tyyy.y;
llh
EPA-520/5-76-015
rrrrrrt

AIR PATHWAY EXPOSURE MODEL
VALIDATION STUDY
AT THE
MONTICELLO
NUCLEAR GENERATING PLANT

U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Radiation Programs

-------
EPA- 5 2 0 /5 - 7 6 - 0 1 5
AIR PATHWAY EXPOSURE MODEL
VALIDATION STUDY
AT THE
MONTICELLO
NUCLEAR GENERATING PLANT
J. E. Partridge
J. A. Broadway
C. R. Phillips
S. T. Windham
Eastern Environmental Radiation Facility
P. 0. Box 3009
Montgomery, Alabama 36109
C. B. Nelson
Environmental Analysis Division (AW-461)
Waterside Mall East
401 M Street, S. W.
Washington, DC 20460
September 1976
U. S. ENVIRONMENTAL PROTECTION AGENCY
Office of Radiation Programs
Washington, DC 2 0460

-------
FOREWORD
The Office of Radiation Programs carries out a
national program designed to evaluate the exposure of man
to ionizing and nonionizing radiation, and to promote the
development of controls necessary to protect the public
health and safety and assure environmental quality.
Technical reports allow comprehensive and rapid
publishing of the results of Office of Radiation
Programs' intramural and contract projects. The reports
are distributed to State and local radiological health
offices, Office of Radiation Programs' technical and ad-
visory committees, universities, laboratories, schools,
the press, and other interested groups and individuals.
These reports are also included in the collections of the
Library of Congress and the National Technical
Information Service.
I encourage readers of these reports to inform the
Office of Radiation Programs of any omissions or errors.
Your additional comments or requests for further infor-
mation are also solicited.
W. D. Rowe, Ph.D.
Deputy Assistant Administrator
for Radiation Programs
i

-------
PREFACE
The Eastern Environmental Radiation Facility (EERF)
participates in the identification of solutions to prob-
lem areas as defined by the Office of Radiation Programs.
The Facility provides analytical capability for evalua-
tion and assessment of radiation sources through environ-
mental studies and surveillance and analysis. The EERF
provides technical assistance to the State and local
health departments in their radiological health programs
and provides special analytical support for Environmental
Protection Agency Regional Offices and other federal
government agencies as requested.
This study is one of several current projects which
the EERF is conducting to assess .environmental radiation
contributions from fixed nuc"
Charles R. Porter
Director
Eastern Environmental Radiation Facility

-------
CONTENTS
Page
FOREWORD		i
PREFACE		ii
LIST OF TABLES AND FIGURES		iii
ABSTRACT		iv
INTRODUCTION		1
OBJECTIVES		1
DESCRIPTION OF FACILITY AND SITE		2
STUDY DESIGN		2
RESULTS		10
DISCUSSION		22
SUMMARY AND CONCLUSIONS		2 3
REFERENCES		25
TABLES
1.	Monticello PIC site locations		3
2.	Calculated relative noble gas releases		8
3.	Differences among exposure models		9
4.	Total exposure		11
5.	Net exposure		12
6.	Predicted and measured exposure values		16
7.	Predicted/measured exposure ratios for four
models		17
8.	Measured and predicted values based on field
gaseous xenon concentration (January 28, 1974)..	21
FIGURES
1.	Pressurized ionization chamber locations		4
2.	Instruments housed in plywood boxes		5
3.	Strip chart recording		5
4.	Comparison of TLD & PIC (total exposure)		13
5.	Comparison of TLD & PIC (net exposure)		14
iii

-------
ABSTRACT
The results of a study designed to improve the
methodology for estimating the population exposures
resulting from nuclear power plant gaseous effluents
are given. The primary objective of this study was
to validate a mathematical model (AIREM) for estimat-
ing radiation exposures due to atmospheric radioactive
releases. This validation was accomplished by com-
paring the model predictions with actual field measure-
ments made using pressurized ionization chambers and
thermoluminescent dosimeters. Use of this model for
predicting external exposures was shown to be quite
acceptable for most applications. The usefulness of
pressurized ionization chambers for making low-level
exposure measurements was also demonstrated by this
study.
iv

-------
Introduction
The proliferation of nuclear power plants, the
increased cost of environmental monitoring, and the
need to measure extremely low population exposure
levels have necessitated the use of new and innova-
tive techniques in nuclear power plant radiation
surveillance. One such technique is the use of
mathematical exposure models to supplement surveil-
lance programs. These exposure models when used as
an integral part of facility monitoring, hopefully
will provide maximum necessary assurances with mini-
mum expenditure of resources. The Office of
Radiation Programs (ORP) of the Environmental
Protection Agency (EPA) has a mathematical model
(AIREM) for estimating the exposure to populations
within 80 km of operating nuclear facilities due to
atmospheric releases of radioactivity. The gaseous
effluent data provided by each reactor in accordance
with plant technical specifications are used as the
basic input data for this model.
Objectives
The primary objective of this study was to vali-
date the ORP mathematical model (AIREM) for estimat-
ing radiation exposures due to atmospheric radio-
active releases. This validation was to be accom-
plished by comparing the model prediction with actual
field measurements. The field measurements were made
using pressurized ionization chambers and thermolumi-
nescent dosimeters. The predictions obtained using
the AIREM model were also compared to the predictions
of several other existing models.
In addition to this primary objective, a secon-
dary objective was to compare exposure measurements
made using the pressurized ionization chambers (PIC's)
and thermoluminescent dosimeters (TLD's).
The problem of dose rates from particulate re-
leases was not addressed in this study. A model vali-
dation study based on particulates is intractable due
to the extremely small quantity of particulates re-
leased.

-------
III.	Description of Facility and Site
The Monticello Nuclear Generating Plant is an
operating boiling water reactor owned by the
Northern States Power (NSP) Company and is located
about 55 km northwest of Minneapolis - St. Paul,
Minnesota (2). It has an authorized power level
of approximately 545 MWe and started commercial
operation in June 1971.
At the time of the study, the plant design
allowed radioactive gaseous effluents to be held
for approximately 30 minutes to permit decay of
the short-lived noble gases and then filtered prior
to release to the atmosphere from a 100-m stack.
These effluents are also monitored prior to release.
An extended holdup system (minimum 50 hours) has
been installed to replace the 30-minute system (3).
The plant site is located about 5 km north-
west of Monticello, Minnesota, (population approx-
imately 2,000) on the south bank of the Mississippi
River in Wright County, Minnesota. The nearest
property boundary is 50 0 m south of the reactor
building and the nearest house is 850 m to the
south. St. Cloud, Minnesota, (population approxi-
mately 40,000) 35 km northwest of the site, is the
nearest large city.
The land surrounding the site is predominantly
rural. There are a few small communities within a
25-km radius of the site. The terrain is heavily
wooded along the river while away from the river
the terrain is relatively level and largely under
cultivation.
IV.	Study Design
To meet the objectives of this study, radia-
tion exposure measurements, meteorological data,
and gaseous release rates were obtained on a con-
tinuous basis for approximately 9 months. The
data collection commenced on August 28, 1973, and
continued until May 13, 1974. The reactor was
shut down for refueling from March 15, 1974,
through May 20, 1974.
2

-------
A. Pressurized Ionization Chambers
The major portion of the field study con-
sisted of continuous ambient radiation expo-
sure measurements using pressurized ionization
chambers (PIC's). The PIC's used in this study
are commercially available instruments similar
to those described by DeCampo, et al (4). The
detector is a spherical stainless steel chamber,
25 cm in diameter with a wall thickness of 0.3
cm. The chamber is filled with argon to a
pressure of 1900 cm Hg (0° C).
The ionization chambers were operated off-
site at the locations shown in table 1 and
figure 1. These locations were selected based
on predominant wind directions and the points
of maximum deposition.
Table 1
Monticello PIC site locations
Site
Distance from stack
(km)
Direction from stack
(degrees)
A
1.4
138
B
3.3
133
C
3.2
156
D
2.3
102
3

-------
Figure 1. Pressurized ionization chamber locations
The instruments were housed in plywood boxes
(figure 2) approximately 2 m above ground. All
locations were supplied with AC power for operation
of the instruments.
The readout of the instruments was in the form
of a strip chart recording (figure 3). The strip
chart operated at a speed of 10.16 cm per hour.
The charts were collected and changed on a weekly
basis and mailed to Eastern Environmental Radiation
Facility (EERF) for data reduction.
Data reduction was accomplished by planimeter
integration of the strip charts. The integral ex-
posures were determined on 2-hour intervals. The
plant contribution was determined by subtracting
the natural background from each 2-hour integral.
During the lattef portions of this study, a
commercially available integrator module was
adapted and tested by the EERF (5). The inte-
grator, when used in conjunction with the PIC, can
be used to determine integral exposures without the
need for tedious planimeter integration.
4

-------

-------
B.	TLD Measurements
Additional radiation exposure measure-
ments were made on a continuous basis using
calcium fluoride :manganese activated (CaF^Mn)
thermoluminescent dosimeters. These dosimeters
are commercially available glass-bulb type do-
simeters, complete with energy compensation
shields to reduce the over response of CaF2tMn
to low energy radiation (6,7).
Three TLD1s were located inside the ply-
wood boxes at each PIC site. The TLD1s were
read out on either a 1- or 2-month interval.
All annealing and readout of dosimeters was
performed in the field near the sites to avoid
any errors that might be introduced by trans-
porting the dosimeters.
TLD monitoring during the 2 months of
reactor shutdown was used to approximate the
natural background for each site. These back-
ground values were subsequently subtracted
from TLD measurements taken during periods of
plant operation to determine the net or facil-
ity contribution.
C.	Meteorological and Gaseous Release Data
The meteorological data used were taken
from the 42.6 m tower located approximately
1.0 km ESE of the 100 m release stack. Data
taken at top of the tower consisted of strip
chart recordings of wind speed and wind direc-
tion and were assumed representative of condi-
tions at the stack release point. The mean
wind direction, wind speed, and wind direction
range were tabulated on 2-hour intervals as ob-
tained from manual analysis of the strip charts.
Temperature lapse rate data were not available
on site and atmospheric stability classes were
estimated from the 2-hour range of wind direc-
tion using the procedure described by Markee
(8) and the tables of Turner (9).
During site visits, the wind direction
recording equipment was observed to indicate
winds 180° out of phase with the true wind
6

-------
direction at unpredictable times. In an
attempt to minimize errors caused by this
anomalous behavior, the wind directions re-
corded at the site were compared on an hourly
basis with those obtained from the National
Weather Service at St. Cloud, Minnesota, and
the site data were thereby validated. During
periods in which the site data were found to
be in error, the recorded site direction was
rotated 180° to obtain the true wind direc-
tion. Due to the data cross-checking procedure
employed, this anamoly did not result in appre-
ciable error. However, the situation does indi-
cate the importance of a facility-maintained
quality assurance program for meteorological
data.
During the latter portion of this study,
NSP was installing a new meteorological system.
This system consisted of a 100 m tower and auto-
matic data reduction equipment. Data from the
new tower were not available during this study.
The availability of summary meteorological data
in future studies would significantly reduce
the manual efforts required.
The gaseous release rates used in this
study were obtained from the plant operating
reports (10) for the period of the study. Pre-
liminary sample calculations indicated that the
majority of the gamma exposure rate at ground
level was due to six nuclides emitted from the
stack, therefore, all subsequent calculations
of exposure rate were based on releases of
these nuclides. These six gaseous nuclides
which were released in measurable quantities,
together with their fractional abundances and
the approximate fraction of ground level expo-
sure rate at 2 km from the stack, are given in
table 2. This information indicates that at
2 km, approximately 90 percent of the exposure
rate was due to the three nuclides krypton-87,
krypton-88, and xenon-135. The fractional ex-
posures were estimated assuming D stability
and 4 m/s wind speed.
7

-------
Table 2
Nuclide
8 smKr
8 7Kr
8 0Kr
1 3 3Xe
1 3 5Xe
13 8Xe
Calculated
relative noble gas releases
Calculated fraction of
Fractional abundance	ground level exposure-
based on total activity rate at 2 km
.079	.02
.200	.27
.210	.56
.147	.01
.289	.10
.075	.04
D. Model Description
A brief description of the four calcula-
tional models considered in this study is
given in table 3 and discussed below:
1.	AIREM (11) is the atmospheric dis-
persion model developed by the Office of
Radiation Programs for long-term exposure pre-
dictions. This program uses an "exposure inte-
gral" data file obtained from execution of the
program EGAD (12) to evaluate the exposure rates
from the external gamma emitters. The EGAD cal-
culation involves numerical integration over the
cloud geometry for each respective point of
interest on the ground.
2.	The calculation using AIREM.SI uses the
same structure as the model above but the expo-
sure rate estimate is made using the traditional
sector-averaged Gaussian calculation (13) of
ground-level concentration at receptor points
and conversion to exposure rate by an exposure
conversion factor (14) using average gamma ener-
gies from Lederer, et al (15).
8

-------
Table 3
Differences among exposure models
AI REM
Sector
Averaged
Plume
Dispersion
Output summed
over:
Nuclide
Stability
Class
Multiple Wind
Field Input
Variable lid
height with
stability
class
az evaluation Internal
Core	182K
Requi rements
Relative	$2
cost to run
Yes
Yes
Yes
No
AIREM.SI
Sector
Averaged
Yes
Yes
Yes
No
Internal
182K
$1
RRR
Sector
Averaged
No
Yes
Yes
Yes
Input
Data
140K
$5
ACRA
Single
Plume
Yes
NO
No
Yes
Input
Data
314K
$16
9

-------
3.	The RRR Model was written by Reeves,
et al (16) and has been used by the U. S.
Atomic Energy Commission in the preparation of
environmental impact statements (17,18) for
nuclear power plants. The calculational pro-
cedure employed is the sector-averaged Gaussian
estimation of ground-level concentration simi-
lar to that employed in AIREM.SI. Values for
vertical standard deviations are entered as
data rather than calculated internally as in
AIREM.SI. In addition, output is left in con-
centrations by each nuclide without direct con-
version to exposure rate. Such minor differ-
ences in input data requirements and output
format variations should lead to inconsequen-
tial differences in the results or applica-
bility of the two models.
4.	The ACRA model was written by Stallmann
and Kam (19) for reactor accident analysis. The
differences between calculational procedures em-
ployed in this model and those previously men-
tioned are quite significant. Three dimensional
Gaussian diffusion is assumed and exposures are
determined by numerical integration over all
points in the cloud which contribute signifi-
cantly to the respective receptor points. Sin-
gle runs consist of data for a given release
function, a single stability category, wind
speed, and wind direction. Consequently, the
results are intended for analysis of relatively
short-term releases when dispersion character-
istics are relatively constant.
V. Results
A. Field Measurements
The results of the field measurements are
presented in tables 4 and 5. The PIC results
were first tabulated on 2-hour intervals and
subsequently summed for the intervals shown.
The time periods shown in these tables corre-
spond to the TLD readout intervals. The data
collection was incomplete during some of the
time periods due to power failures and instru-
ment malfunctions.
10

-------
Table 4
Total exposure
(mR)
Site A	Site B
Time Period	PIC	TLD1	PIC	TLD1
08/29-10/02/73	10.45	9.90	8.02	7.95
10/02-11/06/73	14.54	14.03	NC	9.71
11/06/73-01/07/74	26.52	24.55	16.70	15.92
01/07-01/27/74	12.10	11.66	6.45	6.54
01/27-03/18/74	17.60	15.95	12.16	12.25
03/18-05/13/74*	11.27	11.64	10.89	10.59
Site C	Site D
Time Period	PIC	TLD1	PIC	TLD1
08/29-10/02/73	7.57	7.45	NC	7.97
10/02-11/06/73	8.32	8.64	8.40	8.91
11/06/73-01/07/74	15.07	13.96	NC	14.47
01/07-01/27/74	NC	4.97	5.51	5.13
01/27-03/18/74	NC	10.90	11.54	11.21
03/18-05/13/74*	NC	11.40	NC	11.20
NC Data collection not complete.
* Reactor shut-down for refueling.
1 These values represent the mean of three dosimeters at each site.
All standard errors of the mean were less than 2%.
11

-------
Table 5
Net exposure
(mR)
Site A	Site B
Time Period	TLD1	PIC	TLD1	PIC
08/29-10/02/73	2.61	3.15	1.32	0.91
10/02-11/06/73	6.66	6.39	3.01	NC
11/06/73-01/07/74	11.43	12.99	4.01	NC
01/07-01/27/74	7.47	8.02	2.73	2.75
01/27-03/18/74	5.29	8.51	2.55	3.20
Site C	Site D
Time Period	TLD1	PIC	TLD1	PIC
08/29-10/02/73	0.32	0.81	1.13	NC
10/02-11/06/73	1.44	1.33	1.69	1.33
11/06/73-01/07/74	1.13	3.20	1.90	NC
01/07-01/27/74	0.84	NC	1.09	1.66
01/27-03/18/74	0.47	NC	0.94	NC
NC Data collection not complete for this site.
1 These values represent the mean of three dosimeters at each site.
All standard errors of the mean were less than 2%.
12

-------
Table 4 gives the total exposure, natural
background plus facility contribution, as mea-
sured by both the PIC's and the TLD's. A sta-
tistical comparison of these values is given
in figure 4: a linear regression analysis
using the form y = bx+a was performed with the
PIC measurement as the y values and the TLD
measurements as the x values. The results of
this analysis yielded a regression equation of
y = 1.16 (x) - 0.93 with a correlation coeffi-
cient of 0.996.
The natural background measurements were
subtracted from the total exposure measurements
to yield the net exposures shown in table 5.
Shown in figure 5 is the comparison of net
TLD to net PIC measurements for the period of
the study. Regression analysis of these data
resulted in a regression equation of y = 1.089
x +0.48 with a correlation coefficient of 0.962.
PIC TOTAL EXPOSURE (mR)
Figure 4. Comparison of TLD & PIC (total exposure)
13

-------
(Z
PIC NET EXPOSURE (mR)
Figure 5. Comparison of TLD & PIC (net exposure)
B. Model Validation
Each of the four computer models was
used to obtain predictions of monthly gamma
exposures which were compared to their respec-
tive observed exposure as measured by the
PIC's.
The model validation and comparison por-
tion of the study applied the data from the
4 months September - December 197 3. This por-
tion of the study was not extended beyond 4
months due to the manpower and computer time
required to set up data files and execute each
of the four calculational model studies. The
adequacy of this length of time for estimat-
ing plume exposure has been shown in a pre-
vious report (20) , which indicated the predicted/
observed ratios for 1 month of data are not sig-
nificantly different from those for 2- and 4-
month periods.
14

-------
The observed net exposures at each of
the four sites for the 4 months, September -
December 1973, are shown in table 6 to range
from 5.88 mR at Site A in November to 0.32
mR observed at Site D in September. Corre-
sponding values of monthly predicted exposure
for each of the four computer codes are also
shown.
Table 7 gives the ratios of monthly pre-
dicted exposure to the corresponding observed
exposure for each of the four models. The
mean value of predicted/observed ratios range
from a low of 1.25 (a = 0.518) for the AIREM
model to a high of 2.402 (a = 1.18) for the
ACRA model. To formalize the comparison of
the performance of each of the models, several
statistical tests were run on the predicted/
measured ratios obtained. These tests were
divided assuming the predicted/measured ratios
to be a random variable. The tests were di-
vided into two groups; the first set of tests
assumed that the variable was distributed nor-
mally and the second set assumed that the vari-
able was distributed log-normally. Results of
both sets of tests are given below.
Normal Distribution Treatment: To test the
assumption of homogeneity of the four within-
group variances, a Bartlett Test (21) was per-
formed. The null hypothesis of equal variances
was rejected at the 99 percent level and, there-
fore, the one-way analysis of variance could not
be used as an exact test of the results from the
four models. Consequently a pooled-variance
range test (22) was applied to the results.
This test showed that the results from AIREM,
AIREM.SI and RRR were not significantly differ-
ent at the 99 percent level, but the results
from ACRA were significantly different. This
result is not at all surprising after a cursory
examination of the results in table 7 and one
recalls that the single plume calculation in-
herently gives higher predictions for elevated
releases than does the sector-averaged
calculation.
15

-------
Table 6
Predicted and Measured Exposure Values (mR)*
Site A	Sice B

Mea-

Predicted

Mea-

Predicted

Mo/Yr
sured
AI REM
AR-SI RRR
ACRA
sured
A1REM
AR-SI RRR
ACRA
9/73
2.49
2.04
0.74 0.92
3.71
0.78
0.90
0.80 0.95
2.44
10/73
3.02
3.08
3.30 3.32
4.68
1 .02
1.08
1.25 1.25
2.77
11/73
5.87
5.24
3.43 3.80
7 .02
2.48
2. 15
2.44 2.30
3.94
12/73
5.70
4.62
5.48 6.95
5.12
1.66
1.61
1.94 2.39
2.81
Total
17.08



5.94





Site
C


Site
D


Mea-

Predicted

Mea-

Predicted

Mo/Yr
sured
AI REM
AR-SI RRR
ACRA
sured
AIREM
AR-SI RRR
ACRA •
9/73
0.75
0.58
0.58 0.52
1.15
0.32
0.60
0.55 0.53
1.53
10/73
0.52
0.74
0.89 0.81
1.65
1.13
0.81
0.76 0.88
1.86
11/73
1.52
3.42
4.11 3.28
6.43
0.80
1.69
1.78 1.78
2.88
12/73
2.20
2.99
3.36 3.04
4.87
1.53
2.54
2.45 2.40
4.08
Total
4.89



3.78



*Values
given are
above
background at
the respective
sites.


16

-------
Table 7
Predicted/measured exposure ratios for four models
Mo/Yr Site	AIREM	AIREM.SI	RRR	ACRA
9/73 A	0.82	0.30	0.37	1.49
B	1.16	1.03	1.22	3.14
C	0.77	0.77	0.69	1.53
D	1.88	1.72	1.66	4.78
10/73 A	1.02	1.09	1.10	1.55
B	1.06	1.23	1.22	2.72
C	1.42	1.71	1.56	3.17
D	0.72	0.67	0.60	1.65
11/73 A	0.89	0.58	0.65	1.20
B	0.87	0.98	0.93	1.59
C	2.40	2.89	2.30	4.53
D	2.12	2.29	2.24	3.62
12/73 A	0.81	0.96	1.22	0.90
B	0.97	1.17	1.44	1.69
C	1.36	1.53	1.38	2.21
D	1.66	1.60	1.57	2.67
Mean	1.25	1.28	1.26	2.40
Range 0.72-2.40 0.30-2.89 0.37-2.30 0.90-4.78
Standard
deviation	0.52	0.66	0.55	1.18
Standard
error of
the mean	.13	.16	.14	.30
17

-------
Perhaps the most important outcome of
these tests is that the three other models
AIREM, AIREM.SI. and RRR produced results
which were not significantly different as
measured at these four measurement sites.
Log Normal Distribution Treatment: The ratios
in table 7 were transformed with the expres-
sion z = In (r). Analysis of variance was then
performed using the model
Zijk = ^o + ai + 6j + ^k +	+ eijk where
a, 3/ and y correspond to method, site, and
month effects respectively. The only interac-
tion considered was (a, $)ij which represents
the interaction between method and site - a pos-
sibility where AIREM and ACRA consider the ex-
pected plume distribution above the receptor
while the other models assume a uniform concen-
tration equal to the ground level concentration.
A 0.05 level of significance was chosen to
define the critical region for the test sta-
tistics. The interaction (a,3) has F = .232(9,45)
so rejection of the null hypothesis is not sup-
ported by the data. Similarly F = 1.568(3,45)
for the monthly effect and so it too is deemed
insignificant. The method and site effects a
and 6 provide test statistics of F = 9.15(3,45),
p < .0002 and F 9.19(3,45), p < .0002 respec-
tively and so both effects are considered signif-
icant. At this point the analysis of variance
was repeated removing the interaction from the
model. The tests for a, 3, Y were F = 10.49(3,54)
(p < .0001), F = 10.54(3,54) (p < .0001), and
F = 1.80(3,54) (p < .15). Once again the method
and site effects are significant but the monthly
effect is not statistically significant by this
three factor test.
The underlying assumptions of homogeneity
of variance and additivity were investigated for
a two-way classification of the data, the monthly
observation providing four replications for each
call. Bartlett's test (23) was used to evaluate
the homogeneity of variance. The test statistic
indicated that the hypothesis of homogeneity
18

-------
should be accepted. Tukey's test (24) for
additivity was then performed. The result-
ing F = .075(1,53) indicates that the effects
can be considered additive under the log trans-
formation. Arranging the means in order:
Method AIREM.SI RRR AIREM ACRA
Z	.1164 .274 .1477 .7665
The difference between models can be studied.
Using a Q test (24), the critical difference
between mean D is given by:
D = Q 05 s/ VS = 3.76 • 3939/ VT6 = .369
(a = 4, f = 54)
where a is the number of methods (4) and f is
the degrees of freedom for s (54) , and n is the
number of observations for the mean (16).
It is apparent that the differences be-
tween the means of the first three methods are
insignificant but that the fourth model ACRA
has a mean significantly greater than any of
the other methods.
In summary:
(a)	The log transformation is consistent
with the need for additivity and homo-
geneity.
(b)	Any interaction between site and
method is not supported by the study.
(c)	The monthly effect is negligible.
(d)	Site effects are not negligible. The
site variance is comparable to the
residual variance.
(e)	The mean for ACRA is significantly
greater than for the other methods
which do not differ significantly.
19

-------
C. Measured and Predicted Gaseous Concentrations
In addition to the direct exposure mea-
surements, two field measurements of gaseous
133Xe concentrations have been examined with
regard to relating these to predicted values.
These samples were collected by evacuating a
34-liter tank and subsequently filling to
atmospheric pressure after locating the plume
centerline using mobile PIC field measurements.
After collection, the samples were re-
turned to the laboratory for analysis using a
noble gas separation apparatus and counting the
concentrated samples in a liquid scintillation
counter. The hemispherical immersion approxi-
mation was used to predict exposure rates from
the measured concentrations. The procedure in-
volved converting the measured concentration of
noble gas nuclides using the noble gas ratio
nuclide mix as stated in the July - December
1973 operating report (10). Next the total ex-
posure rate was calculated using the exposure
rate conversion factors for each of the noble
gas nuclides. For the two plume centerline con-
centrations of 0.032 and 0.036 yCi/m3 the expo-
sure rates of 96.4 and 108.4 yR/hr respectively
were obtained. The 95% confidence limits ob-
tained from those values is 102 ± 108 pR/hr.
This value compares reasonably well with the
measured exposure rate of 150 yR/hr at plume
centerline using a PIC during the interval in
which the air samples were taken (see table 8).
The field measurements show that ground-
level concentrations and predictions are rea-
sonably consistent considering that only two
measurements were made. This indicated that
both the dispersion calculation and the conver-
sion from concentration to exposure rate by the
computer models appear to agree within a factor
of two. A larger number of field gas concen-
trations in subsequent studies could place a
better confidence interval on agreement of mea-
sured and predicted values.
20

-------
Table 8
Measured and predicted values based on field gaseous
xenon concentration (January 28, 1974)
Measured total noble	AIREM calculated	AIREM calculated
gas release rate Ground-level concentration Exposure rates
115,500 yCi/sec(a)	0.069 yCi/m3	212 uR/hr(b)
Measured ground-level
concentration of 133Xe
0.032 yCi/m3
0.036 yCi/m3
Exposure rate calcu-
lated from ground-level
concentration
96.4 yR/hr
108.4 yR/hr
95% confidence
on exposure rate
calculated from
measured 13 3Xe
concentration
102 ± 108 uR/hr
Exposure rate measured with PIC during the air sampling: 150 yR/hr
Stability class (calculated from variance of crosswind direction): B
Estimated fraction due to 133Xe: 0.133
Distance from point of collection to stack: 2.5 km
Effective stack height: 100 m
Wind speed: 1.8 m/s
(^Estimated total noble gas mixture.
(b)Calculated from July - December 1973 noble gas mixture.
21

-------
Discussion
The three sector-averaged computer models were
quite close in predicting the exposures observed
over the 4-month interval at each of our measuring
sites.
For the case where the data were treated as
normally distributed, the AIREM model has the lowest
mean predicted/measured exposure ratio of 1.25. When
the data were treated as coming from a log-normal
distribution the calculated mean for the AIREM.SI was
closest to unity. For both of the above cases the
only model producing results significantly different
was ACRA. This single plume accident model was found
to be overly conservative for both cases when esti-
mating monthly exposures.
These results are consistent with those obtained
by Martin (25) in his application of AIREM to 1 year
of field data and by Gogolak (1) in his study of
several atmospheric codes applied to 2 months of data.
Since the range of measurement sites used in
this study varied from 1.4 to 3.3 km, care should be
exercised in extrapolation of the results of the
inter-code comparison to a different range of dis-
tances .
Based on the observed correspondence between
predicted and measured exposures, the use of vari-
ance of horizontal wind direction for estimatinq sta-
bility classes over the period of this study appears
quite satisfactory.
The excellent agreement between the gross TLD
and PIC measurements shown in table 4 and figure 4
demonstrates the ability of this type TLD to accu-
rately measure low total environmental radiation
exposure levels. However, the comparison of the net
exposures (table 5) as recorded by the PIC's and
TLD's shows significant differences. By examining
the background subtraction methods for both PIC's and
TLD's, the net exposure differences were determined
to be a result of an inability to accurately measure
and "subtract out" the natural background component
from the total or gross exposure in the TLD's. The
22

-------
background was measured by the PIC on a continuous
basis and the plant contribution was determined by
integrating only the peak or additional exposures.
However, the TLD background was determined with the
TLD's during only one period of plant shut-down,
3/18-5/13. Therefore, any seasonal or daily varia-
tions in the natural background due to rainfall,
snow, etc. were not accounted for in the determina-
tions of net exposures. Although variations in
background with time were shown not to be signifi-
cant in a three factor experiment using model, site,
and time, there is evidence from comparison of the
total and net results to suggest that the fixed back-
ground subtraction adds error into the net TLD esti-
mates .
The manual reduction of PIC strip chart data in
this study has apparently led to no serious errors
in evaluation of the net exposure observed from the
gaseous plume. However, a machine-oriented data re-
duction process as discussed by Gogolak and Miller
(26) to account for variations in natural background
is under study and may be useful in subsequent
studies.
VII. Summary and Conclusions
The results of the intercomparison study of
four widely used models have demonstrated the use-
fulness of each model for predicting ground-level
exposure rates.
For three models AIREM, AIREM.SI, and RRR, the
mean predicted/measured exposure ratios for external
exposure were 1.25, 1.29, and 1.26, respectively, and
standard errors of the means were less than 0.3. The
use of the short-term accident code ACRA did result
in markedly increased error. In summary, based on
the assumption of normally distributed data, the model
AIREM demonstrated the predicted/measured exposure
ratio closest to unity. Use of this or other similar
models for predicting external exposures has been
shown to be quite acceptable for most applications.
Further studies in this area should emphasize extend-
ing the analysis to a wider range of distances from
the release point. Furthermore, analysis of particu-
lates from a source having a higher release rate should
be included.
23

-------
The usefulness of PIC's for making low-level
exposure measurements was demonstrated by this
study. The ability of this instrument to accu-
rately measure exposure rates of a few yR/hr above
natural background was clearly evident.
The results of this study exhibit the ability
of this type TLD to accurately measure total environ-
mental exposures in the range of a few mR per month.
However, the ability to accurately measure an in-
crease of 5 mR per year above natural background is
questionable. This is due to difficulty in deter-
mining the natural background portion of the total
exposure. The method used to account for natural
background in this study is not as desirable or
accurate as an extended (1 to 2 years) pre-
operational survey. Unless seasonal and other vari-
ations in natural background can be accurately de-
termined, it will be impossible to measure facility
contributions in the range of 5-10 mR/year with this
type TLD. Due to the low cost involved the TLD's
might serve as an integral portion of an external
radiation monitoring system including both TLD's
and PIC's.
24

-------
REFERENCES
1.	GOGOLAK, C. V. Comparison of Measured and Calculated
Radiation Exposure from a Boiling Water Reactor Plume,
HASL-277 (1973).
2.	NORTHERN STATES POWER COMPANY. Final Safety Analysis
Report, Volume 1, Docket Number 50-262 (October 1968).
3.	NORTHERN STATES POWER COMPANY. Gaseous Radwaste
System Modification Report, USAEC Docket 50-26 3,
Minneapolis, Minnesota (October 1971).
4.	DE CAMPO, J. A., H. L. BECK, and P. D. RAFT. High
Pressure Argon Ionization Chamber Systems for Measure-
ment of Environmental Radiation Exposure Rates, HASL-
260 (1972).
5.	WINDHAM, S. T., J. E. PARTRIDGE, and H. L. KELLEY. An
Integrating Device for Use with Pressurized Ionization
Chambers, EPA 520/5-73-007 (December 1973).
6.	PARTRIDGE, J. E., S. T. WINDHAM, J. L. LOBDELL, and
J. A. OPPOLD. Suitability of Glass-Encapsulated
Calcium Fluoride:Manganese TLD's for Environmental
Radiation Surveillance, EPA, ORP/EERF 73-1 (June 1973).
7.	PARTRIDGE, J. E., J. A. BROADWAY, S. T. WINDHAM, and
C. R. PHILLIPS. Low Level Population Exposure Measure-
ments Using Thermoluminescent Dosimeters, CONF. 741018
Health Physics Society, 8th Midyear Symposium (October
1974) .
8.	MARKEE, E. H. On the Relationship of Range to Standard
Deviation of Wind Fluctuation, Monthly Weather Review
91(2), 83 (1963).
9.	TURNER, D. B. Workbook of Atmospheric Dispersion
Estimates, Office of Air Programs, OAP-OTIP-AP-2 6 (1970).
10. NORTHERN STATES POWER COMPANY. Report to the United
States Atomic Energy Commission, Directorate of Licens-
ing, License No. DPR-22, Six Month Operating Report No.
6, July 1, 1973, to December 31, 1973, Section V. A.,
Minneapolis, Minnesota (1973).
25

-------
11.	MARTIN, J. A., C. B. NELSON, and P. A. CUNY. AIREM
Program Manual - A Computer Code for Calculating
Doses and Depositions due to Emissions of Radionuclides.
EPA 520/1-74-004 Office of Radiation Programs, U. S.
¦Environmental Protection Agency, Washington, DC 20460.
12.	COOPER, R. E. EGAD - A Computer Program to Compute Dose
Integrals from External Gamma Emitters, DP-1304,
Savannah River Laboratory, Aiken, South Carolina (1971).
13.	GIFFORD, F. A., JR. An Outline of the Theories of Dif-
fusion in the Lower Layer of the Atmosphere, Meteorology
and Atomic Energy, p. 113, D. H. Slade, Ed., TID-24190,
U. S. Atomic Energy Commission, Oak Ridge, Tennessee
(1968).
14.	COWSER, K. E., et al. Dose Estimation Studies Related
to Proposed Construction of an Atlantic Pacific Inter-
oceanic Canal with Nuclear Explosives, ORNL-4101 (196>7) .
15.	LEDERER, C. M., et al. Table of Isotopes, 6th Ed.
John Wiley, New York (1967).
16.	REEVES, N., et al. A Computer Code for Analyzing Routine
Atmospheric Releases of Short-Lived Radioactive Nuclides,
ORNL-TM-3613 (1972).
17.	U. S. ATOMIC ENERGY COMMISSION. Draft Environmental
Statement Related to the Arkansas Nuclear One Unit 1,
Docket No. 50-313 (1972).
18.	U. S. ATOMIC ENERGY COMMISSION. Draft Environmental
Statement Related to the Arkansas Nuclear-One Unit 2,
Docket No. 50-368 (1972).
19.	STALLMANN, F. and F. B. K. KAM. ACRA - A Computer
Program for the Estimation of Radiation Doses Caused
by a Hypothetical Reactor Accident, ORNL-TM-4082 (1973).
20.	BROADWAY, J. A., J. E. PARTRIDGE, S. T. WINDHAM, and
C. R. PHILLIPS. An Examination of Design Parameters
in a Field Study for Exposure Evaluation, CONF. 741018
Health Physics Society, 8th Midyear Symposium (October
1974).
21.	OSTLE, B. Statistics in Research, 2nd Ed. Iowa State
Press, (1963), p. 136.
26

-------
22. NATRELLA, M. G. Experimental Statistics, N. B. S.
Handbook 91, U. S. Department of Commerce (1963) p. 3-40.
23.	SERVICE, J. A. User Guide to the Statistical Analysis
System, Student Supply Stores, North Carolina State
University, .Raleigh, NC (1972).
24.	SNEDECOR, G. W. and W. G. COCHRAN, Statistical Methods,
Sixth Edition, Iowa State University Press, (1967).
25.	MARTIN, J. A. A Comparison of Calculated and Measured
Yearly Doses in the Vicinity of the Humboldt Bay Nuclear
Power Plant, to be published (197 3).
26.	GOGOLAK, C. V. and K. M. MILLER. Method for Obtaining
Radiation Exposure due to a Boiling Water Reactor Plume
from Continuously Monitoring Ionization Chambers, Health
Phys. 27, 132 (1974).
27

-------

£
5
D
O
\

ra UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
2
lii
O

-------