ML20141D327
| ML20141D327 | |
| Person / Time | |
|---|---|
| Site: | Point Beach  |
| Issue date: | 08/31/1994 |
| From: | Ramsdell J Battelle Memorial Institute, PACIFIC NORTHWEST NATION |
| To: | NRC (Affiliation Not Assigned) |
| Shared Package | |
| ML20141D306 | List: |
| References | |
| CON-FIN-P-2028 NUDOCS 9705200068 | |
| Download: ML20141D327 (28) | |
Text
7 q r , ,
' ~ ~1 1.~ .-
1 L .
F/24/iv l
1 DISPERSION ESTIMATES IN THE VICINITY OF BUILDINGS 1
I .. ,i 1
l l
l J. V. Ramsdell, Jr C. J. Fosmire August 1994 l
Prepared for the l Division of Radiation Safety and Safeguards Office of Nuclear Reactor Regulation U.S. Nuclear Regulatory Commission Washington, DC 20555
- I NRC JCN P2028 Pacific Northwest Laboratorf Richland, Washington 99352 I .
{
i .
, y
- k. .
CONTERTS INTR OD U CTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . . ,
1990 Model . . . . . . . . . . . . . .
Pe e r Revie w . . . . . . . . . . . . . ..........
2 Experimental Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
................'.......... 3 REVIS ED M ODEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . .
General Form for DNfusion increments' . . . . . . . . . . . . . . . . . .5 . . . . . . .l Low Wind Speed inorement . . . . . . . . . . . . . . . . . . . . . . . . .
6 :
High Wind Spee d increment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Complete Model . . . . . . . . . . . . . . . . . . ............... 7 10 Model Evaluation . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
............................. 11 i
ALTER NATIVE M ODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13. . . . . . . .
Wilson.Chui Model . . . . . . . . . . . . . . . . . . . . 14 Wils on Lamb M odel . . . . . . . . . . . . . . . . . . ..........................
15 Comparison of the Revised Model with the Ahematives . . . . . . . . . . . . . . . . . . 17 .
NEAR FlELD CONCENTRATION ESTIMATES . . . . . . . . . . . . .19. . . . . . . .
t CON C LU S I O N S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21. . . . . . . .
R EFER E N C E S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 . . . . . . . . .l 1
1 j
9 4
s il
W) vu i ,
L . \
. 4 FIGURES Figure 1.
Increase in lateral turbulence as function of wind spee'd . . . . . . .8. . . . . . . ) 1 Figure 2.
Increase in vertical turbulence as function of wind speed . . . . . .9. . . . . . . .
Figure 3. Comparison of revised value . . . . . . . . . . . . .model concentration predictions with observed Figure 4.
............................. ....... . 11 i
Bias in Murphy Cam wind speed . . . . . . pe model concentration predictions as a function of Rgure 5.
........................................ 12
- Bias in Regulatory Guide 1.145 model concentr function of wind speed . . . . . . . . . . . . . . . . . ation predictions as a
..................... 12 Figure 6.
Bias in revised modelconcentr speed . . . . . . . . . . . . . . . . . .ation prediction as a function of wind
................................ 13 Rgure7.
Comparison of cumulative frequency distributions of predicted to observed concentration ratios for the Murphy Campe, Regulatory Guide !
1.145 and revised models . . . . . . . . . . . . . . . . . . . . . 14 ........
Rgure 8. Comparison of Wilson Chui model l observed values . . . . . . . . . . . . . concentration predictions with !
............................. 16 Figure 9. Comparison of Wilson Lamb model observed values . . . . . . . . . . . . . . concentration predictions with
............................ 17 !
Figure 10.
Comparison of revised model concentration estimates with observ values in the building surface data set . . . . . . . . . . . . . . . . . . . . ed
....... 18 Figure 11. )
Comparison of cumulative frequency distributions for the ratios of -
predicted to observed concentrations for the Wilson Chul, Wilson-Lamb, and revised modeis based for the bu,lding surface data set . . . . . .18. . . . .
Figure 12.
Ratios of predicted to observed concentrations for the Wilson-model as a function of normalized distance . . . . . . 20
......... . . . . . . .Ch Rgure 13.
Ratios of predicted to observed concentrations for the Wilson-model as a function of normalized distance . . . . . . . .20. . . . . Lam Rgure 14.
Ratios of predicted to observed concentrations for th a function of normalized distance . . . ................ . . . . . . . . . . . e revised 21 mod Rgure 15.
Cumulative frequency distributions of the ratios between predicted and observed concentrations for the Wilson-Chul, Wilson-Lamb an model for all data near the release point . . . . . . . . . . . . ........
. .,. . d revised 22 l l
El
l
( -
DISPERSION ESTIMATES IN THE VICINITY OF BUILDIN
- INTRODUCTION Control .com habitability assessments and the evaluation of the co design basis accidents involve estimating dispersion of emuents level stacks. Nuclear Regulatory Commission guldance to staff and several acceptable methods for estimating dispersion in building wa Regulatory Guide 1.145, and the Murphy Campe procedure (M referenced in Standard Review Plan 6.4 (NUREG4800). These m the concentration at the conter of the plume downwiruf of the rele based on the straight litie Gaussian model. uations According to the Gaussian model, the concentration at center of a plum 1
x/o -
s e,e,U g) where x is the concentration, O is the release rate, o, and o, are horizo diffusion coefficients respectively, and U is the wind speed. Diffusio the effects of turbulence and are generally estimated on the y and basis of a distance using empirical relationships derived from experimental data diffusion models have the same form but represent the effects e of th diffusion coefficients. These models are typically written as x/O * ~
sZ yE,U (2) wherey I and I, are diffusion coefficients corrected for building wake effects.
l Plume centerline concentrations predicted by the various NRC bui were compared with experimental data under NRC'JCN 82970 ' Atmosphe Control Room Habitab!!!ty Assessments'. The results of this work -
5055 Atmospheric ChYusion for ControlRoom HabitabWty Assess ,
showed that the models did not predict variations in concentrations building area and atmospheric conditions very well it also showed that th significantly overpredict concentrations at low wind speeds. '
I i
I I .
M90 Model <
{ . )
/4 updated building wake model (Ramsdell 1990) provides y new defin I,. The definition ofI,is 2, = (c' + a ch" N) where o, describes the diffusion in the absence of a building wake, and Ap p describes the increased diffusion in butiding wakes. The increased diffusion in wakes is esti!
mated by .
Ao ,' = **^(1 -(1 +.**-)axp(.~E)]
u.' /A /K (4) where k is a constant with a value of about 1; Ae m is the increase in the horizontal component of turbulence caused by the building; A is the cross sectionalarea of building; u. is a turbulence scaling velocity in the atmosphere that is r ,
atmospheric stability, and surface roughness.
The expression in brackets on the right side of (4) controls the expansio l plume as x, the distance from the release point increases. It is equa to 0. For x less than 0.2e/A", the expression increases a, approximately a greater than Sa/A" the expression has reached its maximum value of 1.0.
is a proportionality constant between u. and U and is therefore a func stability and surface roughness. For near neutral atmospheric stability, the about 0.09. Similar expressions, with explicit dependence on atmosph derived for I,.
- Two assumptions of note related to I, and I, were made in develop!
the 1990 model. The first of these was that any of the standard sets of dl j
algorithms could be used to estimate diffusion in the absence of the wake j diffusion). The second, and more critical, assumption was that o,is indep speed. With these assumptions and the parameterization for normal diffusl j
1ound in most NRC computer codes, the 1990 model resufts in better predict .
{
eenterline concentrations in wakes than the models described in curren l Commenting on the 1990 model, Briggs /et al.,(1992), point odt that th i
turbulence associated w!th building wakes should be a function of wind .
i
~
4 2
l > .
. ph.! out ttat the improved predictions of the updated model at low wind speeds a
,(.
i +' 5 5 related to better treatment of meander than of builtjing w.'cas.
j -
. #ser Review . .
l A panel was convened in May 1994 to conduct a peer review of the 1990 w e *Ti.e primary recommendations of the peer review panel were: -
1)
- The turbulence increment generated by buildings should be assumed to p l to the wind speed in accordance wtth accepted theory and physical reason i
- 2) , '
in the model, but the treatment should be separa wakes.
3)
An approach to determining concentrations other than straight line Gaussian
) should be considered when releases are from a building and receptors are on fne building, 1
t
(
i 4)
- ppropriate subsets of the available data should be used to evaluate the mode the suggested changes have been made.
This letter report describes the disposition of the recommendations of the pe pa7el. In addition to this introduction, the report consists of three sections that address the peer review panel recommendations. These sections describe the model rev'slons and compare the revised concentrations predicted by the revised model wit measured concentrations and with concentrations predicted by attomative models. T of these sections discusses revisions to the model to separate the effects o wind speed phenomena on diffusion. It is followed by a section that discusses th attemative models in the for situations in which receptors are on or adjacent to th from which the release occurs. Both sections include discussion of model perfor j
which model predictions are compared with observed data. The last model evaluation section compares concentrations predicted by the revised model and two attemati with measured data that are particularly appropriate for evaluating models used for room habitability assessments.
Ernerimental Data ..
~ The two data sets used in evaluating model performance contain data collected in Sold experiments at seven different reactors. Three of the seven reactors- the Mat Reactor-Engineering Test Reactor (MTR) (Islitzer 1965), the Experimental Breede 3
I ,
, .t .
(EBR 2) (Dickson et al.1969), and the Experimental Organically Cooled R
( (Start et al.,1980) are at the Idaho National Engineering Laboratory. The rem i ,
are the Duane Amold Energy Center (DAEC) in Iowa (Thuillier and Mancuso 1980 1982), Diablo Canyon Power Plant (DC) (Thulliier 1988) and Rancho Seco Nucle Station (RS) (Start et al.,1978) in Califomia, and the Three Mile Island Nucle (TMI)(GPUSC 1972)in Pennsylvania. -
I The erst of these data sets, referred to as the ground level data set, consists o concentration measurements made at regular intervals on sampling arcs from 50 to '
i from ground-level rolesse points. The maximum concentration on each arc w; the best approximation of concentration at the center of the plume as it crossed s!!, there are 379 data points taken during 131 release periods in the ground lev Meteorological conditions during the releases ranged from extremely s class G) to extremely unstable (stability ciass A), and wind speeds ranged from meter per second to greater than 10 meters per second. Of the 379 data point 253 represent measurements made with wind speeds less than 4 m/s,208 data represent measurements made during stable atmospheric conditions, and 138 data represent measurements in low wind speed, stable atmospheric conditions. Concentrat predicted for low wind speed, stable atmospheric conditions are generally used in of consequences of accidental releases in control room habitability assessments site boundary. In some cases the actual release point was offset from the center of th sampling arc. As a resutt, the range of distances of the data points in the ground-leve '
set is 6 to 1200 m.
The ground level data set is described in earlier publications (Ramsdell 1 The data have been used as reported except that stability classes have been mo
- few cases where the stability class determined by the NRC detta T method was in with other reported data, for example wind speed or season and time of day. These modifications typically involved changing extremely unstable or extreme!y stab classes to more nearly neutral stability classes. Neutra! stability was assumed for att experiments in which the wind speed exceeded 6 m/s.
The second data set, referred to as the building surface data set, consists of 265 concentration measurements made at locations on and adjacent to buildings at Rancho Seco, the Duane Amold Energy Center, and the EOCR. Data from both ground lev
(. .
i 4
i
. - - - - . - - _= . _. - . . - . _ . _ - _ - _ _ _ . - - - - - - _ -
l
) .
,s .
elevated rele'ase poirje C.'Mt *od in the data set. Meteorological conditions rep h cover the tull rangrid WI.g:
Esp vat;#inr: speeds from less than 1 m/s to almost 10 m/s.
More than half of the detrpoint.s represent measurements in low wind speed, stable cond$ons. A!! of the measurements were made less than 100 m from the releas The aamplers were not arranged in pattoms that would erhure a reas,on that the concentration at the center of the plume was captured at each distan ,
these data are not appropriate for use in developing a model to predict centerline concentrations. However, the data can be used to evaluate model performance. Plu centerline concentrations predicted by a model abould tend to form an upper boun measured concentrations.
REVISED MODEL The first two recommendations of the peer review group have been ad revision of the 1990 building wake model. The revisions add a diffusion in related to wind speed and modify the existing increment to be more directly rela
, wind speed conditions.
Revision of the model starts by redefinition of diffusion y coefficients, Z now definition ofIy is '
8 Z, = (c/ + A e/ + A og )1r2 (5) where o p represents diffusion from a point source under normal conditions, r Ao g epresents an increment to diffusion associated with low wind speed g phenomena, and an increment to diffusion associated with high wind speeds. A similar To maintain continuity with existing regulatory guidance, the relationships us define the diffusion coefficients in the NRC PAVAN and XOOD applicable for op and o,. The relationships were developed initially by Martin and Tikvart (1968) and Tadmor and Gur (1969) as approximations to the Pasquill Gifford d coefficient curves. The relationships have been extended to include stab the guidance in the February 1983 reissue of Revision 1 of NRC Regulatory Gu . .
Genera 1 Form for Diffusion increments .
Derivation of expressions for Aop ', Ap pcand the corresponding increments to vertical diffusion generally fo!'ows the derivation of the diffuslon increments in the origin 5
i '
3
.1 ? .
..c:h. J'u derivatioa starts by .asuming that some phenomenon or combination of
' $,4aoraana causes an licvesse la turbulence above the tufbulence implicitly assume numal diffusica coefficients. At low wind speeds, meander and possibly uneven building surfaces may be responsible for increased diffusion. At high wind speeds!
mechanical turbulence associated with wakes is responsible. The second assump
'the effect of the turbulence increment on diffusion decreases exponentially as a func tirm relatNe to an appropriate time scale. ~
With these assumptions, a horizontal diffusion coefficient increment is defi '
A9,a,gffg,,s9xp(-1/T)R(t)dtdt (6)
^ where Ao,is the increase in the lateral component of the turbulence, t is the time since release, T is the time scale, and R(r) is the autocorrelation function for the tur is assumed to have a constant value, k, near the release point, the double int the solution ,
Ae,a=2kAa,8Tr{j.(y ),gp(. )) , F)
The first part of the expression on the right side of 7 F),2kAo,8 8, determines the max '
increment to the diffusion coefficient. The second part of the expression, which is in brackets, determines the fraction of the maximum increment that is applied as distance) increases. Note that the terms in brackets in (4) and F) h ,
term is zero at the release point and asymptotically approaches one as the distance increases. Two consequences of this behavior are 1) that the model does not pred instantaneous diffusion at the release point, and 2) that the pI, at approaches large o distances. It is not necessary to apply arbitrary limits to the model to avoid unrealistic asymptotic behavior either near the source or at large distances.
Low Wind Sooed increment Ao,,a The' relationship in F) is general and may be used y to define acr ,8, Ao,,8 p 8,and provided the appropriate turbulence increments and time scales are used in each cas For the low wind speeds, a time scale of 1000 s has been chosen for the horizontal turbulence increment. This time scale,'which is larger than time scales usually assoc
. with mechanical turbulence near the ground, is consistent with time scales associated wit 6
i
.I f the transition from diffusion coefficients proportional to time to diffusio square ruot of time.
Two time scales have been chosen for the vertical turbulence winds. The time scate for stable conditions is 100 s. This time scal the inverse of the Brunt Valsala frequency for the temperature y lap classes D and E as defined in Regulatory Guide 1.23. The Brunt V undefined for neutral and unstable conditions. The time sc
. assumed to be toro. As a consequence, the low wind speed vertical diffu zero for neutral and unstable conditions.
I Hieb Wind somed increment For high wind speed conditions, the time scale for decay of the ho
' increment is i
T.E u, i
where A is the cross sectional area of the structure generating the incr I
', u'is the friction velocity upwind of the structure. The time scale for the d turbulence increment is
$ 4
' T- $
u,(2 +2/L) i (9) !
k where z is the wind speed measurement height and L is the Monin Obuk '
j unstable conditions, the time scales for decay of the horizontal and vertical tu 4 increments are about equal, while in stable conditions the time scale turbulence increment is about a factor of 2.5 less than the time scale for d horizontalincrement. This stability correction is discussed in more d the original model(Ramsdell1990).
Turbulence data published by istitzer (1965), Dickson et a!. (1969), a i Oikawa (1982) have been used to develop parameterizations for turbulen during high wind speeds. Changes in o, and 0, were computed as '
1 i 7 l
f i -
-- =
. t ba = fol-of .
('0) where the u and.d subscripts represent upwind and dowrruind .respectively. In instances, the subtractiori resulted in negative differences. Thes; differences cortsidered to be unreal an were set to zero. The resulting turbu'ence inc against the square of the upwind wind speed in Figures 1 and 2. As s review peneI, these increments are functioris of wind speed. The c betwson the turbulence increments and the square of the upwind wind s O.01 for aa ' .
o g and ao , respectively. Thess ooemeients are dimensional with units of seconds / meter. The correlation cosmcient for the relatioftshipGnwn in Figu for the relationship shown in Figure 2, it is 0.81. ,
2.5 a
2.0 ii -
e 4
8 1 .5 g /
, e
$ /
1.0 .
.h *
- V)
, e c c.5 :
t a - Dickson. et al.1969 8, he a b - Ogawa & Oikaws 1982 c -Ishtrer 1965
' ~
c.o 20.0 40.0 so.o 80.0 100.0 120.0
' Wind Speed Squared (rn/s)^2 Figure 1.
Increase in lateral turbulence as function of wind speed.
f 8
I e =~
. t
} .
1.20 '
~~ -.
1.00 c.so g 3 g ,
b o.60 ,
.Ei ,.
E '
o.40 *-
- I i
s e
- i gyg : a - Delmon, et al.1e69 6 j
- j _
' b-Osews and Okews 1982 m g ,. c-lantzer le65 e i c.co ?a .
i
. 3 i o.0 20.0 40.0 so.o so.o 100.0 120.0 Wind Speed Squared (m/s)^2 Figure 2.
Increase in vertical turbulence as function of wind speed.
With these relationships, and the assumption nthat Aa and 40,3 are equal, the remaining unknowns in the model aregk, Ae . Values for these parameters were determined '
by minimizing the function P = ZWiiog( )g-log ( )J (11) i where (X /0),,,, is the normalized concentration predicted by the model, (X)O). is the normalized concentration measured in the ground level data set for the sams conditions, a W, is a weighting function. The weighting function is equal to 1.0 when the model ovetpredicts the measured concentration, and equal to 2.0 when the model underpredicts Optimum values of k and n aa could not be determined almuttaneously. However, search for an optimum value of Ap g converged rapidly when a value was specified for k.
Optimum values of nao were determined for k equal to 0.25,0.5,0.75, and 1.0. The values of P were not particularly sensitive to the value of k, but they were sensitive to the pro the parameters that appears in the low wind speed diffusion increments. The optimum va for the product is approximately 0.5 with lower values of P being associated with small value
, 9 e
- r
)
of k. Ultimately, k was set to 0.5, and the optimum n value for Ao was determ The search algortthm may have talled to converge on an optimum value fo are there are relatively few data in the data set that represent high wind speed Comotete Model Ap a Selectiona of values for kgand'Ao completes the model revision. The equat p , gg s,a, ggp ,and Ap,,aare: .
A ep' = (2)(0,5)(0.7)8(1000)8[1 -(1 + )ay(- ))
8 (12)
- 5.0x10 [1 -(1 + )exp(q * ))
A o,,' = (2)(0.5)(0.7)8(100)8[1 -(1 + )exp( })
(13)
- 5.0 x10sgg .(g . ),xp(. ))
- Aog t n = (2)(0.5)(0.02U8)8 u =8
/4/u. A (1-(1 +1)exp( /4/u=))
i (14)
- 4.0 x10s U 8A a 8 g3 gg/4g),xpg-[ax))
4 ,
l and I
A o,8 = (2)(0.5)(0.01U'):_ A u=s(2+z/L)8gy gg u .(2 + q ) ,xp( r u (2 + M ))
/A /K A (15)
= 1.0 x10d a8(2+z/L)8[1 -(1 + a(2.Wx),,pg_-a(2Wx))
/K (A -
respectively.
Figure 3 compares the predicted and measured concentrations for the data se
= 0.5 and Ao g = 0.70. The median ratio between predicted and measured concentrations is 1.25, and slightly more than 85% of the predicted concentrations are within a facto the measured values.
e
. A 10 l
1
n.-.
,' ) -
.s
- 1E42
( 1E. . 4
, / i
/
{ 1E 02 ^
M;-
1E 04 *
, s a ,
1E 06 ff .
1E C8
/
1E 10 1E-08 1E 07 1E 06 1E-05 1E od 1E-03 1E 02 1E-01 1 1 Observed XIQ f; Figure 3.
Comparison of revised model concentration predictions with observed values.
1' \
1 l l Medal Evaluation
\
Early evaluation of the current NRC wake models indicated that the models significantly overpredicted concentrations during light winds (Ramsdell 1988). This tenden is shown in Figures 4 and 5. NRC guidance related to the use of these models states that ~
the models should be used do determine X/O values that are exceeded no more the time. Typically these highest values are associated with wind speeds of about 1 m/s.
Figures 4 and 5 show that the current models almost always overpredict concentration an order of magnitude when the wind speed is 1 m/s or less and that on the average they predict concentrations that are about 2 orders of magnitude too high.
The revised wake modelincludes corrections to the diffusion coefficients specifically addressed to improving model performance at low wind speeds. Figure 8 shows the variation of ratio between predicted and observed concentrations of the revised model as a function of wind speed. Compared with the current NRC models, the revised model has less l
^
11 e
O
,_.r- --
_ _ , _ - . . . - . . - ~ - - - - " ~ ~ ~ ~ ~ " '
e - '
-l' 5 . -~
' - .t .--1..r. ; am.. ; :
t 1E+03 * ^
i _ . , ,
_m pg .__
-P ,
a r, . _
t.
h Mae;5E
- s. . r - - -
95 == ::
r-_
7 -4. --- .C._.. 3" f.I. MM; j.p g.e,.cs . - q; ,
. _..._ : - pl.'rs,
-H
- gg ,g, AtMr T ~
- - .... . . = . :- ,
s ~
~ ~ ~ ~
i,
= -as, mn 2- 3
{
-T ; d5
_ g. 4 6 --
_a
- - 1 i m _.
- .;u = ,. .
w~M84 ama _ ' = -
a g
c
== p a
M'"** _-^p e=a a -
)
M es"""$ I'-
l i . _ . .
_- ~
__ l 1
1E42 g -- _- .
.y.' _ .. . - - . - . p : ,.
- .-- - * - _1
- e;; .* g gggg.g.j. 3 i _. : :: - n. .::: : ::. ... .._ ..
1E43 - _ . . . _ ... ...
i 4 0 2 4 4
} 8 10
, 12 i Wind Speed (m/s)
Mgure 4.
\ Biasspeed. in Murphy-Campe model concentration predictions a wind g
1E+04 1 . _ -__ , . _. - - -
e
_m t=im m- -
l.
g' i
e r_-, - i
\
1E+03 4., _m . _ - . - i
{ .
t
~'
- _ _u ._ _ _- -
8 . - __. ; ... 4 ::~:._m.
= ~r::
< a>
J . _ "l> 3 1E+02 ^^
j "if' e :.=: .
-T -
1
} .-
,7 T . : ;= --;; : +h == in t ::=i i
- c-
- 1E+01 . b a. as tt <. : *::~
.. . T**...
.c . m-_ ,e_
4 an,,
j g gg..
g 58-c 1.u-.m- -~ : .y .m - - .
"7g .:4; --_ * ,_ ;_ ,: - 2d m ------
3,w. m _ .
n ,
I 9 - _
1E+00 .- - --. ^ '**I ww = m'
~ "
2 j
e- ==
- . _ m. - - -
- w _4
_e i . . , . A g.q_ -- ;
- 1E41 -- -
. w"-j _ , p_:__
l y -
=n-- .u.
t
. = i
- 1E 02 , _ __
- z =_ _
. - __ ..-.=:- _..._.__..:
i
. . . . .=... - - - - ~ - - - - -
1E43 '
1 c 2 4 "
{
e a 10 12 i
- Wind Speed (m/s)
Mgure5. .
function of wind speed.Blas in Reputatory Guide 1.145 m '
s 12 i .
4 4
a n
4 j
i
)
_ I
)
1E+04 3=
n-_w n =-
my.yp g
/ 1E+03 m_,
h p 4 __ _ . . . . _ -
11E+02 . _ ___ __
y_3 _
gy .
$1E+01 esy-ygg **
~.- g. g- nm --
-m 2; w . _w -c =; w g.
~ _. . --==
w m ..
E 1E+00 8-d"" g *
- r 2 - *e I
- A N$5[#5$~ ' A * *2 ~ M '~~~ L J= -
" ~ '
^
7 -
-u -
1
"5"NEN 1E-01 * ._ . -
- g. = . - - =
a,,= ,*,. ^ "W'd ' ' i .,-
gpw E .
wp "_ aI , _
1E 02 pg_ g_. m..
u #-
, a 1E !
o 2 4 6 8 10 12 Wind Speed (m/s)
Figure 6.
Bias speed. in revised model concentration prediction as a func t
tendency to overpredict concentrations at low wind .
oes appear that speed However there is still a slight tendency toward overprediction for wind speed improvement in model performance is furthers illustrated an 3 m/s. The requency in the cum distributions shown in Figure 7. The median reratios cted by between c current NRC models and the maximum observed concentrations .
addition, concentrations predicted by the current n u e of models ar the observed concentrations only about 60% ofrythe . time Figure 7 clea shows'that the improvement in model performance is gained bywithout reducing mod algn!!icantly increasing underpredictions.
ALTERNATIVE MODELS
.a .
Gaussian plume model for estimating s. Two o e co minimum dilution models were identified as potential attoma 13 1
.;" . 1
= 1 1E+04 !
g.;;p %g,13.ui. 7am,y,,= ,
m [gi3pgmm ;133g331.._., um
. == .;;,q;1 g
-g
( 1E+03 pgg.eu
- m. _
m.. .au
.'4
. -...m.
. .. ~, .u:.1 . ..ggs. n.u. ;... m . . . . ' . . g, I
k 1E+02pi;j!; pigs us.gg I I I '
t" - - a ~un ==
__35;Reguistory "mg:wii5 a::it afic I
- n.i;p:;gpg3i...g.,.g
' Guide 1.145 + ' f _ f..::
' + a 1E+01 lhakifiF.!!?;
2.:i-ii3.:dilii
- Ediifreir:ii?
ii:!:iHi 4 pF.i:::n: .
I p'-
G;n liiiiij: jiQg_'u. dii:ithm I
, yy y 7 ,
-T- --
1E+00 -
r, T sm %
. . - pg. .'
WeaMiB iw i i
- p. _.lsE .:._53!!!!:
p, e_. Revised Model y '
1E41
- i i ;
q :.1ti jg"- "".:95 r3 ";@mpe ,
p.
' _ !!il aut ". ." _
9 1 i . . ,
t i 1E02 g,I t l I
........= f I
.....<j..- m.m.u. mag ..gu 33..
i 1
- .mm.
ea.es . t I '
1E43 ! ' ' ! '
' i 0% 10% 20% 30% 40% 60% so% 70%
40% 90% 100 %
Cumuistve Frequency Figure 7.
Comparison of cumulativo frequency distributions of predicted to
. observed concentration ratios for the MuPhy Campe, Regulatory Guide 1.145 and revised models.
model. Minimum dilution models attempt to provide a lower bound to dilution. When '
minimum dilution models are used to estimate the concentration, they should estima upper bound on the concentration. This is in contrast with typical diffusion models which
~
attempt to estimate the average concentration at a position in the plume.
The first shamative model (Wilson and Chul 1994) predicts maximum concentrati "
as a function of wind speed, building area, and downwind distance, and the second m (Nilson and Lamb 1994) predicts maximum concentrations as a function of wind direction fluctuations, and stretched string' distance. Stretched string distance is the distance between the release point and receptor without passing through the structu models have a correction term for initial dilution for releases through stacks and vents.
Wilson-Chul Model -
The Wilson-Chui model was developed and tested using wind tunnel data to estim minimum dilution in plumes released from buildihg vents and short stacks'. The mode '
development does not make or depend on a Gaussian assumption. When reformuiste 14
e t~
t '. .
estimate concentrations assuming equal eMluent and ambient air densities, the Wilson Chul g
modelis (dQ) mas *
(M+ DD)' (16) where F, is the vent flow, D, is an initial dilution, and DD is the downwind dilution. Initial dilution, which is a function of the ratio between the exhaust extt velocity and the wi b -
~
D = 1.0 + 7.0( )'
(17) where W,is the exhaust velocity. The downwind dilution is given by DD = 0.25SX( )'" .
(tB) q where s is the stretched string distance.
[ Maximum concentrations were predicted for each of the 265 concentrations in the building surface data using the Wdson-Chui model to evaluate model performance in the vicinity of full scale structures. The resutts of the calculations are compared with the observed concentrations in Figure 8. The Wilson-Chui model consistently overpredicts the observed concentrations, as expected. No observed concentrations were higher than the model predictions. Therefore, it is reasonable to conclude that the model provides an of the !
upper bound to maximum concentrations in the atmosphere as well as in the wind tunnel.
Wilson 1.amb Model The Wilson Lamb model was developed using field data rather than wind tunnel data.
The form of the Wilson l.amb model when reformulated to calculate max is identical to the form of the Wason-Chui model, i.e., Equation (16). The differences betwe the models comes in specification of the initial and downwind dilution in the Wilson Lamb model initial dilution is given by Briggs (1975) relationship -
D = 1.0 + 13.0( ) .
(19) 15
)
,,,,. , 1Ek1 - _ _
_ _ _ .e i i
(
1 -
1E+00 w = r.-
2
. . . iu i i . . . < '
n- m. -
m u. .. _ a m -- _ a ww33%gg
, , . . . . . t i .. .,:
1E41 ,
ar..u"
-~
_ .g . 6 . . . . . .
,.--uwp 4 4 n _gygg
. 2
,yw
. _ . _- - z . = --- EF ~~~'
D . i . , ,c .....a m-=W1M8%
~
R ~
z= k1 . . . .j 1' = -a, ...
a s m +.a w r mytey4Qa.-w- g q =gA!iild m3==e
- 3'ME5*.':is~--
' ' a 2 "*-"W ^
v '
1E43 ' ' ' *
^ J
w - ~ n -- mue _- =
^
w _m"~
0 4 4 $4444 -
4 4 4 4444e d~ 4 4 & 8446 [a ( & 8 446e
- m. - % , -
i 4 8 6 4444
-..--.._.r s -----r ,...
-m f # - ;
& { l* s 1
l i8 i a )4 [} I 1 l 48a8 l j6 e i !Il *
$ } #*l
__ . uc- mismased H-
= amu 1E 06 == m'g' ' ' " 'g_._x';g e e' ' ". .- ' _e' m ' .
- ___,': m 'ine 4fy,assa= '
1E-07 ' i" ' ' ' * ' ' !' ' ' '" '
1E47 1E46 1E45 \
1E44 1E-03 1E42
. Observed X/Q Figure 8.
Comparison of Wilson-Chui model concentration predictions with observed values.
l i
The downwind dilution is given by l i
DD = jus .
@)
where p is related to the o, the standard deviation of the wind direction fluctuations (i radians) by
$ = 0.039 + 0.17o, .
(21)
If o,is not svallable, the ensemble mean value p = 0.089 may be used. Note that in the W" son L.amb model the maximum concentration is not a function of building area.
Maximum concentrations predicted by the Wilson-Lamb model are compared with the measured concentrations in the building surface data set in Figure 9. The relationship between the Wilson-Lamb model predictions and the measured concentrations is almilar t that between the Wilson Chui mode 1 predictions and the measured concentrations. The model generally overpredicts the concentrations and provides o a ress' nable estimate of the maximum concentrations.
( .
~
16 I
. _ . - . . _ -.--- - - - - - ~ ' ' ' ' ^ - ~ ~~ ~ ~
l~ U
)
1E401 - :
,_ - 6=- _ _- .
_-2 y _ ,,, , m
, ,_ . . . , , . i .. .t . , , . i4 4n meaf.ae w mn.c.m m w-we, n %=. a o ,
l p ' '
-i. .
.. . m
- .+ a 1E' 41 ' ' # ' ' ' ' '
lumia mi6erm N.T. :E s m=3 ,.
7-
. m_sC 4-se rrr
- q 43-f
^
^ _
{1E42 ' ' ' *^ ' ' - T' ?
.- - :n y'e g ggjy' p' g'p'* *
.ug,I y'
~
- y. . = 3 4 m .; -
-8 1E43 ' * *7' * 'A ' ' "
-e--
_* 3 .'n -_%g' w- == - 'm u un
'y 2
1E-04 _ _ ' ' ' ' .* f_ ' ' ' " 'Ti*'"'
"l
'i' ' '
t p g a 46 6
i t i e ai f, wm_ w g
- 4 ) i8l e
( 4 eit6-
, a, -
4 a 6 , a s ,.:
1 i e ll a fi l i i l ill I w-=_. emy ^
( i i t lit
! l 6 l illi I .I Iitii gg I . e i t 1i; i e i i ; iii' i e i' fil i I ' I: iii i i i ':i.
1E47 1E46 1E45 1E44 1E43 1E-02 Observed X/Q '
Figure 9.
Comparison of Wilson l.amb model concentration predictions with observed values.
t Comnarlson of the Revised Model with the Attornatives l
Figures 8 and 9 establish that the two minimum dilution models provide reasonable estimates of maximum concentrations on and adjacent to buildings in the vicinity of a release point. Neither modelis a Gaussian plume model. Figure 10 compares concentrations predicted by the revised model with the set of concentrations used to evaluate the minimum '
dilution model. The pattom of revised model predictions is not much different than the pettom for either of the other models. The revised model underpredicts eight concentrations, in comparison with no underpredictions by the Wilson Chul model and one underprediction by the Wilson Lamb model. However, the revised model should underpredict, more fre trhn the minimum dilution models because the revised modelis designed to everage maximum value in a plume rather than the absolute maximum.
Figure 11 compares the predictions of the three'models directly. The revised modelis most likely to underpredict conoontrations. Yet, only about 3% of the concentrations are 8
9 1'T e.
, m.
URN nm } mm u, -m,um ww-=m yLtisi 3.ew.q g u yi
. - ....i ,,, i...n ._
.. . - , 1E+00 -- -
i 4
...m , e..- i . -
i n = .+ ... ,.. w.,m m - -
- m w w w: .= wen,a i t IE-01 . _3
, ,ii ,
- .
- ( d
--. p I
\ e z.
--~s-e.Y m myg;w
,~4'~'
.
4
. i .,
- 3 s . -- .
c:
1E44
- p. . , . . og, , , , ,g .4i.n,ep- e i ,
i i ,' ii+ei*
=
ae ii++.i
. . -u
~
," _' .y ' ' ' '--n ' ' '"
1E45 , ' '" '"
2 i
- = - ~.wn m .u r. = .y w wu i l M4
{ it i 46 1E-06 4 ll ll l s 6 alti l I III ll l I i l
- umM3mEE":::f'MEBE4MWE!'LT' P 1E47 ' *' ' ' ' "i ' i ' i ' ' i' 1E47 if 1E46 1E45 1E44 1E43
. 1E42
. Observed XC Figure 10.
i Comparison of revised model concentration estimates with observed l values in the building surface data set. ~
i 1E+05 g . . _ . . . . . . .
.. ...=
afei. um :.
q ;.;l 2 .
,2, i _.
, 1E+04 ,
- mm, e 4
.g 7
e..==
) 1E+03 u .
1 j .=._1 mu
] e ;;=
\ >
j ;
{ j1E+02 ! .;. Waso m ' >
w m ms
=== 5 " ~
)
1 l 1E+01 a g s g, , .
p' xe =
1E+00 ,
Ik Revtsed Model
, . 3
.w : sawuw- =
enu. m.im-1E41 )y . . _ . > > =~ '
mi .;e.. <
1E42 .
o% 30% 40%
- 80% 30% 100%
Cumulative Frequency Figure 11.
Comparison of cumulative frequency distributions for the ratios of predicted to observed concentrations for the Wilson-Chul, Wdsor.. Lamb, and revised models based for the building surface data set.
18 9
- 1
( .
i underpredicted by the revised model, and only 1.5% gre undsrpredict I
factor of 2. With the exception of four concentrations underpredicte
{
of four, cumulative frequency distributions for the revised model an are nearly identica!.
j .
3 i NEAR-F1FLD CONCENTRATION manMATEiit -
l The fourth recommendation of the peer group was to evaluate th a'
} an appropriate subset of the experimental data. Figures 10 and 11 pro l
indication that the revised model is useful for estimating concentr j building surfaces, even though the Gaussian model may not be stri purpose. As a final check, concentrations in the building surface data we
{
j concentrations measured near the release point tr the ground level release retutting data, consisting of 402 concentration measurements were com predictions as a function of distance from the release point.
Figures 12,13 and 14 show
{
- ratios of the predicted to observed concentrations for the Wilson Chu I revised model, respectively, as functions of normalized distance. In
! level release data, shown by the near field markers in these figures, 1
l is the downwind distance dMded by the square root of the building a l building surface data set, the normalized distance is the stretche I
the square root of the building area. The ratios shown in the figures ind models are conservative near the release point because they tend to overes
{ concentrations.
4 i .
i 4
1g 4
i
.l 4
3 1E+05 '
o Near Field . Bldg Surfaces 1E+04 ^ (* )
- ( . ."' . - - *
- l 1
g -
.t..
. l i
}
1E +03 -
, , ,* .. n l',g, *
- .. , i
.. ,i
- ^4 g ,.! ,
a-l 1E+02 g
^
+
a
. ! p, 1E+01 Q4a.g{.....A.:.,
. 4!, @. ', g' _.
i II r : 3., 52-1E+00 t4 ^ ,! .:; ,". :g3n . , . -
1
~
r..
I-
-l '
- g i
' 1E 01 l -
l i
I. \
1E-02 i 0 1 2 3 4 5 d Normalized Distance l Figure 12. Ratios of predicted to observed concentrations for the Wilson Chul model as a function of normalized distance.
1
- 1E+05 , , , ,
1
) .
.' .i !
- Near Field
- Bldg Surfaces
]' 1E+04
.l ,
i k
1E+03
} '*..] ^
! j
.i . & . -.T . . . i
' ; *': . g '2'.s *
.. l .:'
^ '
- 1E+02 " ". b' " '*
8 g .
- j .et at ie . - ,
a
- 's . .'*' *** . .* '5 ^*, '. * *
. e' S 1E+01 E
- I I
5-
.t 1,", )' t j '.. . < .
4 l l1E+00 "* "'. t' ' * . .8l'. * ' ,',,
t ,i;-
i p**;.
- g e - -
l .
- 1E-01 e
- 1E ; O 1 2 3 4 5 Normalized Distance i
- Figure 13.
2 Ratios of predicted to observed concentrations for the Wilson Lamb model as a function of normalized distance.
20 4
i
-t O
a a
.a- .- - ._ - - -
,-~~~~^~~ ~ ~~ '
i
, .. . - SE+05 <
t
, . Near Field
- Sids Surfaces
( 1E+04 - i --
$# > *h k
- 1E+03 11 I
)
, .. ;ti '
1E+02 'l *J 'W 8 '
- I
{ 15'$ ([ ktj*o u
, i
- N '* .
f1E+01
- m ,
r, ',', F g. .* . ..** .' . . . ..
1 3 g.oo x- ,
. i %" !:y... 1. .
1
~
. - ,. n'. . :
f-? gl 1E-01 ' *
. 6 1E42 0 1 2 3 4 5
Normalized Distance 1
Figure 14.
Ratios of predicted to observed concentrations for the revised mo a function of normalized distance.
! The final comparison between models is a direct comparison of the cu frequency distributions in Figure 15. The distributions of ratios from the revised model are almost identical. This indicates that these two models resutts N used in control room habitability assessments. The Wilson C '
more of the concentrations than either of the other two model i that might be responsible for the larger number of overpredictions by th is that the Wilson-Chui model does not have a means of accounting at low wind speeds due to meander. Both the Wilson Lamb model and the account for enhanced dispersion at low wind speeds.
- CONCLUSIONS 1
Evaluation of building wake dispersion models beginning in the mid 198 that models currently recommended in NRC guidance to licensees ten 1 overpredict concentrations during low wind speed conditions. As a result, the p used in evaluation of control room habitability q and the conse' uences of design
~
21 -
a
_. .-- - - ~ ~ '
. )
4y e e e
- gp _ ii.ig;;i_ ;;;a::: q;;i,;;;i;,i.,i
'"* -. : ;,3,,
,gg i. gg,,, ;j; , ; ,. g g .g y
f ..:!!.::b.:ni g.pmi:-isd ::::z.::na_
- sua
- n .;n:n:n;ii.:ii h
- d:id d}. _
, k1E+03[.ii!"lii !!@:Ed i g../*"
"r
- _'ifgj'_'E""9
- ;;f::.:-ni'di!'iE #
ni 3. .. . . . .. . .
, , ~ . .
u;. , f ! l..
g-g h....
.u
. m . . . it ";". . ,u
..na::. =:=
1E+01 ' 0 '
^=++1:::s mi e
- 3 *""~'
-'- . '..# a.m=-w #- - eue-e 1r ,'
mmmm: =e=m:ig 1E+00 # ' '
A in-i.=.p. ...:asiaiis; mm=i=.e" = - = ' -"
3== m l -
9 ,
=u.
=m:ga 1E41 '
,y=:-:.m =3::
- = g g3 e gg
_ _ _ a.c 1E C2 I l 0% 20%
! 40% 60% 80% 100 %
! Cumulative Frequency ~
Figure 15.
Cumulative frequency distributions of the ratios between predicted a observed concentrations for the Wilson Chui, Wilson Lamb, and revis
. model for all data near the release point.
r accidents were fett to be overty conservative. A new model was developed predicting concentrations near buildings that did not overpredict concentration
! speed. The 1990 model has roosntly undergone peer review. This letter repo disposition of the primary recommendations of the peer review panel. Those recommendations were 1) l The turbulence increment generated by buildings should be assumed to to the wind speed in accordance with accepted theory and physical reas 2)
, , The effects of meander during low wind speed conditions s'hould b in the wakes. model, but the treatment abould be separate from the treatm -
- 3) '
An approach to determining concentrations other than straight-line Gaussia should the be considered when releases are from a building and receptors are building.
4)
Appropriate subsets of the available data should be used to e' valuate the m the suggested changes have been made.
( .
- 22 i ,
4"
~
y" , ,
t .
% ) ,
- \
e
( in response to the recommendations, the 1990 model was revised to explicitly treat I enhanced dispersion in the vicinity of buildings as a combination of the effects of low and high wind speed phenomena. The low wind speed component of'the enhanced dispersion in the revised model decreases with increasing wind speed. in contrast, the high wind speed 1 component increases with increasing wind speed. Turbulence data collected in the vicinity of bu!! dings has been used to model the increase of turbulence in wakes that is responsible for enhanced dispersion at high wind speeds. DMuston data coliseted in experiments at seven reactors indicate that the revised model is a significant improvement over the building wake models recommended in existing NRC guidance to licensees.
Two attomative, non Gaussian models developed tu estimate minimum dilution (maximum concentrations) in plumes from building stacks and vents were identified and tested using a different set of data from experiments at three of the reactors. These models appear to predict an upper bound for concentrations in the immediate vicinity of the release point. Concentrations predicted by the revised model for the same data also tend to be higher than the measured values. The differences between the predictions of the minimum dilution models and those of the revised model fall within the range of differences that is to i
be expected given the intended bias of the minimum dilution models. Further comparison of the revised model with the minimum dilution modelindicates that all of the models tend to be l
conse'.vative near the release point and become less conservative as the distance from the release point increases. Cumulative frequency distribution of the ratios of predicted to l observed concentrations for the revised model is nearly identical to the distribution for the Wdson Lamb minimum dilution model.
The revised model incorporates the changes recommended by the peer review panel and concentration predictions near release points that are comparable to the concentrations predicted by minimum dilution models. Therefore, the revised modelis considered to be appropriate for use in estimating concentrations for control room habitability assessments.
The revised model is also considered to be appropriate for use in estimating concentrations in the near field for use in evaluating the consequences of design basis' accidents.
e r
l :
' 23 I
! )
, -} . .
- REFERENCES
! \
. Arripact Analyses. American Meteorological .
vronmentaf Socie i
i Atmospheric Envitcament26B(4): 513 517.
Briggs, G. A.n. A. H. Hube
\
i Reactor Complex on Effluent Concentration.' N .
i Nuclear Station under Low Wand Speed 60289, Gene,ra! Public Utilities Service CorporatJon. ,
e Island inv ocktt No.
i s
j tsiltzer, N. F.1965. Aerodynamic EWects of Large Reactor Comp mospheric 1
Turbulence Commission, andIdaho.
Idaho Falls, Dmusion.10012041, Idaho Operations y Offi i
i Martin D. O. and Tikvart J. A. (1968 A general
! the effects on air quality of one or m) ore source. atmospheric dispersion m APCA, St. Paul, Minnesota,18p. presented at the 61st Annual Meeting of the i ,
Mumhy, K G. and K M. Campe.1974.
- Nuclear power a on system plant contr !
i
{ ,-
i Conference, San Francisco, Califomia, Washington, D.C. eaning ss on, CONF i
i Model Cube.' Atmospheric Environment 16():207 und a 222.O
! i
! NUREG/CR 5055, U.S. Nuclear Regulatory , ..
ments. Com Environment. 24B(3):377 388.Ramsdell, J. V. 1990.mospher/c .
' Diffusion in B '
Start, G. E., J. F. Cate, C. R. Dickson, N. R. Ricks,"G. R. Ack . .
1978. RanchoCommission, Nuclear Regulatory Seco Building Wake Washington, D.C. EWects on Atmospheric
.S.
Dm Start, G. E., N. F. Mukari, J. F. Sagendorf, J. H. Cee, and C. R.
Building Wake EWects on Atmospheric DMusion. NOAA Tech Air Resources Laboratory, Silver Springs, Maryland. -
coefficients in atmospheric diffusion. Atmospher 24
l .. M v I. )
l . , . . .. ,
- ') -
.A i -
[ Thulliier, R. H. and R. M. Mancuso.1980. Building EMacts on EWluent Dispersion from Roo '
Vents at Cailfomla. NuclearPowerMants. EPRI NP 1380, Bactric Power Research institute, Palo I
- Thuillier, R. H.1982. " Dispersion Characte'ristics in the Lee of Complex Structures." Joumal of the Nr Pollution Control Association 32:526 532. .
Thulliier, R. H.1988. .
l
),
l Wihon, D. J. and E. H. Chui.1994. "Innuencie of Ning Size on Rooftop Dispersion of i
Exnsust Gas.' Atmospheric ErMronment 288l): .
{
Wilson, D. J. and B. J. Lamb.1994. " Dispersion of Exhaust Gases from Roof Lsvel Stack and Vents on a Laboratory Building." /4,,@,aric ErMronment 28B():.
i i .
) .
t as l
\