Streamline Height,(H ) •, .. . •
The prop wind speed and temperature profiles from Tower A were used to
compute the 5-minute values of HC by means of the integral formula presented
in the CTMD First Milestone Report, and first suggested by Sheppard (1956).
These 5-minute values were then averaged to 1-hour values. The effective
height of HBR was taken to be 85 ra.
Brunt-Vaisala,Frequency (N) '
For each 5-minute period, the Brunt-Vaisala frequency (N) was estimated
at plume height by evaluating the temperature change along the splined
temperature profile in the immediate vicinity (zr..i 0.1. m) of the plume
height to obtain the local temperature gradient. These 5-minute values were
then averaged to 1-hour values.
Stability Class
The Turner dispersion stability class was calculated from cloud cover.
and wind speed data. The cloud cover was estimated from photographs and the
wind speed was a 1-hour scalar average measured by the props at the 10-m
level on Tower A.
Turbulence Parameters
A comparison between the 5-m and 40-m turbulence data from the sonic and
the prop systems (Table 19) at the same "elevation proved that the two systems
were inconsistent. Due to uncertainties "with proper prop response correc-
tions, the sonic data were used to calculate the turbulence parameters.
One-hour values of u, v and aw obtained from the NOAA/WPL sonic
anemometers on Tower A at 5m and 40 m were used to derive the vertical
intensity of turbulence (ig) at plume height. The sonic data were
linearly interpolated to the plume height if this height was less than 40 m;
otherwise the 40-m iz was used. Sigma-w (aw) at plume height was calculated
by multiplying the interpolated sonic iz data with the Tower A propeller
wind speed interpolated at plume height by means of the "spline under
tension" algorithm used in the MDA (see Section 2). This methodology
preserves the sonic iz data within the model. Five-minute average sigina-v
data were set equal to the 1-hour sigma-w data for input to the PDF form of
CTDMU1083 and 11083-E) .
5.1.3 Summary of Model Input Data .
Table 20 lists for each tracer gas all of the inputs t'o the models
evaluated with SHIS #2 data in this report. The time and date for each of
the 17 experiment-hours is followed by the release location (r,9), the
emission rate (Q), the release height above the local ground .elevation, the
estimated plume rise, the effective plume elevation above the zero contour
(1600 m,MSL), the critical dividing-streamline height (H^, the hourly
average wind speed and direction, the Brunt-Vaisala frequency '(N), the
Turner Stability Class, and sigma-w (aw).
189'-
-------
IT! CM 0*
CM 1-1 r-l
O i-l CM
VO CM
CM r-< i-l
O CM
•* CM
00
.§
CM O
CO O O VO CM VD O COCM
en en
oo
-3- o>
r~- o
co
o co
oo
en co
-* O
a
I
I
03 03
o o
r-l O
C3 O
O r-4
CM O
O CM ,-1 00
O O
O i-t
Or-Hr-l OO
O r-4 i-HCO
OOO OO OO OO
• • •
in r-l r-l CM VO
II
O O
- _ _
OOO OO
J.CM
o o
00
a) a)
« B-5
m
cd CM
00
s.
190
-------
to to to 'to to to Q to
to W to to ' to t=* to
en en r-i
cd
r-* vo co o o m o
en CM T-I cM-en r-4 t~t
TO CM CO
O O O O O O O
en m o m -'-d- co co
c- 01 ** i-* oi en m
,-1 CM CM -t o
a
< CO
in m r-i
CM r-( en
CM vO CM tn
i—i en en m
o op
CM 00 CM
-* -d- •*
co in c?\ \o
vo d- CM
-------
•rf 0
A r-t
a o
O \D O en CM CM CO
(U
§
O CO ,-t
00 CO O
CM *-i en
O o o
o
o
01 •
ra u
O r-~ o*\
M
CO
< a
•a ^>
•»*1
3 a
n -o
u 01
> a
•< eyj
CO
erf
s
« .8
I S3
0) 3 AJ
S5 >,
Ed CJ
co en oo r^. en en
CM -^ CM en
en
m
o
o
o
o
0
en
00
u-i
•a-
o
o
o
o
0
o
0
o
o
o
o
o
0
en
en
CO
S
•»
o
o
o
o
o
o
o
o
192
-------
5.2 Complex Terrain Models Evaluated
Valley, COMPLEX I, COMPLEX II, CTDM(11083), and CTDM(11083-E) have been
run on the preliminary 17-hour data base for SHIS #2 at HER. Valley,
COMPLEX I, and COMPLEX II required no modifications for application to HBR,
aside from a change in the sampler grid data. CTDM, however, required
several changes.
CTDM needs terrain data to select stagnation streamlines, impingement
points, and to identify -which receptors are subject to Wrap-type calcula-
tions. For CCB, the polar coordinate system is fixed at the center of the
hill, and a general terrain profile is specifed by
= 970 h~°'38
(59)
where h is measured in meters from the 944.9 m MSL height contour. For HBR,
the polar coordinate system is centered-at Tower A, and the hill profile
perpendicular to the ridge from Tower A is assumed to extend to infinity
along the local tangent to the ridge. The hill profile is fit by the
equation
h = -.42 (807-r) (l-exp(-253/(807-r))) + 91'
(60)
where r is measured from Tower A along the perpendicular to the ridge, and h
is the terrain height above 1600 m MSL. Note that Tower A lies 807 m from
the ridge top, at an elevation of 1603.6 m ASL.
Figure 86 summarizes the HBR geometry contained in CTDM. The height
contours across the sampler grid on the eastward face of HBR define a tangent
plane oriented approximately 2,7° from truet north. The ridge is considered
to be two-dimensional. For Lift computations, the losbal radius of curvature
is set to infinity, and no streamline deflection is allowed in the horizon-
tal. For Wrap computations, the stagnation streamline is defined to follow
a 117° wind, and the distance to a receptor is the sum of the distance to
the receptor-height contour along 117°, and the distance measured along the
ridge from this point of intersection to the receptor. If a receptor lies
on the lee side of HBR, then no Wrap computations are made because the layer
below Hc is assumed to remain on the windward side of the ridge. Receptor
heights are adjusted to equal the dividing-streamline height for performing
a Lift computation at lee-side receptors below Hc.
5.3 Overall Performance Statistics
The data listed in Table 21 were used to run CTDM(11083), CTDM(11083-E),
Valley, COMPLEX I and COMPLEX II. The model calculations were compared to
1-hour average SFg and CFgBr concentrations taken from the 17-hour HBR
data base. Table 21 summarizes, .the residual' statistics. The results are
based on both tracer gases (17 hours each). Columns labeled "Peak
Concentrations" refer to the highest CQ/Q and Cp/Q values unpaired in
space, but paired in time, and the columns labeled "All Concentrations"
refer to values paired in space and time. Tables 22 through 25 give the
193
-------
Figure 86.
Height section and orientation of idealized HBR relative to
Tower A, as represented in CTDM.
194
-------
w
PQ
Q
Pi
P-.
a
CO
§
•I-l
4J
cd
n
4J
a
Hi
u
e
o
o
CO
g
•t-4
fi
(U
O
§
o
in
• *
.
co
CM
O
CO
co
i— i
VO
CM
•
O
CO
o
r-l
m
CM
S.X
/-N
1-1
o
^~s
CN
I-l
r-l
r-l
•
O
o£
•
o
^ — \
^x
/--N
0
•
0
-**~s
CO
vO
M
I
CM
in
CM
VO
vO
CM
X
§
CU
cd
>
«
Q)
^
o
S
(^
co
a>
3
r— 1
cd
^
o
u
ft,
o
(J
0)
4J
cd
r-l
3
U
r-l
td
o
M)
CO
m co
0 0)
-- 3
CO r-l
3. Cd
I4_l
O c)0
CO
CO
4J T3
•H a
a «
3
bO
cu E
cd co
,£ V
4-1
cd o
rJ C
CU
td
CO x— s
cd
S v*^
r-l CM
CO
CU
4-1
O
i3
CD o
•J3 >
4-1 Cd
too
O
4-1 >\
a r-i
*r-t t
as
cd w
r-i o
cd
CO
CU CO
CO -r-t
cd &
V H
0)
ca
13 0)
0 g
r-l -r)
4-J
CU
4J «
cu cu
CO 60
3 M
cd cd
0 r-l
cu
.£> to
cd
cu 5
M
w'x
cd v»x
CJ b3
g h
^
ctf^^*«
H 0
1—i " •
cu a
4J CU t"l
^ O
a o
0 <4-i (3
cu cd
II5
4-J ^^
4-1 0.
cd • *^ c*_^
cu ^-*
r-l tlO
CU C O
4-1 cd
cu cu
.Q 4J 60
co o cd
CU f r-l
O co
13 cd
4-1
>*i cd (>,
cu cu
, 1 I 1
r-l rH r-l
c« -H cd
^ ^c ^
*
195
-------
results of the 34 individual hourly simulations for all the models except
Valley. It should be emphasized that these results are very preliminary
because the meteorological measurements used to derive the model input need
further editing and refinement.
The residual statistics show that CTDM(11083) and CTDM(11083-E) under-
estimate peak CO/Q values by roughly a factor of 2.0. On the other hand,
Valley and the COMPLEX models overestimate the peak observed values by
roughly a factor of 4.0 or more. ma for the space/time-paired concentra-
tions is lowest for the two COMPLEX models. The noise and resolution
statistics are large for all the model calculations.
The tables of the results for each hour show that the overall perfor-
mance statistics are generally dominated by extreme values. For example,
Table 23 shows that for seven of the 34 tracer-hours the ratio of the peak
CQ to the peak Cp is greater than 9.0. The input wind directions and
modeled concentrations for these hours suggest that CTDM (11083-E) largely
missed the hill. Similarly, Table 25 shows that COMPLEX I completely missed
the hill for six of the hours.
Several of the problem simulations occurred for wind directions nearly
parallel to HBR and for releases below the calculated Hc. Further analysis
of the variability of the 5—minute average wind directions for a few hours
suggests that CTDM should be reformulated for releases below Hc at HBR.
For example:
• Is the concept of a single stagnation streamline wind
direction of 117° relevant for these highly variable flows?
• Should the PDF form or Gaussian form of CTDM be applied?
• How should the recirciilation of pollutant material from the
south (or north) of the sample grid be handled by the model?
• In addition to the data refinement issue, are the Tower A data
representative of the plume dispersion conditions,
given the highly variable nature of the flow below Hc?
Experiment 11 (0200-0300) exemplifies some of these concerns. Estimates
of 5-minute winds for the release height and from the 5~m, 10-m and 30-m
levels of Tower -B are listed in Table 26. The wind data estimated for the
release height indicate that the CF^Br plume should have impacted on HBR
for only about a 10-15 minute period during the middle of the hour. The
Tower B 10-m data suggest a more prolonged impact, from about 0215 through
0245 (see Figure 56). In any event, the CF3Br plume was probably swept
from the south side of HBR to north of the sampler grid during the course of
the hour, so that the samplers were influenced by plume material for a
considerable part of the hour.
The PDF reflects the wide range (18.2 - 225.0°) of wind directions
measured on Tower_A; hence, CTDM(11083) and CTDM(11083-E) produced very low
concentrations (Cp =0.6 and 0.2, respectively). The COMPLEX I and
COMPLEX II models produced relatively high concentrations (Cl = 50.3 and
196
-------
I
o
(\j o vy.
'^rO— ih-
«-< -« ru ru »•-
»-< 53-
to
c
i
o
tn j"> in x>
inN-ssr\j3-ruot«liin
-< ru
*-< *-4 "-o'*-* ^ in co
c it
o -o IB »M
to
o
X X
(C (D
X s:
o c.
o o
«n co >^ in •*"* co in co co
,£> CO i
«• AJ
m —i -t
ru AJ •-« ru —i -* ru
oo
ru
ru
3
I-H
to
X
(0
a
CJ
Ajcoo-4!>r«-»^<«a-Ln
o- zt *-> rvj
ru »
ru r\j
(M
»i in ru
00 r~>
O tU
•-< CO
S
e
u
CQ
a
to
i I 'o
I g
U-, O
o tc
I
CO 1^ X
CO
53" *-« o O* O* c** «••* "^ r*^ f\j ^ ^ ?y r^ rO ru o
ru-*o.'M''M:1\jru^or^'~txi-*co—««o-*>o
o-»^rux»H(n—«-oin!o»-*oin
X> if
o
ru
oooruru^'-io^in
5t ^, ^, _, ^, nj ru in
^jox>-- o o
-o o> ru
CNl
CNJ
W
CQ
ft
i O
I U
a~ v *o o r>- ^
—• «»• in «-« co ru i
^5 3^ O^ ^5 ^^ ^^* in 2Q
>»o«a**ea
ru ^< ru =r in co
^) in ^o -^ c^ ~~*
o r\j ^ 'XJ
—i ^ ru
4 ~-
AJ ru c-
«4 «-« a-
ru
ru
ru it ^>
cu ru oo
>-<•-«•«- ooru^r o
.
CQ
197
-------
o
«•« ru
CO
lo
I I
I o
10
ru in o r^. oc -* i
f~ ru ru •** i ft
—i i ru
I i
ro I I in
*-« *~* I ^-* GO
o o ru r-- ru >c I *-<
-* ~o i i i i
(D (C
£ Z
O C.
O CJ
ru ru ru —« r*Y
m -- *-• -<
•^ *H 0s ^ ru ru t^
ru ru i-> ru >o «^ «•<
ro
eo /-»
O r^ Q.
r-t W /T
r-l CO ".
*-* < tX
— ru ru xr o -i ru
•-« r^ in
rvj
S °
D! Crf
i a
to
in »^ ">
so ru
53- «-• ru
O, 3
O Q
f- x
3 «—'
^-i oo cc ru ^-» =3- •-«
to
OJ
hJ
03
in co t»» »H ru i
-H ru ru
ru
ru
10
orufiru ^ru^ru'-truru
-4 ~* ru
ru ^» ^ OD ^< —< h» ooru^1 o
•^ -^ ru ru so
•H >£>
ru
a b
\o
198
-------
i
o
•oomoo-eef».^aoooin?rac
o o
I a
lo
I i
I o
10
til
it
en
O
i—i
X X
1C (5
c a
? i j_j
o
o
a- o-
c^ c?*
&•
o-
&
o*
S
i— i
en
a
o
nj co
i-«.O'.
r\i
OO
f\J
(\J
X
h-4
a,
I
«
0
u
]—1 Q£j
O CQ
K K
tL. Crf
o no
O
en a
E- i
•~«-*r\l>-i —t ix^^,,^ .^
»H f\) t\l
en
BJ
crf
m
i-4
c
CJ
rg
*-ii'j"'. nj
=>-vj~n-M
•^ <*~ r\j ^i ,c ®
•^ —i r\j AJ oo
•^ -C
o
nj
, t/J
LlJ
CQ
ru ru a- in,tn ptT
199
-------
o f\j o ^* o zf o f^ '3O oo -0 ^ ^r ^ r*»
i
o
u
K1
«O ^^ O* ^"* NO fy *"^ *O O^ NO *™* t^ *™^ O1*
-*• ,TfUK1»HNO»-ifUrOrU CO «•«
a o *-i co »•«
rvi r»- o &• t^ in
II NorvjNONi irururur-^o«-iNO5TJOru
to iii-i>H I I =r —< —« I —< I l—i
to I I I I I
ru
l
II
00
o
X X
N3-ooor~Ok
2T r^-NOiro j^coNoo.somor^»r**-oxNO3'
O ON. QN, ^^ ^p^ ^^ ^N JQ gjj f^ f^. ^Q -Q ,plf n^ f^ ^Q
tj inrutoro ir»-"fUenNico«-ti-ii-< o
OO
ru
ru
i •
NO
X U3
3 3
O— CQ
O •<
^
>2 Q
Ctf Cc£
O E
^3 —^
b 3
O Q
a
o
I a
O *F«« O 5T »•< O^ ^ *•*
ro
5T ^^ 3O
tO «T >O
^^ ^^ ^O
ru *\ «-<
—< ru KI ru
cotJL^lO•^«-^ooxoxln^u'7•coof^J'^Jco
NO cr o o-
-H »-< ru
mNO=TO-ocoi-iNoruo
^- -H ru •rt ru
N-< ru
NO
rvi
f\j
J ^-i x
3 "^^ (D
tO 2
C
s
••* r*- NO rO i-*
O sO 9- -O in
in NO in oo i^ o
NCoruruNT»^o^-in
,*,*_,_ ru ru in
••H coeC-fV •^^•^^STNO >-< »i
in
CN
tu
ca
un ^^ ^r "O ^s ^^ 5T~ ^^ ^D ^^ ^^ ^*
t---<-Ttni-'r')rut>»r'iru«-i(\j
-« N> in CD
w in cc r- •- ru i
•^ ru ru
ru
ru
o ru MI ^u
1-1 i-< ru
ru=rru^ruru ^ —•
^*« f^ •*^ -O flC *•* *^ ^^ ^^ ^^ ^U -3" ^?
•»^ (\i AJ CC *-» >O f\I
\o
pa
to
oJ
200
-------
O
O
CO
o
o
o
CM
O
CO
a
0
>-i
CM
PQ
a)
u
O
H
in
g 0
O vO
O
. j
CM
6 o
O —
i™H G\
o
CM
,_,
H o
m vo
r-l
r-
o
r-.
00
[
CM
O
"-^
-*
O
r-.
CM
r-l
0
CM
CM
oo
-*
0
m
o
t [
•*
0
^^.
ON
ON
: r^
CO
o
^
•*
vo
-------
62.4, respectively) because of the narrower simulated crosswind distribution
of the CFBr about the mean direction.
It is obvious that the CTDM(11083-E) calculations would have better
simulated the observations if the Tower B 10-m winds had been used. Photo-
graphs and visual observations of the oil-fog plume during other experiments
suggest that during highly variable wind situations aged plume material is
often transported from the south (or the north) over the sampling array.
This is probably what is happening during experiment 11 (0200-0300). This
type of "sloshing" phenomenon is not handled by the current models and
should be considered during subsequent model development efforts.
Because of the uncertainty in modeling the CF3Br data with the wind
directions based on Tower A data interpolated to release height, a second
series of model runs was made with the bi-variate Gaussian form of
CTDM(11083-E) , assuming that the mean wind was directed from the source to
the area of greatest observed concentrations on the ridge. These changes
tend to place the peak of the horizontal distribution closer to the
stagnation wind direction assumed in the model for material below Hc, and
they place the peak of the distribution for material above Hc at those
receptors with the greater observed concentrations.
As expected, these changes cause peak modeled concentrations to
increase considerably in many cases. The net effect can be seen in the
overall performance statistics. When the results of modeling the CF3Br
data with the bi-variate Gaussian form of CTDM(11083-E) and the modified
mean wind directions are combined with the previous model results for SF6,
the mean bias (mg) of the residuals of peak concentrations is nearly 1.0,
although the noise remains large (sg = 4.3, sa = 75.6). The resolution
improves considerably: ra (peaks) drops from 4.6 to 1.5; ra (all
concentrations) drops from 4.7 to 2.2; and rg (peaks) drops from 1.6 to
1.2. This exercise serves to demonstrate that there is plenty of latitude
for improving the performance of CTDM at HBR.
In summary, the results of the initial HBR modeling are tentative.
None of the models reproduce the observations very well and no model is
better than any other. Several new issues must be considered in simulating
dispersion below l^ at the Hogback.
202
-------
SECTION 6 .
SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS FOR FURTHER STUDY,
This Third Milestone Report has documented the evaluation of several
complex terrain models using the complete SHIS #1 MDA. The model performance
measures suggest that CTDM(11083-E) is significantly better than the other
models. The report also describes in detail the further progress made to
date in analyzing the SHIS #1 experiment-hours and in improving CTDM. It
provides an overview of the SHIS #2 at the Hogback and describes the pre-
liminary results from this field experiment as well as the modeling of a
17-hour subset of the HER data base.
6.1 Principal Accomplishments and Conclusions
The Small Hill Impaction Study #2
The Hogback Ridge field study has achieved its objective of extending
the modeling data base to include a detailed set of meteorological data,
tracer concentrations, lidar data and photographic data from a two-dimen-
sional ridge site. The field program has produced a set of about 179
tracer-hours for model testing, evaluation and refinement.
An initial analysis of the SHIS #2 data base suggests that the concept
of a critical dividing-streamline height is appropriate for stable flows
toward HER. Tracer gas released below Hj. disperses in a highly variable,
"blocked" flow and produces relatively large ground-level concentrations on
the windward face of the ridge. Tracer gas emitted above HC disperses in
a flow that travels over the ridge and produces peak ground-level concen-
trations near the ridge crest and on its lee side.
Like SHIS #1, SHIS #2 has verified the basic concepts of the experi-
mental design. The release of gaseous and visible tracers from a mobile
crane or fixed tower, using real-time meteorological data to guide the
selection of release locations and heights, has resulted in a data base that
covers a wide variety of dispersion conditions and concentration patterns.
The meteorological data from four towers were archived in real-time via
an onsite system of minicomputers. During the first few experiments, it
became apparent that the archived data contained noise. Several subsequent
modifications to the data system were successful in reducing this noise.
ARLFRO and ERT scientists are currently working on eliminating the noise
from all of the archived data. Initial editing suggests that most of the
noise can be removed although there are subtle high frequency effects that
203
-------
need further investigation. The latter effects will have no influence on
the mean meteorological data but may degrade the turbulence statistics and
covariances.
The SHIS #1 Modelers' Data Archive (MDA)
A Modelers' Data Archive (MDA) has been prepared and submitted to the
EPA Project Officer. It is intended to be used by modelers in simulating
concentrations at Cinder Cone Butte and is available from the Project
Officer. The MDA consists of meteorological data that were objectively
interpolated from the Tower A measurements specifically for the heights of
release of the SFg and CF3Br tracer gases. It also includes hourly emission
rates and ground-level tracer gas concentrations. The MDA is continually
evolving to include the meteorological data that are found to be most
representative of dispersion conditions at plume height. As plume photo-
graphs, lidar data, and meteorological data are analyzed further, the MDA
will be supplemented and changes will be submitted to the Project Officer.
The Complex Terrain Dispersion Model (CTDM)
The Lift and Wrap models reported in the Second Milestone Report have
been combined into one model, the Complex Terrain Dispersion Model. CTDM
represents the first step in developing a practical regulatory model. It
was designed to combine both "Wrap" and "Lift" flow and dispersion assump-
tions so that a plume near Hc would exhibit characteristics consistent
with both, and it was designed to provide a better method for treating
surface reflection of plume material in complex terrain. Two versions of
CTDM were prepared for testing: CTDM (11083) and CTDM (11083-E). 'E'
denotes an enhancement to the effective size of az over the terrain.
This enhancement is similar in some respects to the "terrain factor" used in
other models such as COMPLEX I and COMPLEX II.
Comparative Model Performance Evaluations
CTDM(11083-E), CTDM(ll083), Valley, COMPLEX I and COMPLEX II computa-
tions of 1-hour average SFg and CF3Br concentrations were compared with
measured values from 153 SHIS #1 tracer hours. The model performance
measures indicate that CTDM(11083-E) performs significantly better than the
other models. However, both CTDM(11083-E) and CTDM(11083) tend to under-
estimate the peak observed concentrations on average. Yet the bias of the
residuals based on predicted and observed concentrations paired in space and
time is lowest for CTDM(11083-E).
Valley and the two COMPLEX models tend to overestimate the peak observed
concentrations on average, and have higher biases than the two CTDM models.
The noise and resolution statistics for all the models are large and, in
short, there is still substantial room for model improvement.
Twelve hours from Experiments 201 and 210 were used to test certain MDA
modifications suggested by detailed case-studies, and also to gauge the
importance of simulating the 5-rainute sequence of meteorology in obtaining
1-hour average concentration estimates with CTDM(11083-E). The results
204
-------
indicate that the MDA modifications could significantly improve overall
model performance, but that simulating each 5-minute period of an hour does
not consistently improve the hourly concentration estimates.
Investigations of Plume Growth
Detailed analyses of the Tower A turbulence intensity data, lidar data,
and photographic data collected during CCB Experiments 201 and 210, as well
as comparisons of iz data with lidar measurements from the other experi-
ments, have shown that the following expression for az adequately
simulates plume growth upwind of CCB to distances of about 1 km:
a —
z
a t
w
t/2TT)°-5
Lt
where TL = y2/N.
The data suggest that y ^ 0.5, which is consistent with the Pearson
et al. (1981) estimate that y» a molecular exchange coefficient, could
range from 0.1 to 0.8.
CTDM Modifications and Testing
The case-study analyses of SHIS #1 Experiments 201 and 210 have
suggested three principal areas for modifying CTDM(11083):
• An enhancement in the size of the effective az over the hill
is required for the part of the plume that diffuses or is released
above Hc.
• The form of this enhancement must recognize the relative importance
of increased plume dilution and decreased streamline heights above
the surface, and it must include the effects of lee waves.
• A more nearly continuous transition between Wrap and Lift
computations is needed near Hc.
These model modifications were formulated and implemented to create
CTDM(14083). Because a portion of the formulation was designed on the basis
of the results of case-study analyses of Experiments 201 and 210, CTDM(14083)
was tested on the complete 153-hour MDA as well as on the twelve hours from
Experiments 201 and 210 so that an overall comparison of model performance
with CTDM(11083-E) could be made. The resulting measures of model perform-
ance (ma, sa, ra, mg, sg, rg) were nearly the same for CTDM(11083-E) and
CTDM(14083) so that CTDMC14083), when using the initial MDA, is not a
significant improvement over CTDM(11083-E). However, the comparison using
the twelve case-study hours indicated that the placement of peak concentra-
tions is improved significantly, even though there appears to be a need to
include some degree of plume dilution upwind of the hill in both the Lift
and Wrap modules.
205
-------
Preliminary Modeling of the SHIS #2 Data Base
CTDM(11083), CTDM(11083-E), Valley, COMPLEX I, and COMPLEX II calcula-
tions of 1-hour average SFg and CF^Br concentrations were compared with
concentrations obtained for 17 SHIS #2 hours (34 tracer-hours). The per-
formance evaluations show that none of the models simulates the observations
very well. CTDM(11083) and CTDM(11083-E) underestimate the peak observations
by roughly a factor of 2.0; the other models overestimate the concentrations
by about a factor of 4.0. The performance statistics for all models show a
great deal of noise.
Several questions arise from an analysis of the model simulations:
• How relevant is the concept of a single stagnation streamline (as
in Wrap)?
• Does the recirculation of tracer material from the south (or
north) of the HER target area affect measured concentrations? If
so, how can this phenomenon be modeled?
• How can meteorological data be selected to represent the highly
variable flow below Hc?
These questions must be addressed in the subsequent model development
efforts. It also should be emphasized that the modeling results are
preliminary because of the problems with the tower meteorological data.
6.2 Recommendations for Further Study
Further analysis of the SHIS #1 and SHIS #2 data bases are required to
improve and test CTDM and to extend it to ridge and other terrain settings.
These analyses must be supplemented by new fluid modeling investigations.
6.2.1 The SHIS #1 Data Base
The case-study approach to analyzing the CCB data should continue.
Experiments 201 and 210 have provided insight on refining the Lift modeling
approach. Because the higher x/Q values measured at CCB occurred on the
windward face, renewed emphasis should be given to analyzing the Wrap
modeling approach with data from experiments like 206 (see CTDM Second
Milestone Report). The case-study approach will also help update the entire
MDA by investigating photo- and lidar-derived wind directions and turbulence
intensity data.
More theoretical work and data analysis are needed to refine the
modeling of dispersion close to and above Hc. Numerical potential flow
modeling and physical modeling will help in formulating streamline distor-
tion effects. The case-studies will help modify and evaluate the resultant
formulation for a and T in the Lift module, and also the resultant changes
to the simulation of streamline distortion and plume dilution in the Wrap
module.
206
-------
CTDM must also be modified with a. view towards simulating flow and
dispersion around terrain features more complicated than CCB. We have seen
that successful modeling of the SHIS #2 data will'require some additional
reformulation of such basic components of CTDM as the stagnation streamline.
What terrain height should be used in calculating Hc for less ideal terrain
shapes? How does the "Hc surface" vary over rolling terrain between the
meteorological measurement site and the major terrain feature being .modeled?
Similar questions must be posed and answered within a larger framework for
CTDM so that the refined modeling concepts contained in it can be transferred
to other terrain geometries.
Closely allied to these modifications of CTDM are considerations of how
CTDM, as developed for SHIS #1 and #2, will be applied to the upcoming full-
scale study. Certain components of CTDM will be considered more important
on this scale than others. For example, with the relatively slow growth of
az for elevated plumes in stably stratified atmospheres, az may be
wholly determined by the initial buoyancy-dominated growth near the source
so that the proper specification of nonbuoyant plume growth over distances
of 1 to 2 km, which is so important in understanding the SHIS #1 data, may
have little relevance. Similarly, certain aspects of the formulation of
Lift designed to simulate concentrations over the crest and in the lee of
major terrain features may assume considerably greater importance than the
Wrap formulations if the plume is expected to be near or above terrain height
much of the time. The characteristics of new versions of CTDM should be
explored via sensitivity studies to understand how the model will perform on
larger scales.
The question of model sensitivity to the detail of the available input
data must also be addressed. The SHIS #1 and #2 data have underscored the
need for exceptional resolution in the meteorological data in order to
understand and model the observed concentrations. As CTDM becomes more
refined, it will necessarily become more dependent on the availability of
detailed input data. For practical regulatory application, parameterized
"cousins" of CTDM must be developed to retain the more important features of
CTDM while reducing the input data requirements. This process will not only
yield a version of CTDM capable of producing "screening calculations"; it
will also help define the absolute minimum information needed for "refined"
modeling within the current regulatory permitting framework.
6.2.2 The SHIS #2 Data Base
Before significant progress can be made in modeling HBR, segments of
the SHIS #2 data base must be studied and intercompared. The work falls
into two principal categories: refining the measured data and assessing the
representativeness of that data for model-ing the tracer gas plumes.
Four types of noises have been identified in the tower meteorological
measurements taken at the Hogback. Considerable additional work will be
required to filter these noises from the data and to refine the meteorologi-
cal data base. The so-called 60-HZ and 2-to-4-sec noises have yet to be
characterized and no methods have been developed to eliminate them from the
data.
207
-------
A necessary initial step in the data refinement process is to obtain
more propeller response and F460 cup-and-vane response characteristics from
wind tunnel tests, especially at low-wind speeds. We recommend using the EPA
FMF wind tunnel to perform these tests. Another necessary step is to compare
the sonic, propeller and cup-and-vane anemometer data. A careful evaluation
of the turbulence statistics from the sonic vs. propeller anemometers is
especially important for modeling. Once the instrument response character-
istics are better understood, the data can be adjusted to the extent pos-
sible, and the overall accuracy of the various measurements can be assessed.
The issue of data respresentativeness must be addressed if we are to
construct a reasonable MDA for SHIS #2. Many pieces of data are available
to help us construct a detailed spatial and temporal picture of the varia-
bility in the flow.
Temperature data from the three main towers and the tethersondes must
be integrated to create a cross section of the vertical structure of the
flow upwind and over HBR, and this structure must be compared with the
acoustic sounder records. This structure will also be particularly useful
in modeling the plume rise of the jet-fogger plume containing the SF6
tracer gas.
Wind data from the doppler sounder well upwind of HBR should be studied
to characterize the large-scale incident flow, and the wind data from all of
the towers, the tethersondes, and the wind directions obtained from images
of the smoke plume will be needed to understand the relationship among the
vertical temperature structure, the incident flow stucture, and the wind
variability measured just upwind of and over HBR. It is the flow and
turbulence in this zone that determined the concentration patterns measured
at HBR.
With a refined data base, and a rudimentary understanding of the forces
affecting transport and diffusion of plume material upwind of HBR, the case-
study integration of SHIS #2 data can proceed. Key elements in successfully
modeling the observed tracer gas concentrations include:
• extending the latest version of CTDM(14083) to HBR,
• verifying plume rise and CTZ development algorithms with plume
photography and lidar data,
• creating the HBR MDA, and
• summarizing spatial and temporal patterns in measured
concentrations, with special emphasis on the 10-minute data and
the concentrations measured on Towers B and C.
208
-------
REFERENCES
Burt, E.W. 1977. Valley Model User's Guide. EPA-45Q/2-7-7-018. U.S.
Environmental Protection Agency, Research Triangle Park, NC.
Cline, A.K. 1974. NCAR, Comm. ACM _1_7, 4, p221.
Crow, L.W. 1975. "Meteorological Data Analysis Based on Monitoring
Stations and Meteorological Data, January - December 1974. Joint
Environmental Program," No. 153, Loren W. Crow Associates,
Denver, Colorado.
Environmental Protection Agency 1981. "Regional Workshop on Air
Quality Modeling: A Summary Report," Source Receptor Analysis
Branch, OAQPS, Research Triangle Park, NC.
Fox, D.G. 1981. Judging Air Quality Model Performance: A Summary of
the AMS Workshop on Dispersion Model Performance. Bull Am.
Meteorol. Soc. . 62: 599-609.
Gifford, F.A. 1980. Smoke as a Quantitative Atmospheric Diffusion
Tracer, Atm. Env. , 14: 1119-1121.
Greene, B.R., and S. Heisler 1982. EPA CTMD Quality Assurance Project
Report for SHIS #1. ERT #P-B348-350.
Halitsky, J. 1961. Single Camera Measurement of Smoke Plumes. Int.
J. Air & Water Poll. , 4_: Nos. 3/4, 185-198.
Holzworth, G.C. 1980. The EPA Program for Dispersion Model Development
for Sources in Complex Terrain. Second Joint AMS-APCA Conference
on Applications of Air Pollution Meteorology, New Orleans. LA.
Horst, T.W. 1973. Corrections for Response Errors in a Three-
Component Propeller Anemometer. JAM, 12: 716-725.
Hovind, E.L., M.W. Edelstein, and V.C. Sutherland 1979. Workshop on
Atmospheric Dispersion Models in Complex Terrain. EPA-600/9-79-041.
U.S. EPA, Research Triangle Park, NC.
209
-------
REFERENCES (Continued)
Hunt, J.C.R. and R.J. Mulhearn 1973. Turbulence Dispersion from
Sources Near Two-Dimensional Obstacles. Journal of Fluid
Mechanics 61_: 245-274.
Hunt, J.C.R. and K.J. Richards 1982. Stratified Shear Flow over Low
Hills. To be published.
Lavery, T.F., A. Bass, D.G. Strimaitis, A. Venkatrara, B.R. Greene,
P.J. Drivas, and B. Egan 1982. EPA Complex Terrain Model
Development Program: First Milestone Report - 1981.
EPA-600/3-82-036, U.S. EPA, Research Triangle Park, NC, 304p.
Londergan, R.J. 1980. Validation of Plume Models: Statistical Methods
and Criteria. EA-1673-54. Electric Power Research Institute,
Palo Alto, CA.
Moore, G.E., R.G. Ireson, C.S. Liu, R.E. Morris, A.B. Hudischewskyj,
and T.W. Tesche 1981. Air Quality and Meteorology of
Northwestern New Mexico, Draft Final Report No. 81203. Arizona
Public Service.
Pearson, H.J., J.S. Puttock and J.C.R. Hunt 1981. A Statistical
Model of Fluid Element Motions and Vertical Diffusion in a
Homogeneous Stratified Turbulent Flow. J. Fluid Mech. 129;
219-249.
Rhoads, R.G. 1982. "Addendum to Workshop Summary Report," OAQPS,
U.S. EPA, Research Triangle Park, NC.
Sheppard, P.A., 1956. Airflow over Mountains, Quart. J. Roy. Meteor.
Soc., 82_: 528-9.
Smith, R.B. 1980. Linear Theory of Stratified Hydrostatic Flow Past
an Isolated Mountain. Tellus, 32: 348-364.
Snyder, W.H. and J.C.R. Hunt 1983. Turbulent Diffusion From A Point
Source in Stratified and Neutral Flows Around A Three-Dimensional
Hill; Part II, Laboratory Measurements of Surface
Concentrations. (Submitted to Atmospheric Environment).
Strimaitis, D.G., A. Venkatrara, B.R. Greene, S.R. Hanna, S. Heisler,
T.F. Lavery, A. Bass and B.A. Egan 1983. EPA Complex Terrain
Model Development Program: Second Milestone Report - 1982.
EPA-600/3-83-015. U.S. EPA, Research Triangle Park, NC, 375p.
Venkatram, A. 1982. A Framework for Evaluating Air Quality Models.
Boundary-Layer Meteor., 24: 371-385.
210
-------
REFERENCES (Continued)
Williamson, H.J. and R.R. Krenraayer 1980. Analysis of the
Relationship Between Turner's Stability Classifications and Wind
Speed and Direct Measurements of Net Radiation. Second Joint
Conference on Applications of Air Pollution Meteorology, March
24-27, AMS, Boston.
211
-------
APPENDIX A
LABORATORY SIMULATION OF NEUTRAL PLUME
DISPERSION OVER CINDER CONE BUTTE
212
-------
LABORATORY SIMULATION OF NEUTRAL PLUME
DISPERSION OVER CINDER CONE BUTTE
Comparison with Field Data
by
Roger S. Thompson,
William H. Snyder*
and .Robert E. Lawson, Jr.*
Meteorology and Assessment Division
Environmental Sciences Research Laboratory,
Environmental Protection Agency
Research Triangle Park, NC 27711
April 1983
*0n assignment from the National Oceanic and Atmospheric
Administration, U.S. Department of Commerce.
21S-
-------
INTRODUCTION
The purpose of this series of experiments was to duplicate, in the
laboratory, the field experiments performed by Environmental Research &
Technology, Inc. (ERT) at Cinder Cone Butte, Idaho. In this particular
study, a one-hour period (1700-1800 MST) of Case 202 was modeled.
Meteorological measurements during this period indicated that atmospheric
conditions were neutral. The laboratory study was performed in the
Meteorological Wind Tunnel, a facility well-suited for modeling neutral
atmospheric flow. Measurements of vertical profiles of wind speed,
ground-level concentrations, and vertical and lateral profiles of
concentration were made during the wind tunnel experiments. This report
describes the conduct of the study, makes comparisons of field and
laboratory measurements, and discusses some additional observations of the
laboratory study.
214
-------
SUMMARY OF FIELD OBSERVATIONS
The experimental data collection network used in the field has been
described by Lavery, et al. 1982. For the one-hour period considered here
(Case 202, October 17, 1980, 1700-1800 MST), the meteorological data were
fairly complete. The concentration data contained only nine, non-zero,
one-hour average concentrations; however, an additional 14 non-zero,
one-hour values were computed from ten-minute concentration data. There
were five reported zero one-hour concentrations and an additional ten zero
values computed from the ten-minute data.
Meteorological Data
Examination of the NOAA Daily Weather Map series for October 17 and
18 showed synoptic conditions at Cinder Cone Butte (CCB) during Case 202
to be dominated by a weak high pressure system centered over southern
Washington. Skies were clear with surface winds light and variable during
the morning. Surface temperatures ranged from a low ,.of about 1°C to an
afternoon high of about 12°C. Upper level winds (500 mb) were from the
north at 10-12 m/s.
Under these synoptic conditions, atmospheric stability is dominated
by surface heating and cooling. Stable conditions prevail during the
early morning hours due to formation of a radiation inversion and, as
surface heating increases during the day, there is a transition to
near-neutral and finally neutral or slightly unstable conditions. During
the late afternoon and evening, the,reverse process takes place as
insolation decreases and surface radiation begins to dominate. The hour
examined in this case study (1700-1800 MST) falls just prior to sunset
(1800 MST) and, on the basis of these synoptic conditions, would be in a
transition from near-neutral conditions to stable conditions. Examination
of the detailed meteorological data from CCB shows that this trend is, in
fact, taking place as the stability is initially neutral (D stability) and
becoming more stable (E stability) toward the end of the hour.
Wind speed and wind direction.were measured and recorded at five
levels (2, 10, 40, 80 and 150 m) and temperature at eight levels (2, 10,
20, 40, 60, 80, 100, and 150 m) ori the tower (Tower A) located 2 km north
of CCB. These data are summarized in Table A—1. Five-minute averages
were recorded at the time corresponding to the end of the period. The
table contains the u (westerly), v (southerly), and w (upward) components
of the wind as determined from the u,v,w anemometers in m/s. Computed
from these are the five-minute averages of wind speed (WS), wind direction
(WD), along-wind turbulence intensity (IX), cross-wind turbulence
• 215
-------
1
o
o
CO
M S3 in N-tD N —• K> CD CD in CVJ ID
« —' N- VD CD IO N- f in CO M" fO O
tonoj^f. N to to CM <\J CM PJ
O
O
o
oo
en
>- uj — in i*- m en m.(D TT ori oo —
-
>- in in ro ro VD CD —• G) in N-m OD
—• CD to iri r^ ^r — i^ CM en PJ — en
—-N-
ai
U3
oa
X N-criNf t^M to TT CD ^r inn
•-• in co m \n s> *f en \n en N- CD —
(MOJ—D — iniDDV—-U3CD
— VD N. en ED to ID t\HD M in a? ID
' — 6D
CA—'t^N-IO —
'
CM
o
sa
u
§
— (M — -« -• ID IO
tototototontoontoto
rn -
3 \r in — to —
I
I
. CD ID h^ K-K. K. N-
10 O CD D O D CD (D 03 63 O O
cniococncncMcnincoiMioco
Nntncoco — CDTIO —N-^r
tONCVINWIOPJIOIOIOCMN
I I I I I I I I I I I I
ininiDinioiDinniotDiD
3 K) CM ID CO CD ID N CD CD CD ID
CMIM — CM — — IMIM—•-« ~*
i i i i i i i i i i i i
§
O
e
<
la
E
OJ
ti
M
CD ID niO — D ID ID IO (O — tO
ea —NIO ^r in n ID (^ CD m ea
CO — Cn IO tDUDMCTilMMN-CO
E
O
uj
C
co ID n to -- eo co ID in 10 — to
o — • M to v in in 10 h- CD en to
E
O
111 _
MHO
216
-------
o
o
00
1—I
I
o
o
M CQCnEI — r
—• CD CDin \T (M El ED — N-~ ~ in
•tf CO O U CO T. ivi in (vi-— locvicncoinio^rm
—" tvj co — co m in m BD — (vi en to
*-• —• en —' in f^*in ^r ^3" v UD en N-
o
u
O O CD CO E) CD GD G) E) E) O E)
ca
u
o
X tr ro in r^ M — M T m M t^ in
«— (nco—-iDCnuMDM — trxri —
'
—•—'—^—«03ln
-> — en—•
00 tnco •tf — UD
—' E) G> ^^ E)
ISI
(M
El
O
(M
O
co
Q
ID en en UD en — — — — ro r^ en
- fo CM po M tvi r>j oj (M rM — en co
s »
tn ^r to to n N- m 10 in N- EI i^ — •
3 CD o co in o tvj -• — in »D o en
co —
CDcnoDCOCDCDcricriOcncricn
— ro rv in en in UD
co
8
I
1
-] 3
O S
CH -H
O -P
W p!
W U
O
CO
in^rrjininnnnnnnn
i i i i t i i i i i i i
PJ
rOUDSDIOUDDIOIOSDtOlDD
UJ CD ID niO — D CO ID n 10 — D
H BJ — wnv n n 10 K. CD tn D
'yncoiMininincococnMco
inninninininiDiDiriiniD
i i i i i i i i i i i i
enncnionTcoionco — —
— inwioco — in— •
o
in
n
N yj coionio — coibioinio — o
Q) r~ N- K- K- f^ N- N- ffc. ^ K_ N- K- CD
CD
CVJ
II
N
E
CD
*
\\
N
-^ ro in in ro in h- r^ cn co — •
co GI c\i o\i GD cvj ^~ ^3" LO r*— r\i
^D LD in in ^D in ^~ "^T to o^ co
_. _
CO G) CO CM — • ^3" CO
rsi — (D ro r^ in ,r\t
\u in ^o ^3" ^" fo r^
— -• — — -* — E3
CSJ
II
N
03
S_
CD
O.
coujinto
to
-Eicou)inro
. 217
-------
intensity (IY), and vertical turbulence intensity (IZ) in percent.
Table A-lf contains the five-minute average temperatures (°c) recorded for
all eight levels.
One-hour average values of the meteorological parameters were
computed from the available five-minute averages. A one-hour average wind
vector was computed at each of the five levels of the tower as follows:
U
N
1/2
6 = arctan
where u^, v^ are the values of the westerly and southerly wind
components for the ifch five-minute period and there are N periods
containing valid data for the hour. The hour—average values of turbulence
intensity were calculated in the direction of the hourly-mean wind, in the
direction horizontally perpendicular to the mean wind direction, and in
the vertical direction. The five-minute average turbulence intensities
were projected onto these directions and the variance of the five-minute
mean wind data was added in computing the hourly turbulence intensities.
fi? r<
N i=lU-
IX.WS.COSA9.
100 )
IY.WS.sinA9. „
,(•11 1)2
k 100 '
(WS.cosAG .)2_|- U2
U2
1
2
x 100%
IX.WS.sin,^ .
11 A9i
100
lY.WS.cos.
i i A
100
U2
x 100%
"IN IZ.WS ~
» y r L h2
N 4 . ( 100 }
1=1
u2
!_
2
x 100%
where the subscript i again refers to the i five-minute period of the
hour and A9 . = 9 . - 9 .
i i
218
-------
The one-hourly average values for each level of Tower A are contained
in Table A-2. Vertical profiles.of the five-minute average wind speeds
are presented in Figure A-l. The one-hour average wind speed values are
also indicated in Figure A-l and on a log-linear scale in Figure A-2.
From Figure A-2, zo was found to be on the order of 5 to 8 cm. The
boundary-layer thickness appears to be greater than 150 nu Vertical
profiles of the five-minute and one-hour average temperature data are
presented in Figure A-3. A line with the slope of the adiabatic lapse
rate (rd) is included in the figure. The environmental lapse rate
closely follows the adiabatic lapse rate above about 10 m from the
surface. This, together with the previously discussed wind data,
indicates neutral stability for this period.
TABLE A-2. ONE-HOUR AVERAGE METEOROLOGICAL DATA FROM TOWER A
z
U)
2
10
20
40
60
80
U
(m/s)
4.39
5.88
7.89
8.79
Q
(°)
312.6
313.7
316.5
308.2
(316.2)*
1
(%)
18.1
13.6
8.1
7.1
2
(%)
14.1
11.7
8.0
6.4
3
(%)
3.7
3.7
3.0
3.2
100
150
9.09
309.4
(316.4)*
7.7
6.1
3.1
11.21
11.53
11.41
11.31
11.21
11.08
10.85
10.40
*The anemometers at z = 80 and 150 m were oriented approximately 8°
and 7° east of true north, respectively (Lavery, et al. , 1982).
The values in parentheses are corrected for this misalignment.
After the design and data collection of this wind tunnel experiment
were completed, ERT prepared and released their First Milestone Report
(Lavery, et al., 1982) on the field project. The quality assurance audit
reported misalignment of the wind systems at the 80 m and 150 m heights on
Tower A by 8° and 7° east of true north, respectively. Adjustments for
these have been made in Table A-2, but not in Table A-l.
Concentration Data
The tracer experiment in the field was designed to examine
ground-level concentrations from a point source upwind of an isolated
hill. For Case 202, the source was located 1.0 km from the hill center at
319° from true north. Sulfur hexafluoride (SFg) was released at a
height of 50 m above the local ground surface. Examination of the
topographical maps revealed that the local surface at this location was
8 m below the reference height selected as the base of the hill.
219
-------
200
CO
oc
LLJ
J—
ULJ
21
150
100 -
SO -
-i •>- » ,--....-: .} .-, {
average for hour
SPEED (H/S)
Figure A-l. Tower A, Case 202, 1700-1800, 5-minute wind speed values and
hourly average wind speed as a function of height above ground.
220
-------
200
Nl
Figure A-2. Vertical profile of hourly average wind speed measured at
Tower A, Case 202, 1700-1800.
221
-------
200
r. - 9.8 °C/1000 m'
J..4.4.J...M..AIJ
I ! 19 average for hour
M I I I I I
15
Figure A-3. Vertical profiles of five-minute averages of temperature at
Tower A, Case 202, 1700-1800. The one-hour average values and
a line with slope corresponding to the adiabatic lapse rate
are also shown.
222
-------
Time spent in moving the crane supporting the source into this
position resulted in delay of the source beginning to emit until 1716,
sixteen minutes into the hour of interest. For the remainder of the
hour, the source emission rate was 0.080, ( + . 5.3%) g/s of SF6.
on thpn at a network of 100 sampling stations
?able A 3 TV^'i i°Cations of these stations are given in
Fit A / k°th P°lar and rectanS"lar coordinates as defined in
Figure A-4. Sample ports were located on the model at corresponding
positions and assigned identification numbers also listed in Table A-3.
A summary of the field concentration data is presented in
fnt^h , one-hour sample data required an adjustment to correct
for the portion of the hour that the source was not operating and the
sample pumps were. The ten-minute sample data indicate that the
tTfrh8 the SamPl6rS ^^ the be§inning °f the period 1720
that the concentrations measured during this period were
roughly equal to those for the period 1730 to 174?. Therefore the
measured "one-hour concentrations" were multiplied by 1.5 (6
-------
TABLE A-3. LOCATION OF SAMPLERS ON CINDER CONE BUTTE
FMF
• ERT
PORT PORT
1
2
3
4
5
6
7
8
9
10
n
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
37
37
37
37
37
37
37
37
37
37
37
37
37
34
31
31
31
31
31
31
29
28
28
27
26
26
26
63
62
60
60
60
60
60
60
60
60
60
58
58
57
57
57
57
57
56
56
56
55
55
.01
.02
.03
.04
.05
.06
.07
.09
.10
.11
.12
.20
.26
.10
.90
.96
.07
.09
.26
.05
.04
.26
.11
.03
.03
.05
.06
.03
.03
.03
.04
.22
.23
.24
.25
.26
.01
.02
.08
.07
.25
.26
.96
.13
.90
.03
.04
.05
.05
,13
ANGLE
DEG
8
8
8
8
8
8
8
8
8
8
8
8
8
30
52
53
53
53
53
53
68
71
75
83
90
90
90
98
105
120
120
120
120
120
120
120
120
120
135
137
143
143
143
143
147
150
150
150
158
158
•
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
R
m
510
449
391
308
275
250
277
177
154
122
81
193
19
146
146
154
200
154
107
246
187
420
75
260
360
208
142
315
190
362
297
235
164
124
94
43
510
410
195
240
327
140
165
165
162
255
198
86
250
204
X
m
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
' 70.
62.
54.
42.
30.
34.
31.
24.
21.
16.
11.
26.
2.
73.
115.
122.
159.
122.
85.
196.
173.
397.
72.
258.
360.
208.
142.
311.
183.
313.
257.
203.
142.
107.
81.
37.
441.
355.
137.
163.
196.
84.
99.
99.
88.
127.
99.
43.
93.
76.
98
49
42
87
27
79
59
63
43
98
27
86
64
00
05
99
73
99
45
46
38
12
44
06
00
00
00
93
53
50
21
52
03
39
41
24
67
07
88
68
79
25
30
30
23
50
00
00
65
42
505
444
387
305
272
247
224
175
152
120
80
191
18
126
89
92
120
92
64
148
70
136
19
31
0
0
0
-43
-49
-181
-148
-117
-82
-62
-47
-21
-255
-205
-137
-175
-261
-m
-131
-131
-135
-220
-171
-74
-231
-189
Y
m
.04
.63
.19
.00
.32
.57
.79
.28
.50
.81
.21
.12
.82
.44
.89
.68
.36
.68
.39
.05
.05
.74
.41
.69
.00
.00
.00
.84
.18
.00
.50
.50
.00
.00
.00
.50
.00
.00
.89
.52
.16
.81
.78
.78
,87
.84
.47
.48
.80
.14
224
-------
TABLE A-3. LOCATION OF SAMPLERS ON CINDER CONE BUTTE (Continued)
FMF
PORT
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
ERT
PORT
55
55
55
54
53
53
39
39
39
39
39
39
39
39
39
39
45
45
45
48
48
48
48
48
48
50
64
64
65
67
67
67
67
67
67
67
67
67
67
69
69
70
70
70
72
72
73
73
74
74
.96
.26
.90
.04
.03
.04
.21
.01
.02
.03
.05
.07
.12
.17
.18
.20
.24
.07
.10
.12
.02
.03
.05
.07
.10
.07
.05
.10
.03
.01
.02
.04
.05
.06
.13
.23
.24
.25
.26
.07
.26
.08
.26
.25
.23
.24
.10
.11
.03
.04
ANGLE
DEC
158
159
159
165
171
173
187
187
187
187
187
187
187
187
187
187
232
232
232
255
255
255
255
255
255
270
277
277
285
300
300
300
300
300
300
300
300
300
300
315
315
322
322
333
338
338
345
345
353
353
•
•
•
•
»
•
a
#
t
%
B
•
*
.
m
•
•
.
.
.
t
^
•
m
.
m
m
m
.
.
•
.
•
.
.
.
•
9
m
^
f
.
•
m
B
9
*
f
t
•
.
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
R
m
204
165
196
301
139
260
37
510
411
348
291
251
115
180
150
77
107
224
158
63
411
327
258
211
150
238
282
159
252
510
416
343
311
284
230
247
X
m
.0
.0
.0
.0
.0
.0
.0
.0
0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
178.0
130
54
200
325
151
108
275
285
211
127
105
310
197
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
76
59
70
77
21
31
-4
-62
-50
-42
-35
-30
-14
-21
-18
-9
-84
-176
-124
-60
-397
-315
^249
-203
-144
-238
-279
-157
-243
-441
-360
-297
-269
-245
-199
-213
-154
112
-46
-141
-229
-92
-66
-124
-106
-79
-32
-27
-37
-24
.42
.13
.24
.90
.74
.68
.51
.16
.09
.41
.47
.,59
.02
94
.28
.38
.32
.52
.51
.85
.00
.86
.21
.81
.89
.00
.90
.81
.41
.67
.26
.04
.33
.95
.18
.91
.15
.58
.77
.42
.81
.96
.49
.04
.76
.04
.87
.17
.78
.01
Y
m
-189
-154
-182
-290
-137
-258
-36
-506
-407
-345
-288
-249
-114
-178
-148
-76
-65
-137
-97
-16
-106
-84
-66
-54
-38
0
34
19
65
255
208
171
155
142
115
123
89
65
27
141
229
118
85
245
264
195
122
101
307
195
.15
.04
.98
.74
.29
.06
.72
.20
.94
.41
.83
.13
.14
.66
.88
.43
.87
.91
.27
.30
.37
.63
.77
.61
.82
.00
.37
.37
.22
.01
.00
.50
,50
.00
.00
.50
.00
.00
.00
.42
.81
.99
.11
,03'
.25
.64
.67
.42
.69
.53
225
-------
Figure A-4. Map of Cinder Cone Butte showing coordinate systems and
locations of the 100 ground-level samplers.
226
-------
TABLE A-4. GROUND-LEVEL CONCENTRATIONS FROM FIELD EXPERIMENT
a) One-hour sample data
FMF Port
Number
10
13
69
73
31
34
35
81
82
86
87
88
89
99
b) Ten-minute sample data
Measured
Concentration
(ppt SFfi)
189.0
70.0
0.0
0.0
110.0
5.0
61.0
0.0
0.0
0.0
7.0
11.0
26.0
124.0
Adjusted
Concentration
(ppt SFfi)
283.5
105.0
0.0
0.0
165.0
7.5
91.5
0.0
0.0
0.0
10.5
.16.5
39.0
186.0
FM Port
Number
4
5
6
7
9
11
18
19
20
32
52
60
61
64
68
71
74
75
76
79
88
90
95
97
100
Measured Concentration in ppt of SFg
for Ten-Minute Period Ending at
1710
-
-
0
5
-
0
0
0
0
0
0
-
-
'
-
0
-
6
0
0
0
0
-
0
210
1720
0
-
-
-
67
0
0
0
0
0
-
-
-
0
0
0
-
0
0
0
0
0
-
0
194
1730
100
-
265
-
266
119
-
114
101
27
0
-
-
0
0
0
-
-
0
0
. 31
52
.-
0
0
1740
117
87
-
0
352
144
-
178
99
116
0
-
-
0
0
0
-
5
0
0
5
'6
319
- - .
0
. 1750
19 '
-
117
0
116
244
251
294
18
220
0
0
0
0
-
0
0
0
0
0
15
65
490
0
. - .
1800
0
-
0
0
405
-
176
387
8
-
—
0
0
0
-
0
-
-
0
0
47
150
,497
250
35
Average
for
1720-1800
59
87
127
0
285
169
213
243
57
121
0
0
0
0
. 0
0
0
3
0
0
25
68
435
83
— "&
*Data for port 100 were ignored because of high reported
concentrations for 1700-1720 when source was not operating.
227
-------
EXPERIMENTAL DETAILS
Model Specifications
The model was constructed of acrylic plastic by a vacuum-forming
technique. The nominal scale was 1:640; after fabrication, the actual
model scale was determined to be 1:647 in the horizontal and 1:694 in the
vertical (model height = 14.4 cm). A survey of boundary-layer generation
systems used in previous studies in the Meteorological Wind Tunnel
revealed that a boundary layer produced, using 92 cm high vorticity
generators, a 15 cm high fence, and a fine "Sanspray" surface would
simulate the full-scale situation. The model was coated with gravel of
the same size as that used in the Sanspray (average diameter on the order
of 2 mm) and installed in the wind tunnel, along with the boundary-layer
generation devices (see Figure A-5). The center of the hill was located
8.0 m downwind of the fence. The wind tunnel was operated at a freestream
air speed of 4 m/s.
The source was constructed from a 0.32 cm O.D. brass tube that was
bent over to emit the tracer in the direction of the wind and avoid any
momentum plume rise. Ethylene (99.8% pure) was released at a rate of
50 cm-Vs, so that the effluent speed was approximately equal to the
local mean air speed. Sampling ports were made using 0.24 cm O.D. brass
tubing. The ports projected approximately 2 mm above the surface
roughness, except for the mast mounted samplers, which were scaled in
height in proportion to their field counterparts. The sample ports were
connected to a scanning valve which allowed groups of 5 sampling ports to
be connected to the laboratory's system of 5 flame ionization detectors
(FID) for determination of tracer concentration. The FID's were
calibrated before and checked after each experimental run of 100 samples.
The background level of ethylene in the room was measured before and after
each run and a linear interpolation algorithm was used to subtract the
background from'each sample. The analog output of the FID's was processed
on the laboratory's minicomputer system using a digitizing rate of 1 Hz
for 120 seconds, an adequate sampling time to obtain stable averages.
Conversion of Model Concentrations to Field Concentrations
To facilitate comparison of the field and laboratory data, the
laboratory data were converted to equivalent full-scale values. Model
concentration values were recorded in percent by volume of ethylene. The
minicomputer program for collection and analysis of wind tunnel
228
-------
Figure A-5. Cinder Cone Butte model in the EPA Meteorological Wind Tunnel.
Note vortex generators at test section entrance.
229
-------
f\
concentration data calculates a nondimensional concentration x = CUHZ/Q,
where C is the measured concentration, U is the reference wind speed, H is
the hill height, and Q is the tracer release rate. The absolute wind
speed in the wind tunnel is arbitrary. However, to convert wind-tunnel
tracer measurements to field values, a reference wind speed, at the same
locations in the model and field, must be selected and used in computing
X» The reference for this study was selected as the wind speed measured
on Tower A at z = 40 m. These nondimensional concentrations were then
used to convert the model concentrations to field values using equivalence
of the nondimensional concentrations for model and field (xm = Xf
- x)« This results in:
cf -
xmQf/ufHf
where the subscripts ra and f stand for model and field, respectively.
Substituting in the field values for the period of this study gives
C(ppt)
X(0.080 gm SF6/s) (1Q12 ppt)/(6.5xlQ3 gm
(7.89 m/s) (100 m)2
or C(ppt) = 156x-
230
-------
PRESENTATION AND DISCUSSION OF RESULTS
For the sake of easy comparison, all wind-tunnel measurements have
been converted to full-scale values according to the scaling concepts '
previously discussed.
Wind Field Measurements
Verification that the boundary layer developed in the wind tunnel was
representative of that observed in the field was based on comparison of
wind measurments at Tower A in the field with the vertical profile
obtained in the wind tunnel upstream of the model. As mentioned above,
the absolute wind speed in the wind tunnel is arbitrary. Adjustment of
the wind-tunnel velocity profile to field values was based on matching the
speeds for the model and field at the 40-m level on Tower A, the
measurement nearest the height of the source. The mean wind speed
comparison, shown in Figure A-6, is quite good with the exception of the
measurement at z = 80 m on Tower A. Also presented in Figure A-6 are the
data for the Tower B site located on top of the south peak of Cinder Cone
Butte. The mean velocity measurements in the wind tunnel for this site
are also in good agreement with the field data.
Vertical profiles of components of turbulence intensity were also
computed from the wind tunnel data for the two locations discussed above.
The longitudinal (direction of the mean wind) component and lateral
(perpendicular to the mean wind) component of local turbulence intensity
are shown in Figures A-7 and A-8, respectively. The vertical component
was not measured in the wind tunnel. Overall, the turbulence levels in
the wind tunnel are somewhat larger than the field values. The
longitudinal values are about a factor of two greater and the lateral
values are about one and a half times as large. Errors introduced by the
excessive turbulence in the wind-tunnel flow, can be estimated for flow
over flat terrain. Pasquill (1962) has pointed out that for a given plume
cross-section, the maximum ground-level concentration is determined by
only the ratio of the vertical spread to the lateral spread of the plume.
Thus, the magnitude of the maximum ground-level concentration should be
correct. The change in the location of the maximum can be estimated by
assuming plume spread to be proportional to turbulence intensity. For
near-neutral conditions and a source height on the order of the height for
this study, a 50% excess in turbulence intensity can be expected to result
in the maximum ground-level concentration occurring 20% nearer to the
source.
231
-------
cuu
600
500
100 •
2:
rvT
30O •
200 •
100 -
0 -
i i i i i i i 1 i i i 1 i i i 1 i i i 1 i py. '
A Approach Flow, Wind Tunnel
A. Tower A, Field m
U Tower B, Wind Tunnel
• Tower B, Field
B
; ^ ;
• AD
A n
A n
A n
A D
A n
A n
A^ n
A n
A n
A n
A 4 n
4*^* M
A A i — i n c~!* — ' BI
— i 1_^_: 1 1 1 1 1 m *=$ 1 — S 1 1 Ll-!d — uJL : 1 1 1
10
12
U. M/S
Figure A-6. Mean velocity measurements.
232
-------
700
I i i i i 1
600
A Approach Flow, Wind Tunnel
'4 Tower As Field
D Tower B, Wind Tunnel
a Tower B, Field
500
400
DA
D A
D A
D
n
30O
200 -
100 -
n A
n A
n A
n A
n A
n A
n A
n A
• O A
n A
AD A
-AA
10 1^ .15' 17.5
LONG TURB INT
20
25 27.5
Figure A-7. Longitudinal turbulence intensity.
233
-------
700
^gjl
soo -•
A Approach Flow, Wind Tunnel
A Tower A, Field
d Tower B, Wind Tunnel
m Tower B, Field
500 ••
n A
n A
n A
tiOO --
300 • •
200 - -
100 ••
n A
n A
D A
n
n
n A
n A
n A
n A
* n
D
n
A
n A
n A
06 08 iff 12 1U 16 18
LflTERRL TURB INT (ft)
20
Figure A-8. Lateral turbulence intensity.
234
-------
Ground-level Concentration Measurements ...,., . „ ,
Data were collected in the wind tunnel for three different wind
direction and stack height combinations. The first case was for an
ambient wind direction of 312° and a stack height of 50 m (full'-sc'a'lev
equivalent) above the wind-tunnel floor. This wind direction was selected
as an average of the directions reported for Tower A before the >
corrections for alignment of the anemometers was later reported." The
actual terrain near CCB has a gradual slope away from the butte and some
undulations. For the second run, the wind direction was changed to 316°
to correspond more nearly to the direction measured at the source height.
Then, the difference in surface elevation at the source location and the
field reference (8 m full scale) was accounted for by lowering the stack
to match the height above the field reference. These situations, referre'd
to as Runs 1 through 3, are summarized in Table A-5.
For each of the three runs, groutid-level concentrations were measured
at each of the 100 ports corresponding to field sampler.locations. These
data are presented as concentration maps in Figures A-9 through A-ll. The
available field data are shown for comparison in Figure A-12.
Concentration isopleths were hand drawn on these maps to facilitate
comparison. Results for each case can be compared with the field data to
evaluate its success in reproducing.the field results. In addition, the
results from the laboratory runs can be compared with one another to
determine the effects of changing wind direction and"stack height.
TABLE A-5. WIND-TUNNEL EXPERIMENT SUMMARY
Stack Height
Case
Wind
Direction
312°
316'
Above Wind-
Tunnel Floor
7.2 cm
7.2 cm
Above Field
Reference
50 m
50 m
316'
6.05 cm
42 m
Data Collected
100 surface port
concentrations
100 surface port
concentrations,
1 elevated lateral
profile, and 1 elevated
vertical profile
100 surface port
concentrations
The concentrations for all three runs were, in general, a factor of
two lower than the field values. Runs 1 and 2, with the higher source,
exhibit bands of nearly uniform, maximum concentration that extend from a
point on the upwind surface of the hill to the lee side of the hill. The
wind direction for Run 1 differs from that of Run 2 by only 4°; however,
the band of maximum concentration for Run 1 wraps around the side of the
235
-------
Figure A-9. Concentration measurements for Run 1.
236
-------
Figure A-10. Concentration measurements for Run 2.
237
-------
source
MflST R
82.
600 H
x
SCRLE 1:8000
200H
Figure A-ll. Concentration measurements for Run 3.
238
-------
source
SCflLE liSOOQ
200M
Figure A-12. Concentration measurements from field experiment.
239
-------
butte and for Run 2 lays directly across the crest. Run 3, with a lower
release height, shows a more localized region of maximum concentration
occurring on the upwind face of the butte. This is in better agreement with
the field data pattern.
Scatter diagrams of the concentrations for each run compared to the
field values (Figures A-13 through A-15) are another means of evaluating the
agreement of the laboratory and field results. Figure A-13 shows the
wind-tunnel concentrations for Run 1 to be much lower, in general, than the
field measurements. In this type of comparison, Runs 2 and 3 are nearly
indistinguishable (Figures A-14 and A-15). However, there is much better
agreement than for Run 1 in that most data points fall within an error band
of a factor of 2. The difference between Run 2 and Run 3 concentrations can
be seen in a scatter diagram comparing those two runs (Figure A-16).
Concentrations, at the fixed receptors, from a stack 50 m in height are
about 85% of those from a stack 42 m high.
Turner (1970) presented a solution to the diffusion equation known as
the Gaussian plume formula. The plume is assumed to have Gaussian
distributions in the vertical and lateral directions with standard
deviations of az and ay, respectively. The standard deviations are
functions of downwind distance and are obtained, according to stability
classification, from empirical relationships known as the Pasquill-Gifford
curves.
According to Gaussian plume theory, with complete reflection at the
ground, ground-level concentration is proportional to
(l/a2ory)exp(-0.5(H/az)2). If we assume that the plume
geometry does not change (i.e., ay and oz remain the same) and
change only the stack height from H^ to H2, the concentration at the
fixed point should change according to C2/C^ = exp(-0.5(H2/az)2)/
exp(-0.5(H^/az)2). Assuming a value of az of 48 m (that measured in
the wind tunnel over CCB), a change in the stack height from 42 m to 50 m
can be shown to result in concentration decreases by 15%, in general.
This is in good agreement with the comparison in Figure A-16, despite the
presence of the hill.
Another evaluation of the applicability of Gaussian plume theory to
the prediction of surface concentrations is shown in Figure A-17.
Ground-level concentrations along the plume centerline are calculated for
the field source height of 42 m for downwind distances of 200 to
2000 meters. Concentrations from CCB wind-tunnel Run 3 are plotted for
comparison. The concentrations on the upwind hill face (x < 1000 m)
follow the Gaussian plume pattern for D stability. On the lee side of the
hill, the measured concentrations decrease more rapidly with downwind
distance than the Gaussian plume predictions and the concentrations
approach values predicted for C stability.
240
-------
1000
100 -
o.
Q.
e
o
to
O)
o
c
o
u
O)
•- 10 -
c
o:
100
1000
Field concentration, ppt
Figure A-13.
Comparison of concentrations from Run 1 with field data..
In addition, of the fifteen field sampler locations for
which zero concentration was recorded, fourteen were
below 4 ppt in the wind tunnel.
241
-------
1000
10 100
Field concentration, ppt
1000
Figure A-14.
Comparison of concentrations from Run 2 with field data.
In addition, of the fifteen field sampler locations for
which zero concentration was recorded, thirteen were
below 7 ppt in the wind tunnel.
242
-------
1000
Q.
a.
•£ 100 -
£
-------
1000
o.
o.
c
o
•r-
4->
CO
-------
Q.
O.
c
o
s-
c
u
o
»—
(U
o
3
o:
200 500 1000
Distance from source, m
2000
Figure A-17.
Ground-level concentrations along plume centerline for
Run 3 compared with Gaussian plume predictions for flat
terrain. Source height is 42 m; curves are for stability
categories C and D.
245'
-------
Elevated Plume Measurements
The field experiments were limited to direct measurement of
concentration at ground-level only. A lidar system was used, however, to
make oil-smoke plume density measurements from which the vertical plume
spread, oz, can be derived. For this first hour of Case 202, 0Z,
computed as an "hourly value," was found to be 8.2 m at a distance of
200 m from the source and 15.1 m at 520 m from the source. These
positions are upstream of the hill where the plume is not likely to be
strongly influenced by the presence of the hill. These data were not
available at the time of the wind-tunnel experiments and corresponding
laboratory measurements were not made. However, a vertical profile and
elevated lateral profile of concentration in the wind-tunnel plume were
obtained (Figures A-18 and A-19) above the butte during Run 2. The
vertical profile was based at FMF port number 9 (see Table A-3 for
location). The lateral profile was obtained at this downwind distance, at
a height of 46 m above the local surface.
Best-fit Gaussian plume profiles were found for the wind-tunnel data
and are shown in Figures A-18 and A-19. Complete reflection was assumed
at the surface in calculating the best fit to the vertical profile data.
The dispersion coefficients derived from these best-fits are
o = 67.3 m and az = 46.6 m. This ay corresponds to a
Pasquill-Gifford stability category D. The az is shown with the
available lidar measurements and compared with Pasquill-Gifford values in
Figure A-20. The wind-tunnel value falls between Categories C and D,
while the lidar values are between D and E.
246
-------
mo
120 -
100 -
0 10 20 30 liO 50 60 70 .30 90 100
Run 2 Model concentration, ppt
Figure A-18. Vertical profile of concentration.above, -port number.9 of
Cinder Cone Butte model.in Meteorological-Wind Tunnel.
24.7
-------
80
03
OJ
u
o
o
O)
•O
O
-200 -150 -100 -50 0 50 100 150 200
Lateral distance from sampler 9, m
Figure A-19.
Lateral profile of concentration above Cinder Cone Butte
model in Meteorological Wind Tunnel at a height of 46 m
above the local surface at port number 9.
248
-------
© wind tunnel
X lidar
Figure A-20. Vertical plume spread coefficients determined from lidar
measurements in the field and concentration profile in the
wind tunnel compared with Pasquill-Gifford values.
249
-------
SUMMARY AND CONCLUSIONS
A one-hour period of the CCB field study was modeled in the EPA
Meteorological Wind Tunnel. Analysis of the available field data
indicated that the atmosphere was neutral for the period. Three
combinations of wind direction and source height were used in the study.
Concentrations at 100 ground-level sampling positions were obtained for
comparison with available field values. Also, elevated profiles of
concentration were obtained through the plume over the crest of the hill.
The conclusions of this study are:
1)
2)
3)
4)
Modeling neutral flow over this three-dimensional hill in the
wind tunnel. Run 3, reproduced the measured, non-zero, field
concentrations to within a factor of 2 for 43% of the sample
locations; 71% were within a factor of 3.
Changing the approach flow wind direction only 4° resulted in a
significant shift in the concentration pattern on the hill
surface.
Changing only the stack height from 42 m to 50 m resulted in a
15% reduction of ground-level concentrations, in general. This
is in agreement with the Gaussian plume theory relationship for
the dependence of ground-level concentrations on stack height.
Plume spread as determined by computing ay and az from
lateral and vertical tracer concentrations over the hill
top
compared well with Pasquill-Gifford values for flat terrain.
5) Concentrations measured on the upwind face of the hill compared
closely with Gaussian plume theory predictions. However,
concentrations measured on the lee surface of the hill decreased
with distance at a faster rate than Gaussian plume theory
predicts.
The absolute effects of the presence of the hill on the plume and
ground-level concentrations cannot be resolved by this data base.
Additional measurements with the hill and reference measurements in the
absence of the hill must be made to evaluate the effects of the hill. The
wind tunnel data presented here compares reasonably well with the field
data and offers some additional ground-level concentration measurements as
well as measurements within the elevated plume which can be used to extend
the field study data base.
250
-------
REFERENCES
Lavery, T.F., Bass, A, Strimaitis, D.G., Venkatram, A., Green, B.R.,
Drivas, P.J. and Egan, B.A., 1982: EPA Complex Terrain Model
Development Program; First Milestone Report - 1981, EPA-600/3-82-036,
U.S. EPA, Research Triangle Park, NC, 304p.
Pasquill, F., 1962: Atmospheric Diffusion. D. Van Nostrand Company Ltd.,
London, England, 297p.
Turner, D.B., 1970: Workbook of Atmospheric Dispersion Estimates, Office
of Air Programs, Pub. No. AP-26, U.S. EPA, Research Triangle Park, NC,
251
-------
APPENDIX B
SUPPORTING METEOROLOGICAL MEASUREMENTS -
ACOUSTIC SOUNDER AND SONIC ANEMOMETERS
252
-------
SUPPORTING METEOROLOGICAL MEASUREMENTS -
ACOUSTIC SOUNDERS AND SONIC ANEMOMETERS
EPA SHIS NO. 2
W.D. Neff
Wave Propagation Laboratory
Environmental Research Laboratories
National Oceanic and Atmospheric Administration
253
-------
1. INTRODUCTION
In addition to the WPL Lidar, the Atmospheric Studies Program Area of
the WPL provided a number of supporting boundary layer measurements for
SHIS #2. These included two monostatic acoustic sounders, one located just
upwind of the meteorological tower and one adjacent to the Hogback as shown
in Figure B-l. A monostatic/ bistatic-Doppler sounder provided wind
profiles and monostatic facsimile records 1.4 km upwind of the tower. A
portable tethersonde system also provided wind and temperature profiles
first near the Hogback, and during a portion of the experiment, near the
Doppler site.
In addition, just prior to the SHIS #2 a preliminary version of a
portable sonic anemometer system was assembled to provide supporting
turbulence measurements. Two three-axis sonic anemometers were mounted on
Tower A, one at 40 m, a second at 5 m. Electronics for two fast-response
temperature sensors were also in place.
In the sections that follow we will describe these measurement systems,
their operation during SHIS #2, current state of data reduction and some
preliminary analysis and interpretation. In a closing section, we will
outline some proposed improvements in these systems and their deployment in
future experiments.
Sonic Data
SHIS #2 provided an opportunity to test a portable sonic anemometer
data system. At the request of ERT, development of this system began in
early September, 1982. Software was developed on an LSI-11, floppy-disc
based microprocessor system. Raw data were recorded on a 9-track magnetic
tape. Software processing in the field provided 20-min averaged data.
While not fully completed and checked prior to SHIS #2, this system did
provide mean winds, variances, and covariances in a coordinate system
oriented with the sonic device. Post processing of the raw data tapes is
being accomplished by the staff of the Boulder Atmospheric Observatory
(BAO). By adapting this data for input to standard BAD software, a full
range of analysis products can be made available including turbulence
quantities in standard meteorological coordinates, in coordinates rotated
along the mean wind, as well as spectra and cospectra. Addition of a
hard-disc based, multi-tasking operating system to the field version of the
LSI-11 system will eventually allow such products to be available in field
settings.
As operated in this experiment, the sonic anemometer interface
electronics provided both analog and digital outputs. The fast-response
temperature sensor provides only analog data. Two versions of the data
254
-------
u
i
a. ^n
(D
CD
•H
(/I Oi
H
CD ^
•Xl O
S Oj
g-a
w o
O 0
•H
PH
255
-------
acquisition program were developed to allow processing of tnese various
data. The first version provided acquisition of 16-bit digital data
directly from the sonic interface. The second provided 12-bit conversion of
the analog outputs from the sonic interface. The limitations of a 12-bit
A/D led to less resolution of both temperature and analog sonic data.
Evaluation of spectra of temperature will provide some indication of the
usefulness of these data. One may expect, however, that the data resolution
should be better than that from conventional instrumentation such as
propeller anemometers.
In actual operations, the sonic system performed reliably with two
exceptions. Following the power outage of 8 October 1982, a number of
computer system components were destroyed in addition to the ditigal sonic
interface bus ICs. With spare cards from a backup computer system, the
sonic system was operational again by 10 October. Because the replacement
ICs for the sonic digital interface were not available until 15 October, the
program was put into a purely analog mode, bypassing the digital interface.
This provided reliable sonic-derived turbulence data for experiments 4
through 8. Inspection of w-data in the field and in the post-processing
phase show that it is free of spurious noise except for glitches introduced
by the radio transmitter located in the equipment trailer as shown in
Figure B-2. In general, these radio-transmission glitches appear only in
the bottom level of sonic data. Preliminary inspection indicates that these
do not contribute significantly to sigraa-w.
The installation of fast response temperature sensors proved more
difficult. Early in the experiment a platinum wire temperature sensor was
installed with the sonic on the lower boom. This sensor survived the entire
experiment. However, its absolute calibration is off by a factor of two
because of the mismatch between the differential output of the sensor and
the single-sided input to the A/D. The upper sensor presented additional
difficulties. Because the boom was not retractable, installation was
accomplished by swinging from a rope supported at the top of the tower.
With the fragility of the platinum wire sensors, repeated breakage occurred
in this process. The decision was then made to record data from the ERT
thermistor at 40 m. A scaling factor of 8°C/volt combined with the limited
resolution of the 12-bit A/D provided a resolution of 0.04°C, a factor of 8
less than that for the platinum wire sensor.
As of 15 May 1983, preliminary hourly averaged values of wind and
vertical velocity variance had been provided to ERT. In addition, one
reformatted, scaled data tape had been submitted. Complete preprocessing is
expected by the end of May. A number of BAD analysis routines have also
been applied to the SHIS #2 sonic data. Some results from this preliminary
analysis are described in Section 2 of this appendix.
Monostatic Data
Two monostatic sounders with high-resolution facsimile machines
(0.30 m/hr; 150 ra to 340 m vertical range displayed over 0.22 m) were
operated throughout the experiment. One close to Tower A provided a
characterization of the boundary layer near the tracer/smoke release point.
The second, at the base of the Hogback, provided a characterization of the
256
-------
bJ
5>-
UI
X
a.
a:
c£
cs
cc
H-
cr
e
©
Vl
05
v>
{=>
o
t-H
a:
u
e
IS
v-l
\
ui
H:
\f>
Gl
w
I
bO
fi
•H
o
aJ
•P
nj
•ri
(D
•H
I
I
•H
0
TS -P
S -H
aj
o
(D
>
o
•H
in
1/1
• H
a
2 o
en
0) •
fn
257
-------
boundary layer flow as it approached the ridge. Prior to 19 October, the
ridge-sounder was located directly opposite the Tower A access road. It was
then moved 180 m south to avoid acoustic signal contamination from the noise
of the oil-fog generator.
Facsimile recordings from monostatic acoustic sounders show time-height
cross-sections of small-scale (0.20 m) temperature fluctuations as they move
through an antenna beam of some 4° to 8° beamwidth. Such small-scale
thermal structure is formed when turbulent motions mix parcels of different
potential temperature. Examples include regions within temperature inver-
sions where the vector wind shear is increased sufficiently to reduce the
gradient Richardson number below a critical value, leading to the production
of turbulence. More commonly, the shear layer induced by surface friction
combined with radiative cooling of the surface usually produces a well-
defined echo layer on acoustic sounder records. An example, showing both
types of echo layers, appears in Figure B-3. In this case, a ground-based
turbulent layer extends into the overlying inversion to a depth of about
40 m while elevated layers occur at heights extending up to 175 m. In
general, the inversion is much more variable than shown in this example.
Often, the height of the turbulent layers and the character of the turbu-
lence change considerably over the course of an hour. Further discussion
and examples of the interpretation of acoustic sounder data will be found in
Section 2.
Doppler Data
The WPL bistatic Doppler sounder was located 1.4 km upwind of the
experiment site near a power substation. This system, providing vertical
profiles of the wind from 50 m to 250 m, was intended to characterize the
approach flow and provide a general meteorology of the area. It operated in
conjunction with a standard, medium resolution facsimile recorder
(0.123 m/hr; 820 m maximum range). Figure B-4 shows a block diagram for
this system including the layout of the bistatic transmitters and vertical
receiver. Conventional wisdom has always suggested a bistatic arrangement
to be preferable because of the increased forward scatter from velocity
fluctuations that adds to the received signal strength and because the two
components of the wind can be computed in a common volume. The commonly-
noted drawbacks of such a system include the requirement for more open space
to locate the antenna array and increased emission of noise into the
environment. While such systems have increased signal-to-noise ratios,
particularly under adiabatic conditions, the minimum range is somewhat
greater than for monostatic systems. Also as we shall show below, with
strong refraction (due to large wind shear or temperature inversions near
the surface), displacement of the scattering volume and/or increased ground
clutter can bias the Doppler shift from which the winds are computed.
From the results of SHIS #2, it would'appear advisable to use a
monostatic Doppler system for pure temperature inversion flow situations and
where the scale of the flow is greater than the separation of the scattering
volumes (that results from tilting two monostatic sounders in orthogonal
directions).
258
-------
Figure B-3. Monostatic acoustic sounder facsimile record displaying a
ground-based echo layer as well as elevated echo layers
(10/29/83, 0200-0300 MDT).
259
-------
Echosounder
and
Doppler Board
Figure B-4. Block diagram of the WPL bistatic Doppler sounder system.
260
-------
Calibration and Data Quality Checks
Doppler sounders are best checked against fixed tower systems. In^the
field, basic program operation and parameter settings can be checked using a
known frequency input. Calibration against a tethersonde is more diffi-
cult. Holding the balloon at a fixed level causes, .sidelobe reflections
which introduce a zero Doppler shift, usually at the same height as the
comparison. Using a single ascent (descent) of the balloon introduces the
complications of comparing a 20-min time average with a snapshot wind
profile. In addition, the motion of the balloon downwind or upwind can
introduce a bias of as much as 1 m/s under stronger wind conditions.
To eliminate some of these problems, we sought comparison cases where
the balloon ascended and descended in reasonable .proximity to a 20-min
Doppler averaging period. An example of such a comparison is shown in
Figures B-5a and B-5b. In Figure B-5b, the two tethersonde profiles show a
wind speed difference of about 1.5 m/s above 50 m. The reason for this is
that the tetherline has a speed of about 0.5 to 1.0 m/s. At wind speeds of
more than a few m/s the balloon starts moving downwind. As the angle to the
surface decreases, the line speed subtracts from the wind speed. As the
tethersonde descends, the effect is reversed. The net difference then
between ascent and descent ranges from 1 to 2 m/s.
During several cases examined, the presence of a strong low-level jet
resulted in anomalous Doppler shifts. Two possible causes have been
identified. Both are associated with refraction of the bistatic acoustic
beam by the wind shear and/or temperature gradient. This can, for example,
result in a displacement of the scattering volume off the vertical axis of
the receiving antenna making the Doppler shift reflect the vertical
component of the wind more than the horizontal. The second possibility is
that increased refraction near the ground, results in increased ground
clutter and hence a bias towards zero Doppler shift. These cases were
evident only with a wind maximum below 40 m and appeared only to affect the
range gates below 60 m. Following final quality assurance on these data,
they will be submitted to the CTMD data base. These results do suggest the
advisability of using monostatic Doppler geometries for measuring winds near
the surface under shallow drainage conditions.
261
-------
ACOUSTIC DOPPLER DATA
STARTING TIME
10/20 6 0 MDT
MIND
INDEX HEIGHT SPEED DIR
22. . .
21. . .
20. . .
19. . .
IB. . .
17. . .
16. . .
15. . .
14. . .
13. . .
12. . .
11. . .
1O. . .
9. . .
8. . .
7. . .
6. . .
5. . .
4. . .
3. . .
2. . .
1. . .
. . 46O
. . 44O
. . 42O
. . 4OO
. . 380
. . 360
. . 340
. . 32O
. . 30O
. . 28O
. . 26O
. . 24O
. . 220
. . 2OO
. . 180
. . 16O
. . 140
. . 12Q
. . 1OO
. . 80
. . 60
. . 40
0. 0
O. 0
0. 0
O. O
O. O
O. O
0. 0
O. O
O. O
2. 9
0. 0
2. 4
3. O
3. 6
4. 6
5. 7
4. 8
5. 8
6. 0
4. 7
2. 3
2. 8
51
32
41
62
334
51
83
49
87
87
88
86
88
95
93
91
88
81
79
1O2
1O3
9O
ID1
•*••
D
D
D
D
D
D
D
D
D
*
D
*
*
*
*
*
*
*
*
#
*
*
DIRECTION (D)
27O O 090 180
+#*#*****+#*####*#+*#***#**+**#*****•*•
*
*
*
*
*
*
*
S
S
D
D
D
D
D
D
D
D
D
D
D
*
*
*
*
*
*
*
#
*
*
*
*
*
10/20
6: O
+*##****#*+*#*****#*+#**#*#**#+*#***+
O 5. O 1O. O 15. O 18. O
WINDSPEED (D)
Figure B-5a.
Wind speed and direction profile measured by the bistatic
Doppler sounder system (10/20/82, 0600-0620 MDT).
262
-------
| L-
o in
o w
-a \f)
c
a>
M
^
Q
2
in
in
in
O
1 4-
0 If!
O «
•a in
-v- o
1 sji
"*'-1^1>.*-)*j+-
*"l*4-..»-.
* **
_
r-T ,(-„-„., , 1 , , 1 1 , . 1 , „,„ 1 , , . . I , . t
1 1 hi 1 1 r
Kl
.
. o —
f-v O
OJ UI
£4
z
o
_ O 1-
03 CJ
i-l UI
^H
a
a
_ O M
O- 3
. o
o in o sn o in o
in IS C> C-i in fN
>*• ro PS r-j i-i
. ( w ) .1. H a i :;;i H
••
"•
.
"*
/\
>J S.
/ \
-^ \ :
-- -i — i i — i - - i- - i ;
1 i i i i \ \
— in
•r-t
•~ "4
O
lu
s.
- o-5
Ui
Ui
0.
- o en
C3
^-t
3
- M
O
o in o in o in o
in r-. o r-4 in rx
•T* tw*
"
*•
| (. 1 j 1_ ( ^ 1 .) ,
HJ
. o <-
r-j w
•z.
c
^-i
- O i-
CD tJ
^i Ui
u:
e
C;
* 0 w
l> 3
O
oinoinoinoino
o N in N o w ii"2 r-
•a iiT <• ro M CM »-»
(• w ) .1. |»j .9 ;i; ;«| |-|
-
-
1
O y*.
s / \
^ / M -
i / \
5 / \
0 / \ ..
5 'i
o \
1 ' ij. t ) ' i i i
! , ]-j— -f — j 1— 1 (• \ 1
- ll"3
'""'
- PJ
0
U!
V
-0.5
iZi
UI
Ui
ft.
- «o to
w
z
^
- M
O
o in o u~ o in o 'in o
o r-r b~ ix o CM w is
T3 O
pj to
cti cd
CD CM
— '
*
LO
1
CQ
CD
•H
< W ) .1. H £) I -il H
( w
263
-------
2. PRELIMINARY ANALYSIS
SHIS #2 provided a unique set of data from which to analyze winds and
turbulence behavior in nocturnal, complex terrain flows. In particular,
sonic anemometer data provide an accurate measure of turbulence down to very
small scales. Monostatic sounders provide detailed flow visualizations that
in general belie any expectation of a steady, simply-structured flow. The
Doppler sounder together with supporting tower instrumentation provide an
appraisal of the variability in the flow above the boundary layer.
Interpretation of Monostatic Data
1. Increasing shear above the surface-based inversion leads to
dynamical instability. Such instabilities are easily identifiable from
monostatic records and provide a measure of the vertical extent of
large-eddy structure in the inversion. The growth of such instability
regions seemed to be independent of the proximity to the Hogback.. In fact,
similar patterns were observed 1.4 km upwind at the Doppler site. From the
Doppler wind data, increases in wind speed above the inversion of from 1 to
2 m/s were observed during these events.
As deduced from second-order closure modeling, turbulence depends
inversely on the Richardson number. Since Ri in turn depends on the inverse
square of the vector shear, small changes in the shear can translate into
large increases in turbulence. From these data, it appears that mesoscale
processes above the inversion can strongly affect turbulence within the
inversion and boundary layer and hence the dispersion of the plume prior to
impact on the Hogback.
2. The flow approaching the Hogback, like most nocturnal inversion
flows, is characterized by intermittent turbulent layers and patches. For a
given turbulence measurement, say using the 40-m sonic anemometer, the
sounder record can provide a good indication of the vertical extent over
which the measurement is valid and the variability of such regions in time.
The data suggest that linear interpolation schemes are not in general
reliable (again, the sounder record can indicate when such schemes might be
valid). In addition, the sounder records show considerable variability on
time scales that are a significant fraction of a typical averaging period of
one hour. In such cases, the sounder record may suggest a more appropriate
averaging period, not necessarily beginning on an even hour.
3. The sounder records, while showing many repetitive features in the
approach flow, also revealed many short-term phenomena such as inversion
height changes due to an apparent local horizontal convergence or divergence
of the flow. In other circumstances such patterns have been associated with
10- to 20-rain period internal waves.
264
-------
Examples of Data from the Morning of 15 October
Figures B-6a and B-6b show two sequential half-hour segments of
acoustic sounder records between 0500 and 0600 MDT on 15 October. Prior to
0500 MDT, the boundary layer, identified as the lowest echo layer on the
facsimile recording, stayed below 50 m. It would oscillate above this level
and would at times be perturbed by motions from above. During the early
portion of the record shown in Figure B-6a, the flow appeared almost laminar
at 40 m. (The absence of an echo on the sounder indicates either a laminar
flow or one in which the potential temperature gradient is adiabatic.
Adiabatic regions bounded by temperature inversions are not usually clear of
echo: in such cases the decay time for the temperature spectrum is greater
than the time required to transport some temperature variance (w02) from
the boundary). Figure B-7 shows the corresponding time series from the
sonic anemometer at 40 m obtained over a 1-min section of this record. In
this case, superimposed on a slow background oscillation are fluctuations on
the order of 0.01 m/s. During this one hour period, the flow aloft
increased in speed. This resulted in a thickening shear layer. Starting at
a height of 70 m it rapidly expanded downward to meet the boundary layer
below. At the same time, an increase in shear farther aloft expanded to
fill a region up to 125 m. In this case Figure B-8 shows the effect on w
(and therefore on sigraa-w).
Evident in the sounder record are descending echo bands that are most
noticeable between the surface and 80 m. These have, in previous research
been identified with Kelvin-Helmholz instability. The question then arises
of the contributions of this large-eddy structure to sigma-w. This can be
estimated from the log-log plot of frequency versus the product of frequency
and power spectrum shown in Figure B-9. In this case, the product is almost
an order of magnitude above the background spectrum, suggesting a
significant contribution, but not an exclusive one. (A number of such cases
will be analyzed with separate integrations over the wave and turbulence
portion of the spectra). Almost evident in these records is the presence of
an inertial subrange. Within the surface layer this tends to occur at
frequencies greater than 1 Hz. In most cases analyzed thus far, the
spectrum at periods greater than 10 s appear pink (freq. times the spectrum
is flat).
A feature noted during the field exercise was the occurrence of
positive heat fluxes in the 20-min averaged covariances. The cospectral
analysis shown in Figure B-10 shows that these apparent positive fluxes
result from low frequency motions. Generally, we have found that the flux
is negative for frequencies above 0.01 Hz. In the particular case shown the
flux for frequencies less than 0.01 Hz is positive. In part this may be due
to a shallow, fluctuating downslope flow of cold air in the lowest 5 m.
With a local terrain slope of 0.03 (based on the slope along the road),' a
1 m/s drainage flow would produce a 0.03 m/s w. This negative w would then
produce a positive correlation with the decrease in temperature.
Continuing Analysis
With acoustic sounder data (monostatic and Doppler), sonic anemometer
turbulence information, and tethersonde data, analysis such as that
described above will continue. It is expected that the analysis of the
primary modeling cases will occur first followed by other cases of interest.
265
-------
50
0530 MDT
Time
0500 MDT
Figure B-6a. Acoustic sounder record (10/15/82, 0500-0530 MDT)
266
-------
0600 MDT
Time
0530 MDT
Figure B-6b. Acoustic sounder record f(10/15/82,•0530-0600 MDT)
267 -
-------
CM
_J
LU
LU
_I
X
a.
or
CD
cc
a
o
LU
CD
1-4
\
CD
U)
UJ
=p
C«J
G>
LO
•
CD
CO
0)
e ^-x
I H
O Q
rt- S
CD LO
rC O
+J LO
O
O •>
H CM
<4-l 00
rt LO
U) > — '
o
•H T3
f-l O
CD .H
U5 f-l
1 CD
i ^
H CD
4-> -P
3
CD -H
CD S
o I i
CD
f-l
268
-------
u
•X
ct:
cc
I—
t-
00
I
CO
bo
•H
2.69
-------
821615. 5.48 LEVEL 2
28 MIN
5y n~ (inertial Subrange)'
-4.ee
-4
00
-3.00
-2.00 -1.00
LOG N
0 .00
1 00
Figure B-9. Log-log plot of frequency versus the product of frequency
and power spectrum (10/15/82, 0540-0600 MDT). \
270
-------
821015. 5.48
LEVEL 1
120 MIN
-7.15
-4.
1 88
Figure B-10. Cospectral analysis of temperature and vertical wind speed
fluctuations measured at 5 m showing that positive heat fluxes
result from low frequency motions (10/15/82, 0540-0740 MDT).
271
*USGPO: 1983 — 759-102/0789
-------
-------