United States
Environmental Protection
Agency
Air and Energy Environmental
Research Laboratory
Research Triangle Park NC 27711
January 1990
Research and Development
The 1990 International
Symposium on Radon
and Radon Reduction
Technology:
Volume II. Preprints
Session III: Radon Measurement
Methods
Session III-P1: Panel Session on
QA/QC of Measurement
Session III-P2: Panel Session on
Short-/Long-term Radon
Measurements
Session B-lll: QA/QC of
Measurement—POSTERS
Session B-IV: Short-/Long-term
Radon Measurements—
POSTERS
February 19-23,1990
Stouffer Waverly Hotel
Atlanta, Georgia
-------
Session III:
Radon Measurement Methods
-------
Ill - 1
Time Series Linear Regression of Half-Hourly
Radon Levels in a Residence
David A. Hull
Center for Energy and Environmental Studies
Princeton University
Princeton, NJ 08544-5263
ABSTRACT
This paper uses time series linear regression modelling to assess the
impact of temperature and pressure differences on the radon measured in the
basement and in the basement drain of a research house in the Princeton area
of New Jersey. The models examine half-hour averages of several climate and
house parameters for several periods of up to 11 days. The drain radon
concentrations follow a strong diurnal pattern that shifts 12 hours in phase
between the summer and the fall seasons. This shift can be linked both to
the change in temperature differences between seasons and to an experiment
which involved sealing the connection between the drain and the basement.
Ve have found that both the basement and the drain radon concentrations are
correlated to basement-outdoor and soil-outdoor temperature differences (the
coefficient of determination varies between 0.6 and 0.8). The statistical
models for the summer periods clearly describe a physical system where the
basement drain pumps radon in during the night and sucks radon out during
the day.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.^
1. INTRODUCTION
Regression modelling of half-hourly averaged radon data is generally
difficult for several major reasons. Radon levels are dependent on a large
number of environmental and house-specific factors, many of which are not
measured on a continuous basis. Despite the fact that important parameters
like temperature and pressure are regularly measured, they often vary widely
from location to location even within or beneath the same structure. In
addition, there is often a time lag between changes in forcing parameters
and the related impact on radon levels. This lag is both natural and a
result of instrumentation and measurement techniques. Unfortunately, even
*This work was funded by the U.S. Environmental Protection Agency under
Cooperative Agreement No. CR-814673.
-------
the length of the lag often varies depending on the season or even with the
time of day. Finally, the magnitude of the influence of the various
parameters on radon concentration does not remain constant from day to day.
While these variations may not visually cloud the patterns in the data, they
create very poor regression models. This highlights the danger of "blind"
regression modelling. Apparent strong patterns in the data may often lead
to poor regression fits for the reasons mentioned aboye. In fact, a very
poor fit over a small fraction of the data can appear to invalidate the
entire model.
2. APPROACH, HOUSE DESCRIPTION, AND LIST OF MEASURED PARAMETERS
Despite the problems mentioned above, it is still possible to pick out
periods during which the influence of certain parameters is particularly
obvious. These periods would not only be easier to model, but the
appropriate physical processes would be equally more obvious. In
particular, during the summer and fall of 1988, the basement and subslab
radon levels in Research House 22 (H22) exhibit strong diurnal patterns that
seem to be linked to changes in temperature and pressure.
H22 is an old balloon-construction three story home with a full
basement. It has radiative heating and zoned cooling, with the particular
advantage that the air-conditioning unit is on the third floor, thereby
minimizing its impact on basement and subslab radon levels. The basement
has three drains: two communicate in the main section (east and west) which
have identical radon levels, and a third in the crawlspace that is probably
blocked. The drain radon level referred to in the rest of this paper is
measured in the west drain. The data are divided into nine periods that
range from 7 to 11 days in length and are spread out over the year between
late May and early October. The period length is a compromise between using
short periods which produce better fits and long periods which provide a
much more comprehensive model. The parameters that are used in this study
are: basement radon, subslab radon (in this case taken to be the radon level
in a basement drain), outdoor temperature, basement temperature, soil
temperature, and the pressure difference between the basement and the
outdoors, all measured continuously and recorded as half-hour averages. The
pressure difference between basement and subslab was also measured, but its
total magnitude was inside the range of instrument error and so it was
excluded. A complete list of the parameters and their abbreviations is
given in Table 1. The goal of this analysis will be to explain the half-
hourly patterns in the radon variables using linear regression models of
temperature and pressure differences.
3. A FIRST LOOK AT DATA
The radon levels in the basement average 20-50 pCi/L, while the average
in the drain ranges between 3000 and 5000 pCi/L. The extraordinarily high
levels in the drain led to th& assumption that this provided the primary
means of radon entry into the basement. So, on Julian day 211, the east and
west drains were blocked to test the amount of communication between the
basement and the drain. This significantly reduced the radon in the
basement for about 10 days. Subsequently, the radon concentration climbed
back up to its original level. This experiment reinforced the hypothesis
that blocking the path of least resistance does not eliminate the radon
-------
problem, it only causes the radon to find a different path of flow.
Although the average radon concentration remained the same, the diurnal
patterns were significantly altered. Unfortunately, this also corresponded
to the change in seasons from a hot summer to a cooler autumn, so it is
difficult to distinguish between the two effects. Thus, the radon data that
will be studied have two distinct patterns: summer with the drains open and
fall with the drains blocked. A brief description of the data periods is
given in Table 2, while complete graphs of the different periods are given
in Ref. [1],
A quick scan of the data reveals a strong diurnal pattern (Figs. 1-4).
During the summer, the basement and drain radon levels seem highly
correlated, peaking in the middle of the night and bottoming out during the
heat of the day. During Periods 4 and 5, the hottest part of the summer,
the levels in both the drain and the basement drop to almost zero during the
day. Some of the summer peaks also have an unusual dip right in the center
of the drain radon levels (Figs. 1 and 3). The fall patterns show a similar
diurnal cycle, but the drain radon now peaks during the hottest part of the
day, 12 hours out of phase with the basement radon (Figs. 2 and 3). The
drain radon continues to show the double peak structure which occurred
during the summer, but now the dips are even larger.
4. MODELLING APPROACH: STRUCTURE AND METHODOLOGY
The patterns described in Section 3 indicate that a model based on the
basement-outdoor temperature difference should produce a good fit to the
data. In the initial analysis, we will compare the explanatory powers of
two variables: the basement-outdoor temperature difference (T^-T0) and the
basement-outdoor pressure difference (dP^Q), using a time series regression
model. Previous time series radon models using half-hourly measurements
indicated very strong autocorrelations between the current radon level and
that lagged by half an hour [2]. Thus, the autoregressive terms tend to
overwhelm the natural driving parameters. Using first-difference models
eliminated the autocorrelation, but they also eliminated most of the
variation in the dependent variable and the resulting models were relatively
poor (the Coefficient of Determination, R3, <0.5). Consequently, in the
framework of the present study, ve decided to use simple moving average
models. The descriptive power of the independent variables will be
evaluated by including lags of 0, 1, 2, 3, and 4 hours. Though the data are
accurate to the half-hourly level, hourly lags will be employed in order to
keep the models simple (i.e., to hold the number of parameters to a
minimum). All coefficients that are not individually significant by a
t-test (based on a p-value less than 0.05) [3] will be excluded and the
overall quality of the model will be assessed using the R^ statistic. The
choice of 4 hour* as the maximum lag vas based on trial and error modelling
of the data which Indicated that longer lags did not significantly improve
the models. In addition, the models were adjusted to that all the
parameters of each variable had the same sign. For example, if basement
radon is positively correlated with dPw0 at a lag of 1 hour, it should not
be negatively correlated at a lag of 2 nours. While not necessary in an
ordinary time series model, this sign correction is important here to
maintain the physical significance of the parameters.
-------
The use of lag terms In a moving-average model does not actually mean
that the current radon levels are directly dependent on the temperature
differences 2 or 3 hours before. It is merely an abstract method of
describing the measured radon as consisting of a mixture of old and new air,
where the air infiltration is directly dependent on temperature and pressure
differences. Thus, a 2-hour temperature lag term reflects the amount of
radon entrained by the air that escaped from the soil 2 hours before which
is still present at the measurement location.
5. DRAIN RADON MODELS
5.1 Summer Period
The statistical analysis was initially conducted on the data from
Period 1 (148-155 Julian), using the drain radon as the independent
variable. The models for this period are given in Table 3. The basic
temperature and pressure difference models (3a and 3b) show very poor fits
(R2 of 0.366 and 0.381, respectively), which is rather surprising and
contrary to what one would expect just by looking at the data. However, it
was unusually cool during the early part of the sample and it rained quite
heavily on the last day of the period. If these periods are excluded, both
the temperature and the pressure models (3c and 3d) Improve tremendously
(R* of 0.816 and 0.669). From these results, it appears that the
temperature difference works at least as well as the pressure difference to
describe the drain radon variation. In addition, the high value suggests
that the temperature difference represents the primary driving force for
radon entry into the drain, which means that the drain must have some
connection to outdoor air. (As yet, an actual exit for the drain has not
been discovered.)
The small dips in the drain radon concentration that are noticeable in
the peaks on days 148-150 can be explained by further refining the model. It
was noticed that they correspond to the times when the outdoor temperature
drops below the soil temperature. Therefore, the model has been expanded to
include the term [Ts-T0] , where '+' implies that the negative values of
(Ts-T0) have been set to zero. The new model (3e) provides only a slight
improvement on the simple temperature difference model (R^-0.853); however,
when it is reapplied to the entire week (3f), it represents a substantial
improvement over the original model, indicating that the poor fits (i.e.,
low R' values) were a result of an inadequate model rather than the
environmental conditions.
Before proceeding, it is important to find a physical description of
the system that explains the results of the statistical model. Vithout such
backing, the statistical models have very little value. The models imply
that there is a three level system consisting of the soil, a pocket of
outdoor air that enters through the drain, and the basement air above the
drain. In order to maximize the stack effect, and thus the radon flow,
there oust be a positive temperature gradient through the entire system;
i.e., Tg < Tq < Tfc. Since the model describes a summer period, the soil
temperature is almost always cooler than the outdoor temperature. This
means that the dominant effect in the model is the basement-outdoor
temperature difference. For this reason, ve ignore the times when Tt is
-------
less than TQ and restrict the model to the term [Ts-T0]+. This corresponds
to the times when the soil-outdoor temperature difference acts against the
main driving force [T^-Tq]. Also, from a statistical standpoint, since both
Tjj and Ts are nearly constant over the entire week, the differences [T^-T0]
and [Ts-T0] would be almost collinear; thus, they should not both be used in
the same model. It might seem that, if the soil temperature were less than
the outdoor temperature, the radon flow into the drain from the soil should
be cut off completely. This may well be the case, but the duration of this
period is short enough that this doesn't substantially affect the overall
concentration in the drain, and thus we do not see this pattern in the data.
The Continuous Radon Monitor (CRM) measuring the drain radon takes
samples from the air a short distance below the basement floor. When the
variable [Tb-T0] is positive, radon levels will tend to increase because air
with a very high radon concentration is being pumped upward from deeper in
the drain. Similarly, when [Tjj-T0] is negative, radon levels will tend to
decrease because cleaner basement air is being sucked down the drain.
Unfortunately, because of instrumentation problems, we have no pressure
measurement which is equivalent to [T8-T0], so it is impossible to construct
a more comprehensive pressure model. Since the current statistical model
does have a strong physical justification, we will extend it to the rest of
the summer data.
As we mentioned previously, during the period which includes the
hottest part of the summer (Julian days 189-204), the radon levels in the
drain drop off to almost zero when [T^-T0] is negative. This means that a
radon model using [T^-T^]* would provide a better fit to the data. In
addition, since [Tg-T0]* is never significant in this period, it can be left
out of the model completely. The temperature model and a similar pressure
model for this period are given in Table 4. (In this and in any future
tables, the parentheses give the value for the model with no intercept.
However, since intercept and no-intercept values are not equivalent, the
intercept values are provided first to allow comparisons between models.) We
note that they provide a very strong fit to the data. A likely physical
explanation is that the temperature rises so quickly that, as soon as the
outdoors becomes hotter than the basement, a strong airflow out through the
drain drops the radon levels to zero.
We can combine the two temperature models given above (3f,4a) to
produce a comprehensive model structure for describing the radon
concentrations in the drain by dividing [T^-T0] into two terms:
Rndr - *1[Tb-T0]i+ + bitVTbli* + c1[T,-T0]1+ + k; i - 0 4 (1)
where i represents the number of hours of the lag and k is the intercept.
Splitting up the basement-outdoor temperature difference makes physical
sense, since the change in radon concentration from that in the air pumped
from deeper in the drain will be different from the air pumped in from the
basement. The fits for each summer period uaing this model are presented in
Table 5. While the coefficients and the significant parameters vary widely
from period to period, we note that, overall, the model structure provides a
relatively accurate description of the radon concentrations In the drain.
The only exception Is model 5c, which fares very poorly. A likely reason
-------
for this is chat the temperatures are much cooler than is typical for the
summer, and so these data do not follow the patterns of the summer model.
However, if we examine only the first part of the period (days 175-179), the
model does fit (R^ - 0.676), although not as well as for the rest of the
data [see Table 6(a)]. The reason that the first part of the data fits the
model while the second part does not may be that it takes several days of
cooler weather for the summer model to deteriorate.
The adequacy of the summer model is further evaluated by comparing
measured and modeled radon levels (see Fig. 1). For the most part, the
models appear to fit very well, showing no consistent bias. A few
exceptions, like day 148 of Period 1, can be excluded and the results of the
models improve substantially [see Table 6(b,c)]. The value of this approach
is that it allows one to see if a model is uniformly poor or whether it does
not fit only a small part of the data. It is clear that the unusually low
peak on day 194 results in a systematic under-fitting of all the rest of the
days in that period. Another observation is that the modelled radon values
drop below zero during Period 2. These can be rounded up to zero to produce
an even better fit for those data. Since there is no systematic under-
fitting or over-fitting of the peaks during any period, other than the
exception pointed out above, a linear model should be sufficient to describe
the data. It is possible that higher order models would provide a good fit
for the cooler periods as well (perhaps the process is only locally linear);
this represents a good direction for future work.
5.2 Fall Period
While the summer period was dominated by the basement-outdoor
temperature difference, the fall period, with the drain blocked to limit
contact with the basement, produces a dramatically different diurnal pattern
of radon flow. Now, the drain radon levels will be dominated by the
difference between the outdoor and the soil temperature. When the outdoor
air (i.e., the air in the drain) is warmer than the soil, then the radon
will tend to flow into the drain. This corresponds to the peaks that appear
in the actual data. These peaks are 12 hours out of phase with the ones
that occurred during the summer, appearing during the hottest part of the
day. Now, the dip in the center of the peak corresponds to the time when the
outdoor temperature rises above that of the basement, indicating that,
despite the sealing, the basement air still manages to find a path into the
drain. Thus, these dips correspond to cleaner and colder basement air
sinking into the drain.
The model structure will be similar to the summer model, but now the
outdoor-soil temperature difference will dominate. The term [T^-T0]+ will
be excluded for similar reasons to those given for the summer model above:
Rndr - •ilT0-T,]1+ + b1[Ta-T0]1+ + c1[T0-Tb]1+ + k; i - 0 4 (2)
where i represents the number of hours of the lag and k is the intercept.
Each of the fall periods is fit to this model, and the results are presented
in Table 7. Obviously, this model does not fit these data nearly as veil as
the previous one. Despite the fact that Period 6 could not be fit to the
model, the same patterns that were mentioned above do appear there as well.
The magnitude of the coefficients do not seem to be consistent enough to
-------
produce a coherent model. This may mean that the Initial effects of
blocking the drain had not yet completely worn off. The comparison plots of
measured versus modelled data (see Fig. 2) also show that the fit, although
reasonable, is not as accurate as for the summer model.
As was mentioned previously, it is difficult to determine whether the
fall model structure is due entirely to the impact of the blocked drains or
whether the cooler weather makes most of the difference. In order to test
the relationship between the two factors, the fall model was run on some of
the cooler summer data (days 179-184 of Period 3, where the summer model
didn't fit) when the drains were open (see Table 8, » 0.596). The fit
was poorer than for the rest of the fall data, but it did indicate that
cooler weather rather than the blocked drains may explain much of the
difference in the patterns of radon flow.
5.3 Conclusions
Thus, the drain radon concentrations during the summer and fall in H22
can be adequately modelled (R* generally between 0.7 and 0.8) using the
temperatures in the basement, the outdoors, and the soil. Although, the
coefficients are quite different from week to week, two distinct model
structures emerge, one for the summer and one for the fall. Vhile sealing
the drain-basement connection serves to accentuate the difference between
the seasons, it is not completely clear whether it has a major effect on the
drain radon levels.
It should be mentioned that the current model structure is designed to
facilitate the physical interpretation of the parameters. Since T^ and Tg
are nearly constant over a 10 day measurement period, the terms [T^-Tc] and
[T#-T0] differ to first order by a constant and thus are equivalent in terms
of regression modelling. Since some of the information is redundant, it
would be misleading to fully use both terms in any model. In fact, both the
summer and the fall models of the drain radon are actually sub-models of the
following general model:
Rndr - •itTb-T0]£+ + b^To-Tbli*- + + *ilV*o]i'' + k (3)
where i ranges from 0 to 4 and represents the hours of lag of the parameter
and k is the intercept. In practice, the collinearity of [Tj,-T0] and
[Ts-T0] forces us to exclude one of these terms from aach model. The terms
chosen for each model are the ones that provide the best physical
description.
6. EASEMENT RADON MODELS
Now that the drain radon concentration is well described by temperature
and pressure models, ve can examine the basement radon. Inspection of Figs.
3 and 4 reveals that it is strongly related to basement-outdoor temperature
and pressure differences. It is not surprising that, during the summer
period with the drains opan, basement radon lavels are also highly
corralated with the drain radon levels, following the aame diurnal cycle.
The first set of models will measure the extant to which the variation in
basement radon can be explained by Rn^r, (dPw0], and [Tfc-T0] for Period 1
(148-155 Julian). The results ara preaented in Table 9. The models show
-------
that the drain radon and the pressure difference alone provide a reasonable
fit, with the temperature difference being much poorer. The best model
combines the drain radon with the pressure difference and results in an
excellent fit (R^ - 0.860). This combined model is necessary because the
basement radon does not show the dips associated with soil-outdoor
temperature difference, indicating that the drain is not the only means of
radon entry into the basement. Therefore, the basement-outdoor pressure
difference in the combined model measures radon entry into the basement from
sources other than the drain.
Given the success of the previous models, we will expand the analysis
to all of the summer periods using the following model structure:
Rnjj - a^IRn^j.]^ + b£[dPt>/0]i + k i - 0 4 (4)
where i is the time lag in hours and k represents the intercept. The
results are presented in Table 10, along with simple pressure difference
models for comparison. The reason that a simple pressure model is markedly
inferior to the model given above is that it ignores the soil-outdoor
parameter that is so important in describing the radon levels in the drain.
The full temperature difference model used to model drain radon could also
be applied here, but since this is represented by the drain radon values, it
is more convenient to deal with the simple model.
The basement model provides an excellent fit for almost every period.
The only period where the R^ value drops below 0.75 is when the drain radon
is very poorly modelled, during a stretch of cooler weather. The measured
versus modelled plots (see Fig. 3) also show the quality of the fit. In
fact, if just a few days are excluded (say, days 150, 166, 168, 188, and 198
Julian), it is likely that the R^ would exceed 0.95. It should be mentioned
that during many periods, particularly the later part of the summer when the
soil temperature has very little impact, the drain radon concentration and
the basement-outdoor pressure difference are highly correlated. In terms of
the models, this means that two factors are likely to be confounded so that
the actual values of the coefficients are unreliable, although the overall
model structure is still accurate.
The fall period simplifies the basement model even further. Since the
drains are plugged, the basement radon no longer shows the strong dependence
on the subslab radon concentration. (Although they now have a reasonable
negative correlation, this is basically incidental.) Therefore, the fall
radon levels in the basement will now be modelled only by temperature and
pressure differences. The results are presented in Table 11. For the fall
periods, the simple pressure models have a much better fit than they did in
the summer. This indicates that, when the drain is blocked, the basement-
outdoor pressure difference represents the driving force for radon entry
into the basement. Even the temperature models produce better fits than
before. Plots of the measured versus modelled values are given in Fig. 4.
7. SIMPLIFIED MODELS
In general, models are designed to explain the maximum amount of
variance in the data. While this method produces the most accurate model,
it does not necessarily produce a model that is easy to interpret. In
-------
particular, the radon models developed previously are extremely difficult to
analyze because the significant time series parameters vary widely from week
to week. We can simplify the radon models considerably and hopefully aid in
their interpretation by restricting each parameter to a single lag. In order
to minimize the information loss from this step, we will now allow for half-
hour lags. The new simplified drain models are given in Table 12 (for
convenience the brackets give the of the original model) while the
basement models are presented in Table 13. The optimum single lag term for
each parameter was calculated by an exhaustive search of all reasonable
combinations for each model. This required a lot of time and effort, one
disadvantage of the simplified model. However, a close approximation of the
best model is relatively easy to calculate.
While the simplified summer drain radon models look very different as
the season progresses, the actual drain radon data remain much the same.
The difference lies in the flat troughs of the daily cycle that appear in
the later periods. They result from the simple fact that radon
concentrations must be greater than zero. The term [,T0-T-|53'4" reflects the
airflow from the basement into the drain, and once the drain radon falls to
zero, as it rapidly does during the late summer, the term has no impact on
the radon level and thus it disappears. Similarly, the term [Tg-T0]+ has no
effect in Period 4 because the outdoor temperature rarely drops below the
soil temperature, and even when It does it is only by a degree or two.
The simplified fall radon models all look remarkably similar, with all
terms having a 2-hour lag. This indicates that the fall drain model should
be able to fit all of the data at once. The comprehensive fall model is
presented in Table 14, in both the complex and the simplified form. Thus,
it seems that during the fall, the drain radon follows a relatively
consistent pattern dominated by the outdoor-soil temperature difference.
Note that, although the term IT0-Tb]+ has the largest coefficient, T0 is so
rarely higher than that this parameter is only of secondary importance.
The new basement radon models are not nearly as consistent as the fall
drain radon models described previously. In the summer models, the drain
radon term remains relatively consistent but the pressure difference term
varies considerably both in time lag and in magnitude. The fall models are
somewhat better, with the magnitude of the pressure coefficient varying
between 30 and 40 pCi/L/Pa.
8. CONCLUSIONS
One very significant element of this analysis is missing: a study of
the residuals. It is excluded since the goal of this study was not to
produce an ideal time series model of radon behavior, but to come
up with rougher but simpler relations that lend themselves to physical
interpretation. It is important not to get overwhelmed by statistics to the
extent that one loses sight of the final goal. Up to this point, good time
series models of short-term radon behavior are almost non-existent. This
has a lot to do with the complexity and variability of radon measurements.
In fact, the residuals of the radon models in this study show a strong auto-
correlation, because the magnitude of the temperature effect varies
considerably, even on a day to day basis. Although this means that the
models have significant problems, it does not invalidate them as guides to a
-------
physical analysis of the system.
The models developed in this analysis show that temperature differences
can explain between 60 and 80% of the variation in radon concentration in
the basement and the drain of House 22. They reveal a relatively complex
system of air dynamics where the drain works as a pump, pulling radon into
the basement during one phase of a diurnal cycle and sucking radon out
during the other phase. As the temperature differences between the soil,
the basement, and the outdoors change as summer shifts into fall, an
entirely new pattern of drain radon flow emerges. The drain sealing
probably has some impact on the dynamics of the system, although its precise
effect cannot properly be measured. Although the models reveal general
patterns, their complexity makes them quite difficult to analyze. The
simplified models sacrifice between 5 and 10% in terms of the coefficient of
variance, but they prove to be much easier both to compare end to interpret.
Because of the quantity of data, particularly when examined at the
half-hour level, this study has been limited to two seasons out of the year.
An expansion into both the winter and the spring seasons would be desirable.
Unfortunately, the clear diurnal pattern that is present in the summer and
fall data completely disappears during the winter, making modelling much
more difficult. In wintertime, Te never rises above T^, meaning that the
drain does not pump radon out of the basement and the levels remain
relatively high. However, the winter season certainly deserves further
study. Despite the pitfalls, time series regression analysis can produce
reasonable models for half-hour radon data.
9. ACKNOWLEDGEMENTS
The author has benefited from discussions with K. Gadsby, A. Reddy,
R. Sextro, and R. Socolow. Critical comments by R. Mosley of
U.S. EPA/AEERL are acknowledged.
10. REFERENCES
1. D.A. Hull, Time Series Linear Regression of Half-Hourly Radon Levels in
A Residence, PU/CEES Working Paper No. 110, August 1989.
2. T.J. Noonan, Modeling Levels of Radon Gas and Progeny, PU/CEES Working
Paper No. 100. April 1958.
3. Sachs, L., Applied Statistics. 2nd Ed., Springer-Verlag, New York, 1984.
Table 1 - Parameters measured continuously In H22
Rnj} - basement radon measured by a Wrenn chamber (pCi/L)
Rndr - radon level in drain beneath the basement measured by a CRM (pCi/L)
Tb - basement temperature (°C)
T0 - outdoor temperature (°C)
Ts - soil temperature (°C)
dPb/o - pressure difference, ambient - basement (Pa)
-------
Table 2 - Description of the data periods
Period
Julian days No. days
Reason for exclusion
2
3
4
5
6
7
8
9
148-155
155-164
165-174
175-184
185-194
195-203
204-210
211-220
221-232
233-242
243-252
253-262
263-273
274+
7
10
10
10
10
9
7
10
12
10
10
10
11
cooler weather / uncharacteristic summer
pattern
heavy rains / probable drain flooding
initial attempt to block drains
drains sealed / radon levels decreased
drains re-opened
Table 3 • Drain radon models (Period 1)
Days 148-155
a) Rndr = 287[Tb-T0] + 5895 R2 = 0.366
b) Rndr = 1138[dPb/0] + 915[dPb/ollhr + 4975 R2 = 0.381
Days 148.5-154
c) Rndr = 248[Tb-T0] + 249[Tb-T0]2hr + 6737 R2 * 0.816
d) Rndr= 1084[dPb/o] + 1088[dPb/ohhr + 5114 R2 = 0.669
e) Rndr = 266[Tb-T0] + 312[Tb-T0hhr - 400[T,-To]+ihr ,
- 301 [Ts-T0]+2hr + 7147 R - 0.853
Days 148-155
f) Rndr = 299[Tb-T0] + 314[Tb-T0hhr - 594[T,-T0]+ihr ,
- 623[T,-T0]+3hr + 7242 R ¦ 0.764
Table 4 - Drain radon models (Days 189-204)
a) Rndr = 793CTb-T0]++ 562[Tb-T0]+2hr r2 - 0.783 (0.860)
b) Rndr «= 2256[dPb/ol+ + 32l9[dPb/0]+lhr + 3310[dPb/o]+2hr
R2 - 0.836 (0.894)
-------
Table 5 • Drain radon revised model (all summer periods)
Period Model
a) 148-155 Rndr = 556[Tb-T0]+ + 590[Tb-To]+2hr - 152[T0-Tb]+
- 209[To-Tb]+2hr - 646[TS-T0]+ - 936[T,-T0]+2hr
- 276[TI-T0]+4hr + 5604 R2 = 0.802
b) 165-175 Rndr = 760[Tb-Tol+ + 462[Tb-T0l+ihr - 132(T0-Tbl+
- 213[T0-Tb]+4hr - 1131[TJ-T0]+
- 871 [Ts-T0]+ihr + 2395 R2 = 0.827
c) 175-185 R2 = 0.400 no consistent model structure
d) 185-195 Rndr - 1093[Tb-To]+ + 445[Tb-T0]+2hr + 222[Tb-T0]+4hr
- 932[T,-T0]+ - 872[TJ-T0]+lhr
- 405[T,-To]+2hr + 202 R2 * 0. 756
e) 195-204 Rndr = 856[Tb-T0]+ + 465[Tb-T0]+2hr R2 = 0.741 (0.832)
Table 6 - Drain radon: reduced modelling periods
Period Model
a) 175-179 Rndr ¦ 772[Tb-T0]+ + 578[Tb-T0]+ihr - 708[T,-To]+
- 1058[T,-ToJ+ihf + 1937 R2 « 0.676
b) 149-155 Rndr = 742[Tb-T0]+ + 334[Tb-T0]+3hr - 237[T0-Tb]+
- 188CT0-Tb]+4hr - 685[T,-T0]+ - 582[T,-T0]+,hr
- 416tT,-T0]+3hr + 6144 R2 « 0.896
c) 185-194 Rndr « 1143[Tb-T0]+ + 502[Tb-To)+2hr + 183[Tb-T0]+4hr
- 1089[Tj-To]+ - 822[Tj-T0]+ihr
- 464[T1-T0]+2hr + 376 R2 « 0.823
Table 7 - Drain radon: fall models
Period Model
a) 233-243 R2 = 0.362 no consistent model structure
b) 243-253 Rndr « 1104[To-T1)+ + 1490tTo-TJ]+2hf + 1069[To-T,l*3hr
+ 857[T0-T,]+4hr - 154[Tj-T0]+ - 260[TJ-To]+3hr
- 19S4[To-Tb]+ - 1704[To-Tb]+ihr - 2570(To-Tb]+2hr
- 2032[To-Tb]+3hr - l966[T0-Tb]+4hr + 5048
R2« 0.718
e) 253-263 Rndr « 1271[To-T,]+lhr + 94ltT0-T1]+2hr + 645[T0-T,)*3hr
- 307[Tj-To]+2hr - 1835[T0-Tb]+lhr - 1537[T0-Tb]+2hr
- 1063(To-Tb]+3hr - 842(T0.Tb]+4hr + 3955
R2« 0.629
d) 263-274 Rn* - 732[T0-T,)+ + 708[To-T,]+ihr + 1177lT0-TiJ^hr
+ 1383[T0-T,]+4hr - 367[T,-T0]*ihr - 170[T,.To]*4l»
- llOiPVTb]* - 1268ITo»TbJ+jhf - 2033(To-Tbl+2br
- 2S72[To-Tb]+4hr + 5429 R2 - 0.760
-------
Table 8 • Drain radon: fall model tested on summer data
Period Model
179-184 Rndr = 201[To-Ts]+3hr - 464[TrT0]+ - 136[Ts-T0]+4hr
- 644[To-Tb]+ihr+ 6824 R2 = 0.596
Table 9 - Basement radon models (Days 148-155)
a) Rnt" 0.0084[Rn
-------
Table 11 • Basement radon: fall models
Period Models
a) 233-243 Rnb = 14.5[dPb/ohhr + 14.8[dPb/okhr
+ 13.7[dPb/0]4hr - 7.3 R2= 0.779
Rnb = 3.4[Tb-T0]+ + 4.4[Tb-T0]+4hr R2 = 0.725 (0.875)
b) 243-253 Rnb « 14.6[dPb/0lihr + 16.3[dPb/ohhr + 5.5 R2 = 0.633
Rnb = 3.8[Tb-T0]+ + 2.9[Tb-T0]+3hr + 12.3 R2 = 0.618
c) 253-263 Rnb = 6.0[dPb/o] + 13.2[dPb/0]ihr
+ 15.9tdPb/0hhr - 4.9 R2 « 0.754
Rnb = 2.5[Tb-T0]+ + 2.8[Tb-T0]+ihr + 16.2 R2 = 0.571
d) 263-274 Rnb = 9.5[dPb/0] + 13.6[dPb/0]ihr
+ 12.8[dPb/0]2hr R2 = 0.820 (0.955)
Rnb = 5.6[Tb-T0]+ + 1.6[Tb-T0]+3hr + 114 R2 ¦ 0.716
Table 12 • Drain radon: simplified models
Period Summer Model
a) 148-155 Rndr - 795[Tb-T0]+ihr - 430[To-Tb]+ihr
- 13l0[Ts-To]+2hr + 6282 R2 = .0.752 (0.802)
b) 165-175 Rndr = 1183[Tb-To]+0.5hr - 210[To-Tb]+2hr
- 1747[Tj-T0]+o.5hr + 2003 R2 = 0.781 (0.827)
c) 185-195 Rn
-------
Table 13 - Basement radon: simplified models
Period Summer Model
a) 148-155 Rnb-0.00&0[Rndrllhr + 12.9[dPb/oh.5hr + 2.6 R2 = 0.864
Rnt>- 23.8[dPb/oll.5hr + 32.6 R2 * 0.598
b) 165-175 Rnb - Q.0059[Rndr]l.5hr + 21.3[dPb/ol3hr + 10.3 R2 «= 0.775
Rnb - 39.7[dPb/ol3hr + 23.8 R2 = 0.628
c) 175-185 Rnb -0L0039[Rndrh.5hr + 18.9[dPb/0ll.5hr + 9.9 R2 = 0.641
Rnb - 17.5[dPb/oll.5hr + 26.7 R2 « 0.452
d) 185-195 Rnb-0.0062LRndrll.5hr + 16.0[dPb/ol2.5hr + 7.6 R2 = 0.822
Rnb - 27.0[dPb/oh.5hr + 20.4 R2 = 0.588
e) 195-204 Rnb- 0.0051[Rndr]ihr + 6.1[dPb/0]2.5hr + 4.0 R2 « 0»78®
Rnb- 24.6[dPb/ol2.5hr + 13.4 R2 - 0.537
Period Fftll Model
f) 233-243 Rnb = 39.5[dPb/ol2hr - 4.1 R2 « 0.751
Rnb ¦ 6.9[Tb-T0]+2.5hr + 3.5 R2 » 0.691
g) 243-253 Rnb * 29.9[dPb/oh.5hr + 7.2 R2 «= 0.620
Rnb « 6.2[Tb-T0]+i.5hr + 15.7 R2 ¦ 0.606
h) 253-263 Rnb « 32.3[dPb/oll.5hr R2 » 0.737 (0.934)
Rnb ¦ 5.2[Tb-T0]+o.5hr + 16.8 R2 ¦ 0.567
i) 263-274 Rnb - 35.7[dPb/0]l.5hr *2 » 0.792 (0.948)
Rnb « 6.8[Tb-T0]+o.Jhr + 14.1 R2 * 0.701
Table 14 - Comprehensive fall drain radon model
Period Regular Model
233-274 Rndr « 579[T0-T,]+ + 751[T0-T,]+ihr + 1026[To-T,]+2hr
+ 722[T0-T,]+3hr + 709[To-T,]+4hr - 388[T,-T0]+2hr
- 878[T0-Tb]+ - 1547[T0-Tb]+ihr - 1669[T0-Tb]+2hr
- 1214tTo-Tb]+3hr - 1841[T0-Tb]+4hr + 4655
R2 « 0.676
-------
loooo -
g 8000
g 6000
*o
«s
4000 -
2000 -
0 -
148
Julian Day
Fig, 1 Comparison of measured drain radon with model predictions
using temperature differences (Period 1)
Julian Day
Fig. 2 Comparison of measured drain radon with model predictions
using temperature differences (Period 7)
-------
Julian Day
Fig. 3 Comparison of measured basement radon with model predictions
using drain radon and basement-outdoor pressure difference
(Period 1)
244 246 248 250 252
Julian Day
Fig. 4 Comparison of measured basement radon with model predictions
using basement-outdoor pressure difference (Period 7)
-------
III-2
Experience with the Wire Screen
unattached Fraction Measurement Technique
Van Cleef, Douglas J. and Windham, Sam T.
U.S. Environmental Protection Agency
Eastern Environmental Radiation Facility
Montgomery, Alabama
ABSTRACT
Polonium 218, historically known as Radium A, is commonly
considered to be a major contributor to adverse health effects from
exposure to radon decay products. The free ions, or unattached
fraction, of RaA, by virtue of deeper tissue deposition in the
respiratory tract, are believed to be responsible for the greatest
portion of this health risk. Because of this, much work has been
performed in the development of measurement methods for radioactive
particles as small as 0.001 microns, such as unattached Radium A.
This work includes development of methods ranging from simple wire
screens to complex diffusion batteries. Most of the previous work
has been performed using high concentrations of radon or its decay
products, providing only theoretical assumptions for the
application of the work to environmental levels of radon. This work
discusses the application, limitations, and assumptions regarding
the use of the single wire screen RaA measurement technique to the
lower environmental levels expected in most homes, schools, and
office buildings.
This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's peer and administrative review
policies and approved for presentation and publication.
-------
INTRODUCTION
In the past 3 decades, many scientists and engineers have
contributed much to the understanding and measurement of unattached
Polonium-218 (RaA) particles. Some of the foundational work for
the methods evaluated here was performed by THOMAS and HINCHLIFFE
and by GEORGE(2) using wire screens as diffusion collection devices.
MERCER{35 had previously developed an effective impaction stage
diffusion sampler. RAGHAVAYYA and JONES(4) developed a method for
estimating RaA unattached fraction (f8) using wire screens and
backup filters. More recent work by VAN DER VOOREN et all5) and
RAMAMURTHI and HOPKE (6) have described more sophisticated
unattached RaA measurement techniques and limitations of many of
the available methods. Some recent work by HOPKE et ali7) has
included the use of multi-stage wire screen diffusion batteries to
enable more complete RaA collection and particle size
determination.
The work described in this paper has sought to determine the
specific limitations and advantages of the single wire screen
collection method, for the following reasons: (1) the method only
requires the simultaneous use of two counters, unlike the diffusion
batteries or impaction stage devices, which require at least three;
(2) the method retains use of the reference filter and its
established analysis methods, unlike the wire screen/backup filter
combination; and (3) the method is simple and easily adaptable to
field work, where the EPA intends to use it.
A secondary goal of this work was to evaluate the use of the
reference filter/ wire screen/ backup filter measurement technique
for use in field measurement applications.
TESTING FACILITIES
Most of the measurements and analyses were performed at the
U.S. EPA's Eastern Environmental Radiation Facility (EERF) in
Montgomery, Alabama. The facility has two radon exposure chambers,
each with unique environments and sampling limitations. EERF's
older chamber, designated Chamber C, is a 3.6 cubic meter volume
chamber of wooden construction. It is a single pass, flow through
chamber with a heated nichrome wire supplying condensation nuclei
(CN) . Incoming air is diffused through a perforated particle board
upon entry to the chamber, and the radon decay products are
typically about 30 minutes old when they reach the sampling points.
The facility's other chamber, designated Chamber A, is a 40
cubic meter volume, recycled environment chamber of mostly aluminum
-------
construction. Condensation nuclei are provided by a wax generator.
Radon and its decay products are normally in secular equilibrium
in this chamber. Each chamber receives its radon from a series of
commercially-available flow-through Ra-226 sources.
SAMPLING DEVICES
Based on the recommendations of GEORGEc2>, a 60 mesh wire
screen was decided upon as the collection device for unattached
RaA. The stainless steel screen was cut to size and pressed into
a custom holder of 2 concentric rings, also of stainless steel, to
provide a reproducible and convenient method of support for the
screen (Figure 1) . A commercially available sample head was
modified to accept the 52mm diameter
screen and a 47mm diameter backup filter,
then adapted for use through the chambers'
sample ports (Figure 2).
The reference filter sampler is a
commercially-available plastic filter
holder, normally containing a 25mm
membrane-type filter of 0.8 micron pore
size. This filter arrangement has been
used extensively in radon decay product
measurements and serves as the EPA's
standard device for such measurements.
SAMPLING AND ANALYSIS
Figure 1. 60 Mesh
Stainless Steel Wire Samples of the chamber atmospheres were
Screen taken under a number of widely varied but
controlled conditions, in an effort to
establish measurement results comparable
to those obtained in previous works. Variable sample conditions
included sample flow rate, radon and condensation nuclei
concentration, and sample duration. All samples taken with the
wire screen device were accompanied by simultaneous measurements
using the reference filter arrangement. Many samples were taken
using the reference filter, wire screen, and backup filter in an
effort to estimate fa from only the reference filter and backup
filter values, as described by RAGHAVAYYA and J0NES(O. Sample flow
rates varied from 13 to 42 liters per minute, with corresponding
screen face velocities of 15 to 48 cm sec'1. Condensation nuclei
concentrations were adjusted within a range of about 3xl03 to 1x10s
particles-cm"3, and chamber radon concentrations varied from under
10 to over 350 pCi-1"1. All condensation nuclei measurements were
performed with a single TSI Model 3020 CN counter.
-------
Screen and filter samples were analyzed
simultaneously in EERF's radon counting
room. The samples were placed particle
side down on disks of ZnS(Ag) on a two-
inch photomultiplier tube which was
enclosed in a light-tight housing. Signal
counts were collected by Ludlum model 2200
scalars attached to and controlled by a
single personal computer. The computer
provided all counter timing and results
analysis accordinq to the Thomas-modified
Tsivoglou method . Analysis of the
reference filter provided individual radon
Figure 2. Wire Screen decay product concentrations, but wire
Sample Head screen decay product results had to be
modified to correct for collection
efficiency and the loss of counts from decay products on the sides
or back of the screen mesh. The collection and counting efficiency
values were estimated from work by THOMAS and HINCHLIFFE(1) and
modified according to actual observations provided by GEORGE(2).
These estimates predict a rapid drop in collection efficiency with
changes in face velocity from 0 to 20 cm-sec"1, then a more gradual
drop with increasing sample velocities. Collection efficiency
estimates, corrected for sample flow rates, were generally between
40 and 50 percent. Calculations included error estimates which
were subseguently used to evaluate the validity of the sample
results.
RESULTS
As previously stated, data were collected in an effort to
reproduce results of previous work with similar eguipment, as well
as to establish a correlation between theoretical and practical
measurement capabilities. One criteria used for comparing our
results was to evaluate the shape of the plot of condensation
nuclei concentration versus unattached fraction. Several previous
works, including those by GEORGE and HINCHLIFFE<9>, RAABE<10>, and
M0HNEN(11> have shown fa to drop rapidly between 3xl03 and 1.5x10*
particles-cm3. Our ability to reproduce these curves would serve as
a good measure of the serviceability of our measurement technique.
An additional measure of the value of the method would come from
the error analysis of the counting results.
The measurements performed in Chamber A provided results that
corresponded very nicely with previous work and theoretical
projections, as shown in Figure 3. The curve shape is comparable
-------
to those of previous works, and the actual values at the extremes
of the measurement range are consistent with expectations.
Counting errors on the wire screens ranged from about 10 percent
at radon concentrations above 20 pCi-liter" and low CN
concentrations to a high of 122 percent at 7 pCi-liter and high
CN concentration. Large sample volumes were necessary to minimize
counting errors, and these were obtained by maintaining high sample
flow rates. There is no evidence of deviation from expected screen
collection efficiencies at the high sample flow rates used.
Estimates of fa by the backup filter method were generally
consistent with those made with the wire screen, but had higher
uncertainties at low radon concentrations and high CN values and
a larger resultant data spread. For this reason, the results from
backup filter measurements were not used as evaluation criteria.
Measurement results in Chamber C were somewhat different than
expected. Where the Chamber A results and previous work
demonstrated a rapid, exponential-shaped drop in fa with a rise in
CN concentration, the unattached fraction values in Chamber C
RaA fa vs. CN Concentration
EPA EERF Chaaber A
~! 1 T
40 60
(TUousands)
CX Concentration (luclel/CC)
+ Measured RaA Fa
Figure 3.
- Dual ParaDola Exponential rit
Chamber A f, vs. CN Concentration
-------
dropped at a much slower, nearly linear rate as CN values rose (see
Figure 4) . The actual fa values at all condensation nuclei
concentrations were somewhat greater than predicted by previous
work or measured in Chamber A. As with measurements from Chamber
A, backup filter results did not provide sufficiently consistent
confirmation of f. estimates to be used as evaluation criteria.
Several theories are currently under consideration regarding
the unusual shape of the fa versus CN curve in Chamber C. The most
likely cause is related to source age and introduction method. In
Chamber A, the radon gas is introduced through a recirculation
plenum, where it mixes with the radon and CN throughout the
chamber. In Chamber C, the radon gas is introduced though a single
pass inlet plenum, where it is mixed with the CN. The gas/CN
mixture is diffused across a large perforated particle board, and
is then assumed to exhibit laminar flow through the chamber. The
total time between introduction of gas and CN is about 30 minutes
before sampling. While this time allows for sufficient ingrowth
of RaA, it may not be sufficient to allow adequate combination of
RaA fa vs. CH Concentration
EPA SERF cnatber C
1 -
0.9 -
0.8 -
0.7 -
B
'rt
S 0.6 -
u
+ + + +
I 0. 5 -
L)
5LW* + +
s
1 °'4 -
D
* / + % k
+ + +
<
a
0.3 -
>^ *+ * r ;———
0.2 -
+•
0.1 -
0 -
i 1 1 1 i r 1 1 1
0 20 40 60 80 100
(Thousands)
CN Concentration (Kuclei/cc)
* Measured RaA Fa - Single Parabola Exponential rit
Fugure 4. Chamber C fa vs. CN Concentration
-------
RaA and CN for representative f versus CN analysis. Further
evidence of the limited mixing time is demonstrated by the low
fraction of RaB and RaC*measured in both the screen and reference
filter analyses.
In addition to differences in measured fa which may result
from the introduction method, it is probable that the measurements
suffer from a bias in sample particle size. The unattached radon
decay products in Chamber C are likely to be much smaller than
those in Chamber A due to the age of the source at the time of
sampling. It is also possible that the different chemical
properties of the condensation nuclei affect the rate of
combination with the RaA.
The data obtained from measurements in Chamber C is valuable
for the purposes of determining counting errors and fa measurement
reproducibility, but not for confirmation of screen collection
efficiencies. Screen counting errors ranged from 4 percent at high
radon/low CN conditions to 38 percent at low radon/high CN
conditions.
CONCLUSIONS
Several specific conclusions can be drawn from the analysis
results, while several additional questions now require addressing.
It is clear that simple wire screens can be used successfully as
collection devices for free RaA within a reasonably well known set
of limitations. Sampling with wire screens and backup filters does
not provide more reliable results in difficult measurement
circumstances. Sample accuracy at low radon/high CN concentrations
is improved by increasing the sample volume. Screen collection and
counting efficiencies appear to be consistent with that predicted
by previous works.
Sampling with the wire screens at very low radon
concentrations or very high condensation nuclei concentrations is
likely to yield large counting uncertainties. To improve counting
statistics under those conditions, large sample volumes are
required, necessitating high sample flow rates or long sample
durations. The same effect can be achieved with larger screens and
high sample flow rates, as long the counting equipment can
accomodate the larger screens. Due to the short half-life of RaA,
sample durations longer than 10 minutes do not appreciably improve
counting uncertainties.
In an effort to better understand the differences in results
from EERF's Chamber C and expected results, measurement work will
-------
continue there, focused on proving (or disproving) the previously
stated hypothesis.
-------
REFERENCES
(1) J.W. Thomas and L. Hinchliffe, J. Aerosol Sci. 3, 387 (1972)
(2) A.C. George, Health Physics 23, 390 (1972)
(3) T.T Mercer and W.A. Stowe, Health Physics 17, 259 (1969)
(4) M. Raghavayya and J.H. Jones, Health Physics 26, 417 (1974)
(5) a. Van der Vooren, A. Busigin, and C.R. Phillips, Health
Physics 42, 801 (1982)
(6) m. Ramamurthi and P.K. Hopke, Health Physics 56, 189 (1989)
(7) P.K. Hopke, M. Ramamurthi, and E.O. Knutson, in press (1989)
(8) j.w. Thomas, Health Physics 23, 783 (1972)
(9) A.c. George and L. Hinchliffe, Health Physics 23, 791 (1972)
(10) o.g. Raabe, Health Physics 17, 177 (1969)
(11) V. Mohnen, AERE Trans. 1106 (1967)
-------
Session Ill-Pi:
Panel Session on QA/QC of Measurement
-------
III-pl-1
QUALITY ASSURANCE AMD QUALITY CONTROL IN THB RADON MKASURBMBNT INDUSTRY
WHAT IS MISSIHB?
by: Terry E. Howell
Radon Reduction & Testing, Inc.
Atlanta, Georgia 30329
ABSTRACT
Since the release of survey results by the consumer advocacy group
BUYERS-UP, QA/QC has been the buzz word among everyone involved with
Radon/Radon Decay Product measurement. This paper will present a discussion
in generality of the roles of Quality Assurance and Quality Control in the
Radon Measurement Industry, the strengths and weaknesses of QA/QC in radon
measurement and the significance of research in QA/QC for radon measurement.
What are the important issues of quality assurance and quality control
in radon measurement and...
WHAT IS KISSING?
-------
INTRODUCTION
" The EPA has established the National Radon Measurement
Proficiency (RMP) Program. This quality assurance program
enables participants to demonstrate their proficiency at
measuring radon and radon decay product concentrations."
Everyone involved in radon measurement has read this statement - it
appears in almost every EPA publication dealing with radon measurement and
radon measurement protocols. But do we understand its meaning? Do we truly
understand the purpose of the RMP Program and it's significance to the Radon
Measurement Industry? If so, then why has the RMP Program left us stumbling
blindly in a state of confusion, distrust, and uncertainty that is near chaos?
Why do we hear statements from measurement companies that..."The RMP is really
a certification program"... or ..."The EPA wants to control the Radon
Industry"... or ..."The EPA has to do more than test our accuracy once every
twelve to eighteen months"...? And why are consumer organizations and state
agencies saying ..."the EPA must make the RMP Program tougher and more
viable"...? WHAT IS MISSING?
DOLLARS AMD SENSE
Perhaps the most important thing missing in Quality Assurance and
Quality Control in the Radon Measurement Industry is simply that we do not
understand the real reason for a QA/QC Plan. The reason that successful
businesses develop and maintain Quality Assurance and Quality Control programs
is because the results directly influence the bottom line of their Profit and
Loss Statements. And it makes good business SENSE to pay close attention to
those factors that directly influence the DQLUUBS.
Statistical studies indicate that most businesses fail during their first
three years and that the overwhelming majority of those failures are the
result of one thing-Under Capitalization. The part that is really disturbing
to financial analysts, lenders and investors is not that new ventures
fail because of under capitalization but that most of the failures did not
realize that they were under capitalized. They had never developed a Business
Plan. If they had the cash flow projections in the plan would have identified
the level of capitalization necessary to sustain them until they were
profitable.
-------
Developing and implementing a quality assurance plan and a quality
control plan is nothing more than one integral part of our business plan. The
results directly influence the two most important areas of our business
plan—The Financial Statement (Reserves for Returned Goods and Services), and
The Marketing Plan (in competitive, educated markets Quality Assurance may be
the only advantage and therefore the core of the entire marketing strategy).
So how do we develop a Quality Assurance and Quality Control Program? The
first step is to understand what the terms mean.
Quality Assurance is the warranty or guarantee of quality and
durability for goods and services that we impart to our customers.
Quality Control is the system of inspections, testing and other
measures that we employ to insure that claims against our warranty do
not exceed anticipated levels (Reserves for Returned Goods and
Services) and thereby become detrimental to our business financially.
As we can see these terms mean two entirely different things. But while they
are not synonymous one is meaningless without the other.
Once the commitment has been made, the development of a Quality
Assurance/Quality Control Plan is very much like the development of other
portions of the business plan. We must perform a series of exercises and
projections that when combined with an honest and objective evaluation of our
capabilities and resources will provide us with all the elements of the plan.
ELEMENTS OF QUALITY ASSURANCE
The following are some of the important areas that should be considered
in developing the QA Plan. More areas may need to be addressed depending on
the complexity of the business structure or the scope of the project to be
addressed.
1. Identify the benefits of a warranty. Will it give a competitive
edge or is it mandatory for the market or the project? Will there
be an implied warranty in the absence of a written warranty.
2. Identify the risks of a warranty.
3. Identify the acceptable level of exposure. To what degree will or
extent will the warranty cover. Are there specific requirements?
-------
4. Identify how failures will be handled. What will be the
procedures for handling claims? What resources and reserves will be
necessary to handle the predicted number of claims?
5. Formulate the written warranty. YOUR ASSURANCE OF QUALITY.
6. Designate personnel responsible for implementation of the program.
The Quality Assurance Officer
7. Formulate a plan for insuring that claims on the warranty do not
exceed the acceptable limits and become detrimental to the business.
The Quality Control Plan.
8. Plan for catastrophic events. General, Product, Professional
Liability Insurance. Is it available? Is it cost effective?
9. Seek out opportunities to demonstrate your capabilities or the
effectiveness of your product. Are their existing programs
available through consumer or government agencies that compare
products or services to established or acceptable standards and
publish their findings?
10. Monitor your performance. Evaluate performance in intercomparison
programs. Evaluate number of warranty claims. Evaluate the impact
on the market position.
ELEMENTS OF QUALITY CONTROL
The QC Plan will generally be technical in nature but must be designed
to conform to the needs of the QA Plan. It is after all the primary insurance
policy against warranty claims. The complexity of the plan will vary in
proportion to the complexity of the work being performed. The following are
some of the important areas to address.
1. Identify the task. This involves a thorough understanding of the
over all scope of 'the project. The project should be divided into
as many sub-categories as possible and then the portion or portions
of the work to be performed should be identified.
2. Identify all the variables or factors within the task that will
affect the performance of the work. What effect will each have?
Can adverse effects be neutralized, standardized, or must they be
accepted as is? How will each be handled?
-------
3. identify how the task will be performed. What equipment, methods or
procedures will be utilized? What are the underlying theories or
principals of operation of each? What factors or variables will
affect the performance? How will these variables be handled? What
are the maintenance and/or calibration requirements of the
equipment? Is the method or equipment appropriate for the task? Is
it practical?
4. Identify personnel requirements. What are the background
requirements? What specific training will be required? What will
their responsibilities be?
5. Identify the control options for each variable. What are the
advantages and disadvantages of each? Are they appropriate? Are
they practical? Are they effective?
6. Identify acceptable limits for precision and accuracy? Are there
industry standards? Do they meet the needs of the Warranty?
7. Establish procedures for performance checks. How will each be
performed? How often? How will failures be handled?
8. Adopt the plan. Put it in writing. Make it official by adoption at
directors meeting.
9. Implement the plan. Make performance checks. Document the results.
10. Review performance. Were schedules met? Were performance
specifications met? Were they sufficient to meet the needs of the
warranty.
THB SirariFICANCS OP RESEARCH
Lets take a look at the process for developing a QC Plan.
Where do we get the information to identify the strengths and weaknesses
of our detectors, methods, and instrumentation?
How do we discover the variables that influence our detectors, methods,
and instrumentation?
How do we evaluate the effects of these variables?
-------
In short, vhere do ve acquire all the information that enables us to
1. Select the appropriate detector, method and instrumentation to
perform the measurement.
2. Identify control options for the variables that influence
performance.
3. Establish acceptable limits for precision and accuracy.
4. Establish procedures for performance checks.
Research is the very core of a Quality Control Program. if we do not
understand the underlying principles of operation, capabilities and
limitations, and the cause and effect of variables on our detectors, methods,
and instrumentation we will never be able to develop much less implement an
effective Quality Control Program.
WHAT IS MISSIMG?
What is missing? An honest objective evaluation of ourselves as an
industry. We are unqualified, and untrained. We are unaware that there is
chaos in our industry due to lack of confidence in ourselves and the absence
of public trust. We are out of control, completely unable to offer an
industry wide assurance of quality. Financial analysts have labeled us as a
low tech, low potential industry and therefore unworthy of investor
confidence. Anyone with a credit card and a business card can be a radon
measurement expert before the sun goes down!
What is missing? Responsibility« We have lost sight of the scope of
the project and its implications. We are dealing with an environmental public
health issue. We have a moral obligation to be correct. Our clients are
making decisions concerning the lives of themselves, their families, and in
the case of public buildings, the general population, solely on the information
are giving them.
What is missing? Training. No more awareness and overview training.
We must have method specific find task specific training that incorporates an
understanding of the underlying principles and theories of operation with
method and task specific QA/QC procedures. Trained, qualified personnel is
essential for quality control because it reduces the largest potential for
error—human error due to Ignorance.
-------
What is missing? Proper evaluation of our Methods, detectors, and
instrumentation. An evaluation of our methods, detectors, and instrumentation
in accordance to specific criteria will identify their capabilities and
limitations, the measurement purpose for which they are most appropriate, the
influence of variables, and the sources of potential error. This would lead
to the establishment of specific QC procedures for each. Combined with a seal
of approval, this would enable us to have the highest possible confidence in
our measurements, and would work to restore public trust by acting as a
deterrent to companies talcing detectors and instrumentation to the market
place without industry evaluation and acceptance.
What is missing? Accreditation. An industry wide quality control
check to insure that our personnel have received the proper method specific
training and demonstrated their understanding of the radon measurement
principals and quality control procedures they will be utilizing. An
accredited industry establishes an assurance of quality for all and is
necessary if we are to squelch the "Tin-Man" image we presently portray.
What is missing? Verification. An inspection of our QC procedures,
performance, and documentation to insure that the industry is practicing the
daily, weekly, and monthly quality control necessary to provide a maximum
assurance of quality deserving of the public trust.
What is missing? Definition. The EPA must define what it feels its
role in radon QA/QC is. Is it political? Is it regulatory? Is it industry
support? is it consumer protection? Is it competition? The EPA must define
what it expects from the radon industry in straight forward "tell it like it
is language" not ambiguous statements filled with political rhetoric. More
importantly, the radon industry must define its role. We must decide just
what we expect from the EPA. Do we want government regulation of our
business?
Both the radon industry and the EPA need to take a hard look at the RMP
Program together and define it for what it is. Is it really a quality
assurance program? If so where is the guarantee and to whom is it being
conveyed? is it a certification program that state and local agencies should
use as minimum requirements (as many already do)? If so, what or whom is
being certified? Perhaps it is time that the EPA stopped calling the RMP a
quality assurance program. More importantly, is it tiae the radon industry
stopped calling it a quality assurance prograa.
What is missing?
AH INDUSTRY WIDE ASSURAHCE OF QUALITY MID THE C0MTR0L3 TO BACK IT OP1
The work described in this paper was not funded by the U.S.
Environmental Protection Agency and therefore the contents do not
necessarily reflect the views of the Agency and no official
endorsement should be inferred.
-------
III-P1-2
LQT-TO-LOT VARIATION IN THE PERFORMANCE OF PASSIVE ALPHA TRACK RADON DETECTORS
by: David L. Wilson
Oak Ridge National Laboratory
Oak Ridge, Tennessee
WITHDRAWN BY AUTHOR
-------
III-P1-3
COMPARISON OF THE ELECTRET-PASSIVE RADON MONITOR SYSTEM
WITH CHARCOAL CANISTERS IN CONTROLLED ENVIRONMENTS
by: James R. Summers, William D. Nicholas, and
Robert L. Clemons
Analysas Corporation
Environmental Technology Division
101A East Tennessee Avenue
Oak Ridge, Tennessee 37830
and
David L. Wilson
Martin Marietta Energy Systems, Inc.
Health and Safety Research Division
Oak Ridge National Laboratory
Post Office Box 2008
Oak Ridge, Tennessee 37831-2008
ABSTRACT
The Electret-Passlve Environmental Radon Monitor (E-PERM) was exposed
to calibrated chamber radon concentrations of background, ~ 5, and ~ 60
pCi/1. The results were used to derive a statistical model to define the
variability of measurements taken with the E-PERM. These results were
compared with the manufacturer's calibration curves. The E-PERM and
charcoal canisters were exposed simultaneously 1n a known chamber radon
concentration held at a constant level. Results from the quality check of
the manufacturer's calibration of the E-PERM validated the accuracy of the
reported values. Data from the E-PERM and Femto-TECH are 1n agreement,
while the data obtained with the charcoal canisters has a greater
variability at Identical radon concentrations.
-------
INTRODUCTION
With recent interest in indoor radon concentrations, inexpensive radon
screening methods have been developed. Charcoal canisters were the first to
be developed and are currently the most widely used short-term (2-7 days)
screening method. Because the charcoal canister relied on existing
technology and the demand for an inexpensive and simple screening method was
great, charcoal canisters quickly became an accepted radon screening method.
However, studies have shown that charcoal canisters are not an ideal
screening method (1-3). Sensitivity to environmental parameters such as
temperature and humidity and the radon absorption characteristics of
charcoal have lead to uncertainties in the results from screening with
charcoal canisters. Even with this information known, screening with
charcoal canisters was still widely conducted because alternative short-term
screening methods were not available. Efforts were made to modify the
design of charcoal canisters and improve performance. However, the large
number of manufacturers of charcoal canisters and the wide variety of
canister designs and materials used further confounded the validity of the
results.
Technological advances with electret, a piece of dielectric material
exhibiting a quasi-permanent electrical charge, have resulted in the
development of a new short-term radon screening method (4-8). The device is
called the electret passive environmental Rn monitor (E-PERM), and it is
currently manufactured and marketed by Rad Elect Inc.1 The E-PERM device has
shown little sensitivity to the environmental parameters that affect
charcoal radon absorption. However, only a limited amount of test data are
available from the literature on the performance of E-PERM and comparison of
the two screening methods. Thus, radon screening with E-PERM is not yet
widely practiced (compared with charcoal canisters).
The purpose of this study is to conduct an evaluation of the E-PERM's
radon screening capabilities and a comparison of radon screening results
between E-PERM and charcoal canisters in controlled radon environments.
Recommendations for further studies also result from this research.
BACKGROUND
It is important to note the differences in how both methods accomplish
the task of measuring indoor radon concentrations. The following discussion
provides a brief description of each method and summarizes the advantages
and disadvantages of each.
7Rad Elect Inc., 5330 0 Spectrum Drive, 270 Technology Park, Frederick,
Maryland 21701, (301) 694-0011.
-------
CHARCOAL CANISTERS
The charcoal canister method takes advantage of the characteristic of
activated charcoal to strongly adsorb radon. Activated charcoal also
desorbs radon. An equilibrium is reached between the adsorbing and
desorbing of radon and, thus, indoor radon is actively determined. As a
result of this behavior, the measured radon concentration from the charcoal
canister has much the same character as a "grab sample", which gives an
instantaneous concentration biased to the radon level at the end of the test
period. A study by UNC Geotech reported discrepancies of up to 40 percent
from charcoal canister measurements due to variations in radon
concentrations (1). Because indoor radon concentrations have been found to
have significant diurnal variations, this could lead to large discrepancies
between measured radon and actual radon concentrations for the exposure
period.
Studies have found that radon absorption is sensitive to environmental
parameters such as temperature, humidity, and air movement (1,3). The
ability of charcoal to absorb radon increases as temperature decreases, and
at very low temperatures (-78'C) charcoal becomes almost an infinite sink
for radon. Discrepancy between measured radon and actual radon
concentrations could occur as a result of the temperature dependence.
Activated charcoal has a strong affinity for water vapor; as humidity
increases, the ability for radon absorption decreases, which can result in a
false negative. Air movements have a tendency to enhance the absorption,
which can result in a false positive. Bocanegra and Hopke also found that
air-borne organic contaminants can lower the dynamic adsorption coefficient
of activated charcoal (3).
Despite these characteristics, the charcoal canister method is capable
of measuring radon concentrations within EPA's acceptable precision (±25%),
especially under ideal conditions. This precision under ideal conditions is
further illustrated by the large number of radon testing companies that have
passed Round 6 of the EPA Radon Proficiency Program. These tests are
carried out in controlled environments, holding radon concentrations at a
relatively constant level. Charcoal canisters used in this experiment were
obtained from an EPA-listed radon tester. For the purposes of this
research, the manufacturer will remain anonymous.
E-PERM
The E-PERM device 1s a small 40-ml interval container made of a semi-
conducting material with a charged electret attached to the bottom of the
container. The electret 1s made of Teflon1® FEP and is initially charged to
750 volts. The device is "turned on" by a spring loaded plunger that, when
opened, allows the diffusion of radon through a filtered opening. When
« ————————m
Teflon*FEP fluorocarbon manufactured by E. I. du Pont de Nemours and
Company, Inc., Wilmington, Delaware 19898.
-------
closed, radon gas can no longer diffuse into the internal volume of the
container, and the plunger covers the electret. With the E-PERM turned off,
the surface potential of the electret remains constant. The chamber behaves
much like an ionization chamber. As the diffused radon gas undergoes decay,
ions are generated. Negative ions (mostly electrons) are collected on the
positively charged electret. The electret loses surface potential as a
result of the deposition of ions. The loss in surface potential is
proportional to the activity of the ionizing radiation (7,9). The device is
also sensitive to background gamma radiation. Failing to account for
background radiation could result in overestimating radon concentration by
as much as 1 pCi/1.
Because the electret acts as an integrator of radon activity over time,
the resulting measurement provides an average radon exposure during the test
period. Integrated radon concentrations are a more useful diagnostic tool
for radon testers because further actions are based on average exposures.
Kotrappa<14) found that E-PERM detectors showed little sensitivity to
humidity (8). However, E-PERM detectors have shown a sensitivity to
temperature. The surface potential of an electret will have a lower voltage
when the electret is cold. The lower reading, because of the different
expansion coefficients of the Teflon and electret holder material (10), will
be interpreted as a result of radon activity and result in a false positive.
The problem can be avoided if the reading of the electret (before and after
the test) is performed within the temperature range of 65*F to 85*F and
electrets are allowed at least 30 minutes to equilibrate to room
temperature. Temperature variations during the testing period and travel
will not affect the results.
Another potential problem that could misrepresent the measured radon
concentrations is improper handling of the electret during the reading.
Dust or dirt within the E-PERM chamber or from the air that contacts the
electret will result in a lowering of the surface potential. Touching the
electret will also lower the surface potential. The result would again be a
false positive. The imposition of proper handling procedures in a clean
environment will minimize the occurrence of false positives.
Even with special precautions in handling and reading of electrets, the
E-PERM can give false positives. Out of the 80 E-PERM tests conducted in
this study, one E-PERM gave a false positive result. EPA guidance states
that false positives are preferred to false negatives (11). It 1s unlikely
that the E-PERM system will give false negatives. False positives will be
identified when confirmatory follow-up measurements are made with the E-PERM
system.
Another benefit of the E-PERM system 1s that uncertainty due to varying
manufacturer's specifications is limited because the E-PERM system 1s
produced solely by Rad Elect, Inc. However, there are uncertainties
associated with variations In the materials used to construct the E-PERM
device. These variations are small and have not proven to contribute
significantly to errors. The results from Analysas' participation in
-------
Round 6 of the EPA Radon Measurement Proficiency Program had a standard
error of 7%.
EXPERIMENTAL PROCEDURE
E-PERM detectors and charcoal canisters were exposed to radon in a 17-
m Hot Pack walk-in chamber located at the U.S. Department of Energy's Oak
Ridge National Laboratory. Radon levels within the chamber are created from
the National Bureau of Standards Traceable Radium Sources. A filtered air
exchange, which is relatively radon free, is used to achieve a given radon
concentration. To create background levels, the radium sources were not
used, and only the filtered air supplied to the chamber. Radon
concentrations within the chamber were determined by the use of four Femto-
TECH Model RT310F Radon Trackers. The chamber was recently calibrated by
the EPA Radiological Health Branch in Montgomery, Alabama. Resulting
concentrations from the four Femto-TECHs were recorded by data loggers every
30 minutes. The results were then averaged over the time of the exposure to
arrive at a time weight average that was used as the actual radon
concentration of the chamber.
Three exposures were carried out in the chamber. The first exposure
was with E-PERMs exposed at background levels. Subsequent runs exposed E-
PERMs and charcoal canisters simultaneously. For each exposure, 20 of each
device were exposed. Radon levels were allowed to stabilize before the
radon detectors were placed. After exposure, the detectors were closed and
returned to Analysas. The surface potential of the electrets was measured
and radon concentrations calculated from manufacturer's calibration curves.
Charcoal canisters were returned to the vendor for analysis following the
vendor's specified handling protocols.
The results of the three E-PERM exposures were then used to verify the
manufacturer's calibration curves. Results from a simultaneous exposure
with E-PERMs and charcoal canisters were used for a comparison of the two
radon measurement methods.
RESULTS
The following section provides the results of the study. The data are
reported 1n two parts. The first gives the results from a check of the E-
PERM manufacturer's calibration curve, and the second compares the E-PERM
and charcoal canister methods.
QUALITY ASSURANCE STUDY OF MANUFACTURER'S CALIBRATION CURVE
The results of the E-PERM calibration study are shown in Figure 1. The
solid line represents a linear regression through data from 60 detectors
exposed at three radon concentrations. Interval estimates for this line
-------
(a - Q.05) are also shown. The broken line is the calibration curve
estimated by the manufacturer. Calibration factors were calculated by
assuming the Femto-TECH data as actual values and calculate back to obtain
the calibration factor.
3.5
3.0
s-
o
¦M
U
10
§ 2.5
J-
«3
o
2.0
1.5
•
•
•
• •
• »
•
• •
•
• • • • •• - —""
' -1
•
•
•
•
• »
• • • •• •
• • •
•
400
500 600 700
Midpoint voltage drop
800
Figure 1. E-Perm calibration curve.
Solid line is linear regression through E-PERM data (n - 60).
Dashed line is confidence interval (a » 0.05) about the regression
line. Broken line is manufacturer's calibration curve.
-------
Data used to generate Figure 1 were from E-PERM and Femto-TECH (shown
on Table 1) and data from a background exposure. The results of the
background exposure were 0.7 pCi/1 for the Femto-TECH and 0.83 pC1/l with a
standard deviation of 0.22 pCi/1 for the E-PERM monitors. A simultaneous
exposure of 300 alpha-track detectors was conducted at the background level,
with a resulting average radon concentration of 0.68 pCi/1.
COMPARISON OF CHARCOAL CANISTER
Simultaneous exposures were conducted at radon concentrations of
approximately 5 and 90 pCi/1. Table 1 summarizes the results from each run.
DISCUSSION
A correlation coefficient of 0.39 was calculated for the data plotted
in Figure 1. This weak correlation coefficient shows that the E-PERM device
responds linearly over the voltage range of the electrets used in the study.
The weak correlation coefficient also means the slope of the calibration
should be essentially zero. The slope of the manufacturer's calibration
curve is 6 x 10'4.
The quality assurance check of the calibration curve verifies the
manufacturer's reported values. The manufacturer's curve falls within the
95% interval estimate from the estimated calibration curve. The data agree
quite well even though a limited number of data points were generated from
the study.
Using the Femto-TECH data as actual radon levels, both E-PERM and
charcoal canister methods performed well for run 1. E-PERM detectors were
more precise (accomplishing a standard deviation of 0.35 compared to 1.25
for charcoal canisters), while charcoal canisters were more accurate. Of
the 20 charcoal canisters exposed at a radon level of 5.23 pC1/l, two
canisters reported a false negative. Values of 1.5 pC1/l were reported for
both false negatives.
For the second exposure, E-PERM and Femto-TECH data closely agreed with
one another. However, charcoal canisters reported values that, on the
average, were twice that of the assumed actual values. At this time the
cause of the error Is uncertain. The manufacturer was contacted, and
various reasons for the discrepancy explored. No apparent cause for the
error could be Identified.
As a quality check, an E-PERM was exposed 1n radon levels of 132 pC1/l
1n the off position. After a two-day exposure, no loss 1n surface potential
was measured.
The conclusion of the study was that the E-PERM device performed well,
closely matching the precision of the Femto-TECH device. The quality of the
charcoal canister's performance varied, with accurate readings at low radon
level and grossly overestimated radon levels at high levels. The next phase
-------
TABLE 1. RADON CONCENTRATIONS FOR SIMULTANEOUS EXPOSURES WITH CHARCOAL CANISTERS
AND E-PERM DETECTORS
E-PERM Charcoal canister Femto-TECH*
Rn concentration Standard Rn concentration Standard Rn concentration Standard
(pCi/1) deviation (pCi/1) deviation (pCi/11 deviation
Run 1 4.20 0.35 4.81 1.25 5.23 0.23
Run 2 66.65 3.12 133.12 7.25 63.62 3.01
*Used as the actual radon concentration.
-------
of the research is to conduct simultaneous exposures of E-PERM detectors and
charcoal canisters under varying radon concentrations. The varying radon
concentrations will attempt to mimic the natural diurnal variation usually
seen for indoor radon levels. Results from this study will be available in
the coming months.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the views
of the Agency and no official endorsement should be inferred.
-------
REFERENCES
1. Pearson, Mark D. Evaluation of the Performance Characteristics of
Radon and Radon-Daughter Concentration Measuring Devices Under
Controlled Environments. DOE/ID/12584-29. U.S. Department of Energy,
Grand Junction Projects Office, Grand Junction, Colorado, 1989.
2. Wilson, Owen J. Radon Transport in an Activated Charcoal Canister.
Nuclear Instruments and Methods in Phvsics Research. A275: 163-171,
1989.
3. Bocanegra, Rey and Hopke, Philip K. Radon Adsorption on Activated
Carbon and the Effects of Some Airborne Contaminants. The Science of
the Total Environment. 76: 193-202, 1988.
4. Gupta, P. C., Kotrappa, P., and Dua, S. K. Electret Personal
Dosimeter. Radiation Protection Dosimetry. 11: 107, 1985.
5. Pretzsch, G., Dorschel, B., and Leuschner, A. Investigation of Teflon
Electret Detectors for Gamma Dosimetry. Radiation Protection
Dosimetry. 12: 79, 1983.
6. Fallone, B. G. and Podgorsak, E. B. Electrostatic Fields in an
Electret Ionization Chamber. Journal of Applied Phvsics. 54: 4739,
1983.
7. Kotrappa, P., et al. Measurement of Potential Alpha Energy
Concentration of Radon and Thorium Daughters Using an Electret
Dosimeter. Radiation Protection Dosimetry. S: 49, 1984.
8. Kotrappa, P., et al. An Electret Passive 222Rn Monitor Based on
Ionization Measurement. Health Phvsics. 54: 1 (47-56), 1988.
9. Pretzsch, B., et al. Measurement of Tritium Activity Concentration in
Air by Means of the Electret Ionization Chamber. Radiation Protection
Dosimetry. 12: 345, 1986.
10. Rad Elect Technical Update Memo Number 7.
11. Ronca-Battista, M. and Magno, P. Standard Measurement Techniques and
Strategies for Indoor 222Rn Measurements. Health Phvsics. 55: 1 (67-
69), 1988.
-------
III-Pl-4
A RADON CHAMBER COMPARISON OF ALPHA TRACK
DETECTORS OVER A RANGE OF EXPOSURES
by:
William M. Yeager, Keith A. Daum, and Robert M. Lucas
Center for Environmental Quality Assurance
Research Triangle Institute
Research Triangle Park, North Carolina 27709
Lisa Feldt
Office of Radiation Programs
U.S. Environmental Protection Agency
Washington, DC 20460
Edwin Senslntaffar and Sam Poppell
Eastern Environmental Radiation Facility
U.S. Environmental Protection Agency
Montgomery, Alabama 36109
M1ke Clarkln
Camroden Associates
Orlskany, New York 13424
Three models of alpha track detectors (ATDs) were exposed 1n EPA radon
chambers to obtain estimates of precision and bias for the National Residen-
tial Radon Survey (NRRS). Exposures 1n this study ranged from 365 to 7300
pC1/L-day, plus blanks. These exposures correspond to annual average con-
centrations of 1 to 20 pC1/L, the range expected 1n most U.S. residences.
Ten ATDs of each model were studied at 12 exposures. The mean and
standard deviation of the reported values for each model were calculated and
compared with the continuously monitored chamber concentrations to determine
the bias and precision at each exposure. Results of this analysis were
discussed with the vendors, who took corrective actions. Changes 1n track
counting procedures and calibrations Improved detector performance.
This paper has been reviewed 1n accordance with the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
presentation and publication.
-------
INTRODUCTION
Alpha track detectors (ATDs) have been used for many years to measure
the concentration of2radon 1n soil gas, ambient air, and Indoor air (1).
ATOs are small (2 cm ) sheets of plastic or film which are damaged by
heavily Ionizing charged particles (e.g., alpha particles from the decay of
radon and Its progeny). After exposure, the damaged material Is preferen-
tially etched by a caustic solution, leaving a scar 1n the plastic which 1s
visible under an optical microscope. These scars or tracks may be counted
by a trained person or by a computerized Image analysis system. The density
(number/area) of tracks 1s proportional to the exposure to radon. Exposure
1s the product of time and concentration. Previous studies have shown that
track density Is linear over a wide range of durations and concentrations
(2,3,4). Given the duration of the exposure (days), the average concentra-
tion of radon (pC1/L) may be calculated. The sensitivity of an alpha track
detector depends on its volume and geometry, as well as on the character-
istics of the plastic and the etching process. Figure 1 Illustrates an ATD.
The sensitive plastic chip 1s enclosed 1n a cylindrical plastic housing,
which has several openings 1n one end to allow radon to enter. A filter
keeps out dust and radon progeny. Detectors are packaged 1n "radon-proof"
foil bags to prevent exposure before placement.
The National Residential Radon Survey (NRRS) 1s measuring annual
average radon concentrations In a statistical sample of about 5000 homes
around the United States. The primary goal of the Survey Is to estimate the
distribution of radon concentrations 1n homes 1n the U.S., particularly the
fractions of homes with annual average radon concentrations exceeding
threshold levels of 4.0 and 10.0 pC1/L. The Survey 1s designed to estimate
the fraction exceeding 10.0 pC1/L with an uncertainty, or relative standard
error, of 50% or less If the fraction of homes 1s at least .005.
Radon concentrations will be measured using ATDs exposed for 12 months
1n the homes. This study of alpha track detectors was performed to evaluate
which models could meet the precision and bias requirements of the NRRS.
Information from this study was used 1n the design of the subsample of homes
for the Survey which have co-located ATDs. The results of those 12-month
measurements will be used to estimate the precision of the Survey measure-
ments.
FACTORS AFFECTING THE QUALITY OF ATD MEASUREMENTS
ATDs are calibrated by exposure to known concentrations of radon. The
response factor determined by these exposures relates the track density to
the product of radon concentration with exposure duration. Track densities
are determined by dividing the number of alpha tracks by the area 1n which
they were counted. These counting statistics may be modeled using the
Polsson probability distribution (5). The random counting error is the
square root of the total count. The net track density Is the difference
-------
between the observed track density (determined by counting the number of
tracks per unit area) and the average track density (or "background"), found
on unexposed material Immediately after 1t 1s received:
Net track density * track density - background correction
This background represents surface defects from the manufacturing process,
not other radiation damage during the exposure period. Thus the background
correction 1s part of the calibration of the detectors; It 1s Independent of
exposure, but may vary with the age of the plastic.
Assuming the net track density 1s a linear function of radon exposure,
the exposures of ATDs may be calculated by multiplying the net track density
by the response of the plastic material to known exposures:
Exposure = (response factor)(net track density).
* R (trkden - B)
where R ¦ response
trkden = track density
B ¦ background
Now, given the duration of the exposure (days), the average concentration of
radon (pC1/L) can be calculated:
Average concentration ¦ exposure/exposure duration
At low exposures, counting statistics limit the precision of ATDs.
Only a small portion of the detector (10 to 50 mm ) 1s routinely analyzed,
but precision at these exposures may be Improved by counting a larger area.
Also, the accuracy of the background correction limits the bias of ATD
measurements at low exposures.
At high exposures, errors 1n the response factor limit the precision 1f
the counting error Is truly random. Relative precision should Improve as
the exposure Increases, asymptotically approaching the precision In the
response factor. The accuracy of the response factor limits the bias of ATD
measurements at high exposures.
DESIGN OF THE STUDY
Four sets of 140 commercially available models of ATDs were purchased
through Camroden Associates, a subcontractor which does a considerable
amount of radon measurement 1n NY and NJ. This blind purchase was arranged
so that the vendors would not recognize that they were being tested. After
delivery of the detectors, one vendor requested that their detectors be
returned because of unspecified quality control problems. That model was
therefore dropped from the study. Camroden shipped the rest of the detec-
tors to EPA's Eastern Environmental Radiation Facility (EERF) 1n Montgomery,
AL, for exposure to known concentrations of radon. There were two radon
-------
3
chambers at the EERF. The larger (42m ) walk-1n chamber was operated at a
nominal concentration of 36.5 pC1/L; the smaller (3.6m ) chamber, at 365
pC1/L. The study lasted for 20 days 1n each chamber. The radon concentra-
tion in each chamber was continuously monitored to assure a reasonably con-
stant level. Grab samples of chamber air were taken dally to calibrate the
continuous monitors.
Each detector was Identified by a manufacturer's serial number, which
was permanently marked on the sensitive plastic chip, on the outside of the
detector housing, and on the "radon-proof" foil bag In which each detector
was packaged.
The detectors were randomly divided Into 14 groups of 30. All 30
detectors 1n each group (10 of each model) received the same exposure. The
nominal exposures ranged from 365 pCI/L-d, simulating an annual average of l
pC1/L, to 7300 pC1/L-d, simulating an annual average of 20 pC1/L. Table 1
shows the chamber concentrations and exposure times. Note that the nominal
chamber concentrations were chosen so that the detectors were always removed
from the chamber at the same time of day.
There were two groups of blank detectors. Group 1 remained outside the
radon chamber. Immediately before these detectors were returned to their
vendors for analysis, their foil bags were opened and they were packaged for
shipment to Camroden 1n the same fashion as the exposed detectors. This
procedure checked for additional exposure in transit after the detectors
were removed from the chambers. Group 3 detectors were left 1n their foil
bags and placed 1n the large chamber. After 20 days, these detectors were
removed from the chamber, the bags were opened, and the detectors were
Immediately packaged for shipment to Camroden. This procedure checked
whether the foil bags completely prevented radon exposure before they were
opened.
EXECUTION OF THE STUDY
At the EERF, the detectors were removed from their foil bags and placed
1n numbered wire trays. Care was taken that the serial numbers of the
detectors 1n a tray were not sequential; again, we hoped to prevent the
vendors from recognizing that there had been anything unusual 1n the place-
ment.
Radon concentrations 1n each chamber were continuously monitored by a-
partlcle counters. In addition, grab samples of chamber air were taken
dally 1n 125-mL scintillation cells, which were held for 4 hours before
counting. The nominal concentration 1n the small chamber was 365 pC1/L; the
average monitored concentration was 366.5 pC1/L with a variation In the
counting rate corresponding to +9.0 pC1/L. The nominal concentration 1n the
large chamber was 36.5 pC1/L; tKe average monitored concentration was 36.1
pCI/L with a variation 1n the counting rate corresponding to +0.9 pC1/L.
The total error (counting plus calibration) In the monitored radon concen-
-------
tratlon for each chamber 1s estimated to be less than 5%. The monitored
exposure for each group of detectors was calculated from the hourly averages
during the exposure period of that group.
EERF personnel removed the detectors from the chambers according to the
schedule 1n Table 1. The detectors were allowed 1 hour to equH1br1ate to
the low radon concentration 1n room air before they were packaged for ship-
ment to Camroden Associates. During this time, the serial numbers of the
detectors were recorded on a data sheet. The 10 detectors of each model 1n
each group were wrapped together 1n two layers of aluminum foil and marked
with the group number.
Camroden personnel unwrapped the detectors and filled 1n dummy start
and stop dates on the labels attached to each detector. These dates Indi-
cated an exposure of at least 2 weeks, the minimum period for which detec-
tors might reasonably be exposed 1n homes suspected to have high radon con-
centrations. Detectors were then Individually wrapped 1n foil and returned
to their vendors for analysis. Exposure reports were received by Camroden
and sent to RTI for evaluation.
RESULTS
Table 2 gives summary statistics of the reported exposures for the
blank detectors. Group 1 detectors remained 1n their foil bags outside the
radon chamber until Immediately before shipment back to Camroden. Group 3
detectors remained 1n their bags Inside the large chamber (36.5 pC1/L) for
20 days. They were removed from the chamber and then from their bags
Immediately before shipment back to Camroden. The vendors of detectors A
and C did not report quantitative values <30 pC1/L-d. Th1rty-f1ve percent
of the model A blanks exceeded this detection limit, while only 15% of model
C blanks did so. All reported readings for model B blanks exceeded 30
pC1/L-d. It appears that the "radon-proof" bags of detector B allowed some
exposure before receipt and especially while Group 3 was Inside the radon
chamber. The bags for detector B reduced the radon exposures for Group 3 by
approximately 25% compared to the Group 4 detectors, which were 1n the same
chamber for the same period.
Table 3 shows the nominal and monitored chamber exposures, summary
statistics of the reported readings, and the recovery (reported/monitored)
for four representative exposure groups. Data for all 12 exposure groups
are Included 1n Figures 2-6. Minimum and maximum values which differ by
more than a factor of 2 may Indicate an error by the laboratory which pro-
cessed the detectors. Other than checking the serial numbers used to match
actual with reported exposures, no attempt was made to resolve these out-
liers. The laboratories were not asked to confirm any reported values.
Figure 2 shows the relative standard deviation (RSD ¦ standard
deviation/mean) versus the monitored exposure for each group of detectors.
In general, detector C had the best precision; detector B, the worst. Large
increases 1n the RSD over a small range of exposures may be due to a transl-
-------
t1 on from automatic to manual counting. As track density Increases, so does
the probability of tracks overlapping. This leads to undercountlng, because
automatic counting equipment has difficulty resolving overlapping tracks.
The human eye, however, can resolve Individual tracks at much higher
densities. Therefore, detectors exposed above some transition point must be
manually counted. Precision decreases after this transition for two
reasons: 1) a smaller area Is counted, and 2) people have difficulty repro-
duclbly counting 50 or 100 things, especially under a microscope. The
vendor of detector A advised us that their transition was at about 1500
pC1/L-d, where we see the RSD Increased from 12 to 52%. We speculated that
the transition to manual counting for detector B was about 3000 pC1/L~d,
where the RSD Increased from 7 to 17%, but later determined this was not the
case. There 1s no apparent transition point for detector C 1n Figure 2.
Figure 3 shows the recovery vs. the monitored exposure for each model
of ATD. An Ideal detector would have a recovery of 1.0 with a small error
bar. Detector A was closest to the monitored value, and therefore was
selected for use 1n the NRRS, although 1t was biased high 1n the middle
range of exposures. Detectors B and C were biased low; detector B had the
greatest bias and the worst precision.
FOLLOW-UP
In May 1989, RTI and EPA personnel met with representatives of the
vendors of models A, B, and C to discuss their performance 1n this study and
what could be done to Improve 1t. The detectors used 1n the study were
Identified, but their Individual exposures were not.
In June 1989, we visited the facility where model A detectors were
manufactured and analyzed. The vendor explained that they had two computer-
ized image analysis systems: "w1de-f1eld" and "conventional." The first
scanned a larger area than the latter, but could not resolve Individual
tracks at exposures above 1500 pC1/L-d. The detectors 1n our study had been
counted by this system, with a transition to manual counting at 1500
pCI/L-d. This resulted In a large decrease In precision, observed between
exposure groups 6 and 7. The vendor had recounted all of their detectors
used 1n our study. Instead of making the transition to manual counting at
1500 pCI/L-d, they had changed to their conventional Image analysis system,
which could resolve tracks up to about 4500 pC1/L-d. At that point, they
made the transition to manual counting. There was no significant difference
1n the reported exposure of the detectors which were recounted 1n the same
way as originally counted, but the detectors which were first manually
counted and then recounted with the conventional Image analysis system
averaged 45% less exposure on the recount. After receiving the recount
data, RTI provided the vendor wtth the monitored exposures for the Indivi-
dual detectors used In the study.
-------
Figure 4 compares the first and second values reported for detector A.
The average precision for detectors exposed between 1500 and 4500 pC1/L-d
Improved from 26% to 10%. Overall, the average precision (RSD) Improved
from 20% to 15%, while the average bias changed from +8% to -9%.
In June 1989, we also visited the vendor of detector B. This vendor
explained that they did not use an Image analysis system; all detectors were
manually counted. Thus, the deterioration 1n precision which we had
observed at around 3000 pC1/L-d for model B detectors was not due to a
change 1n counting procedure. We then hypothesized that 1t might be due to
a difference between the two persons who counted the detectors, but 1t
turned out that all of these detectors had been counted by the same Indivi-
dual. This vendor agreed to recount all of their detectors used 1n our
study. After receiving the recount data, RTI provided the vendor with the
monitored exposures for the Individual detectors. In addition to recount-
ing, the second set of reported values for detector B reflected changes 1n
both the response factor and the background correction. The vendor also
advised us that there had been two typographical errors 1n the first report
that affected the reported values, and that two more detectors were damaged
and should not have been reported.
Figure 5 compares the first and second values reported for detector B.
Average bias Improved from -35% to +6%; average precision (RSD) from 2435 to
19%. The bias of detector B 1s now less than either of the other two detec-
tor models Included 1n this study; however, the precision 1s still the
poorest of the three, especially for low exposures « 1800 pC1/L-d or
5 pC1/L-yr).
After our site visit, vendor B began using an Image analysis system to
count their ATDs. Vendor B now uses this system to count all detectors.
Because they spanned a wide range of exposures, detectors from our perform-
ance evaluation study were used to calibrate this system.
Table 4 shows the reported exposures of blank detectors determined from
recounts. These may be compared with the values in Table 2. The reported
exposures for model B detectors 1n group 3 are even higher than before.
This vendor now reports that problems with sealing the foil bags 1n which
the detectors are packaged have been solved.
Table 5 shows the reported exposures of non-blank detectors determined
from recounts. These may be compared with the values In Table 3. The over-
all average recovery for all 12 groups of exposed detectors 1s 0.91 + .14
for model A, 1.06 + .19 for model B.
REGRESSION ANALYSIS
If the reported values of spiked detectors are precise, but biased, a
linear regression of the monitored values on the reported values may be
performed to check the calibration of the detectors. Table 6 shows the
-------
results of such a regression (based on recounts for models A and B). A
least squares fit was used to determine the coefficients S and I of the
equation
Monitored = S*Reported + I
Ideally, monitored = reported or S = 1.0 and I = 0.0. A value of S signi-
ficantly different from unity Indicates that the response factor 1s off. It
appears that the response factor is low for detector C and high for detec-
tors A and B. This may be related to the fact that the vendors' calibration
exposures are usually less than 500 pC1/L-d. Calibration exposures should
span the anticipated range of field exposures. On the other hand, a value
of I significantly different from zero Indicates that the background correc-
tion 1s not accurate. It appears that the background correction 1s too
large for model A and too small for model C, but about right for model B.
This correction 1s based on the observed track density of freshly manu-
factured plastic or film, not on counts of completely assembled detectors
which have been distributed throughout the country, sealed, and mailed back
to the vendor. Note that at low exposures, the errors 1n the response
factors and background corrections for detectors A and C tend to cancel out.
Tables 7 and 8 summarize the interpretation of the regression coefficients
1n general and for the specific models In this study.
These regression coefficients may be used to correct the calibration
coefficients:
Monitored = S*Reported + I = S*R*(trkden - B) + I
= R'*(trkden - B')
where R' = S*R
B' = B - 1/(S*R).
Alternatively, the regression coefficients may be used to adjust the
reported values:
Adjusted = S*Reported + I.
Figure 6 shows the results of such an adjustment for model C detectors. The
model C detector was designed for non-res1dent1al applications. Although 1t
had the best precision of the three models In our study, 1t showed signifi-
cant bias. These two characteristics made 1t an Ideal candidate for adjust-
ing the reported values based on the regression described above. Overall(
the average recovery Improved from 0.742 * .068 to 1.009 * .099; the mean
absolute relative error (MARE), from 0.261 to 0.078. Results of such an
adjustment for the other models, which had worse precision and possible non-
calibration problems, were less Impressive.
-------
CONCLUSIONS
Performance of measurement methods should be evaluated over the range
of exposures anticipated, before field placement. The importance of various
measures of performance depends on the Intended use of the measurement. The
decision to mitigate an Individual home will be based on a few measurements;
each one is important. It 1s not necessary, however, to be very accurate at
high or low exposures. One would like to maximize measurement quality at
the exposure corresponding to the decision of whether or not to mitigate the
structure. The EPA guideline for mitigation 1s an annual average radon
concentration of 4 pC1/L, or a 12-month exposure of 1460 pCI/L-d.
On the other hand, 1n a survey to determine the frequency distribution
of radon concentrations, 1t 1s Important to know the precision and bias over
the entire range of exposures, so that the distribution can be adjusted on a
statistical basis. For the NRRS, performance will be monitored by the use
of blank, collocated, and spiked detectors which will be mixed with field
exposed detectors returned for analysis.
In either case, the performance of alpha track detectors on a well-
designed set of known exposures may be the basis of discussions with the
vendor about suspected problems with calibration, processing, or counting.
In some cases, calibration coefficients may be adjusted on the basis of a
regression analysis of spiked detectors.
REFERENCES
1. Lovett, D.B. Track Etch Detectors for Alpha Exposure Estimates.
Health Physics 16: 623, 1969.
2. Alter, H.W. and R.L. Fleischer. Passive Integrating Radon Monitor for
Environmental Monitoring. Health Physics 40: 693, 1981.
3. Savage, E.D. Evaluation of Track-Etch Detectors. EPA-520/5-83-010,
U.S. Environmental Protection Agency, Washington, D.C., 1983.
4. George, J.L. and 6.H. Langner. Field Study of Indoor Average Radon-
Daughter Estimation Methods. GJ/TMC-26 UC-70A, U.S. Department of
Energy Report, 1986. p. 54.
5. Bevlngton, P.R. Data Reduction and Error Analysis for the Physical
Sciences. McGraw H111: New York, 1969. p. 36.
-------
0.6
0.4
Q
If)
CC
c
a
(3
'>
-------
* i.s
8
cc
1.0
o.s
i
f. If1
¦ Report 1
a Report 2 ( rec
fi]
i
ounted)
[I I f
ifli
)
iir
i
il I 1
l l
1SOO 3000 4500
Monitored Exposure ( p cl/L-d )
6000
7500
Figure 4. First and second values reported for Model A.
0.0
1500 3000 4800
Monitored Exposure (p cl/L-d)
6000
7500
Figure 6. Reported and adjusted values for Model C.
-------
TABLE 1. NOMINAL CHAMBER CONCENTRATIONS AND EXPOSURE TIMES
Chamb*r
Chamber
Group
Planned Exposur*
Cone.
Tim*
(oCl/L-yr}
(pCt/L-d)
foCl/Ll
(day*)
1
0
0
0
(unop*n*d)
0
2
1
365
36.5
10
3
0
0
36.5
(unop*n*d)
20
4
2
730
36.5
20
5
3
1095
365
3
6
4
1460
365
4
7
5
1825
365
5
9
8
2920
365
8
9
9
3285
365
9
10
10
3650
365
10
11
11
4015
365
11
12
12
4380
365
12
13
15
6475
865
15
14
20
7300
366
20
TABLE 2. REPORTEO RADON EXPOSURES FOR BLANK DETECTORS
Reported'*
Percent!lea
Group*
Modal
Nomlna1
N>30
Mad 1 an
90th
10th
1
A
0
4
24
56
9
B
0
10
64
118
48
C
0
2
<30
43
<30
3
A
0
3
15
52
5
B
0
10
857
403
218
C
0
1
<30
33
<30
* Group 1 deteetora remained In foil bags out*l 80 pCI/L-d were reported for modal C.
TABLE 8. REPORTED RADON EXPOSURES
fi£|SU> M
4^
10
14
B
C
A
B
C
A
B
C
A
B
C
Monitored Exposur*
3.9
10.0
20.1
1432
8640
7380
Reported Exposur*
HT IB i!h»
292
819
1372
1147
1097
4494
2020
2566
7666
4460
5098
802
831
1895
1115
1088
4870
1940
2669
7887
4407
502
T.
.211
0.128
0.118
0.169
0.124
0.151
0.342
0.076
0.124
0.086
0.036
(pCI/L-d)
vm 18"
420 250
410 290
1590 1216
1320 750
1320 690
5160 2990
2820 510
2840 2260
8870 5570
4890 8850
5230 4670
• Raeovary • (man reported ~ etd.dev.)/men I tored
1*PK»
0.17
0.1*
0.12
0.83
0.91
0.97
0.78
0.76
1.20
0.53
0.70
1.01
0.60
0.69
0.11
0.09
0.18
0.18
0.08
0.18
0.08
0.03
/
-------
TABLE 4. RECOUNTED RADON EXPOSURES FOR BLANK DETECTORS
Reported'3
Percentiles
GrouD
Model
Nomlnal
N>30
Median
90th 10th
1
A
0
3
23
74 1
B
0
10
47
109 19
3
A
0
4
25
82 11
B
0
10
420
677 232
* Group X datactora ramalnad In foil baga outalda tha radon chambara.
Group 8 dataetora ramalnad in foil baga inalda a radon chambar for
*22 pCl/L-d.
Only valuaa > 80 pCl/L-d war* raportad for modal C.
TABLE 5. RECOUNTED RADON EXPOSURES
SrSJJ* Modal
Monitorad Expoaura
fDCI/L-vri (aCI/L-d)
Raportad Expaaura
Madtan Maan RSD
(pCl/L-d)
90th 10th
Raeovary*
2
A
1.0
864
860
807
0.281
668 200
1.06 + 0.26
B
890
404
0.260
670 200
1.11 + 0.29
e
A
8.0
1482
1212
1881
0.167
1601 1060
0.98 +0.16
B
1068
1047
0.188
2180 1400
1.20 ± 0.17
la
A
10.0
8646
8606
8178
0.100
8409 2628
0.07 ~ 0.09
B
8816
8744
0.166
4900 2920
1.08 ~ 0«17
14
A
20.1
7880
7190
0042
0.276
12276 6268
1.10 ~ 0.80
B
7786
7672
0.160
9620 6770
1.06 ~ 0.17
Raeavary * (maan raportad ~ aid.dav.)/men 1 torad
-------
TABLE 6. REGRESSION COEFFICIENTS OF REPORTED VERSUS
MONITORED EXPOSURES (MONITORED = S*REPORTED + I)
Model S I R-square
A2 0.851 + .033 690.7 + 115.2 0.8505
B2 0.916 + .026 149.8 + 99.1 0.9154
C 1.456 + .017 -107.1 + 42.9 0.9845
TABLE 7. GENERAL INTERPRETATION OF REGRESSION COEFFICIENTS
MONITORED = S*REPORTED + I
R I
small S > 1.0 I < 0.0
large S < 1.0 I > 0.0
TABLE 8. SPECIFIC INTERPRETATION OF REGRESSION COEFFICIENTS
IN TABLE 6
MONITORED - S*REPORTED + I
Model S --> R
A2 <1.0 large
B2 <1.0 large
C >1.0 small
_J 1Z> I
> 0.0 large
>0.0 satisfactory
< 0.0 small
Comments
cancel at low exposures
cancel at low exposures
-------
III-P1-5
A STUDY OF BATCH CALIBRATIONS ON h" OPEN FACED CHARCOAL ADSORBERS FROM FOUR
DIFFERENT MANUFACTURERS AND HOW THEY COMPARE TO EERF'S PUBLISHED CALIBRATION
CURVES
by: Dallas L. Jones and T. E. Howell
Radon Reduction & Testing, Inc.
1587 Northeast Expressway
Suite 175
Atlanta, Georgia 30329
ABSTRACT
The EERF Standard Operating Procedures for Rn-222 Measurements Using
Charcoal Canisters is quite explicit in stating that the calibration tables
and curves published in this manual are only "typical calibration data that
were developed for a particular batch of canisters from a supplier," and that
"new calibration data are developed and used with each different lot of
canisters."
In this study, batch calibrations were performed in our radon chambers on
so called "EPA style" canisters from four different companies. We then
prepared calibration tables and curves for each batch and compared them to the
EERF published tables. It was our premise that EPA never intended for their
curves to be considered "standard" and that with more and more companies
manufacturing canisters, the calibration factors could certainly vary and
might significantly differ from those derived by EERF several years ago.
-------
INTRODUCTION
During the past year of promoting our calibration chamber services, it
became apparent to us that many labs analyzing 4" open-faced cans are blindly
using the calibration tables and curves programed into the software supplied
with their instrumentation, without ever developing calibration curves of
their own. In most cases, these curves were taken directly from the EERF
publication. We were concerned that companies may be placing a false sense
of security in calibration curves that may not apply to their particular
canisters.
PRELIMINARY QUALITY ASSURANCE CHECKS
Prior to beginning our study, we chose to perform quality assurance
checks on our calibration chambers by performing a current chamber mapping and
an intercalibration with the EERF chamber in Montgomery. Radon Reduction &
Testing chambers were designed to recirculate the radon-laden air to eliminate
stratification of heavier radon gas to the bottom. We must be conscious
however, that air flows are not excessive enough to cause over response by the
charcoal adsorbers. (2)
CHAMBER MAPPING
To doubly insure that the radon concentrations within our calibration
chambers is being evenly distributed and that air currents are not adversely
affecting absorption, charcoal cartridges were placed on every shelf in each
coded "bin" location, approximately 10 inches apart. The average radon
concentration in Chamber 1 for this 24 hour mapping exposure was 33.3 pCi/1;
the average radon concentration in Chamber 2 during the mapping exposure was
5.8 pCi/1.
Comparing calibration factors calculated for each detector exposed
Chamber 1, we determined a standard deviation of less than 4%. Comparing
calibration factors calculated for each detector exposed in Chamber 2, we
determined a standard deviation of less than 8%. These differences are well
within the standard deviations between charcoal detectors themselves and
indicated to us that the chamber environment was consistent throughout. The
slightly higher deviation between detectors in Chamber 2 were attributable to
the low radon concentration in that chamber during mapping exposures.
(Precision between canisters exposed at 6 pci/1 would generally be lower than
precision between canisters exposed at 33 pCi/1.) Another mapping in chamber 2
upon project completion at a higher radon concentration proved this to be the
case. At 30 pCi/1 the standard deviation between cans was slightly greater
than 3%.
-------
INTERCALIBRATION
On August 15, 1989 two Eberline RGM-3 continuous gas monitors used as
back-ups for continuous chamber monitoring and for performing daily grab
sampling to cross check the full time chamber monitors, were taken to
Montgomery and intercalibrated with the EERF radon chambers. Comparing these
monitors with our full time chamber monitors indicated agreement within 3% and
no need for a calibration adjustment.
BATCH SELECTIONS
The charcoal canister batches used in this study were graciously supplied
by four different laboratories or suppliers. Each batch exposed and analyzed
will heretofore be referred to as Company Red, Company Green, Company Yellow
and Company Pink; the laboratories suppling the canisters will not be
identified by name. All canisters were 4" diameter open faced, described as
EPA style canisters and supposedly assembled to EPA specifications described
in the EERF publication. (1)
Upon receiving the cans, we randomly selected 5 from each batch of 100 to
be analyzed for any differences and to determine the initial carbon weight and
moisture content for an average can from each batch. Although all of the
canisters were described by the laboratories providing them as EPA style
canisters, analysis revealed certain obvious differences. Initial moisture
contents of the carbon in each batch varied, even though all were within the
EPA specification of less than 4%. The initial carbon weight of cans from
Company Pink averaged 76.1 grams. (Remember EPA specs of 70 grams plus or
minus 1 gram.) The average initial carbon weight and moisture content of the
charcoal in each batch was determined to be the following:
Co. RED
Co. GREEN
Co. PINK
Co. YELLOW
69.75g initial carbon wt.
69.54g initial carbon wt.
76.10g initial carbon wt.
70.66g initial carbon wt.
2.8% initial moisture content
1.8% initial moisture content
2.4% initial moisture content
1.9% initial moisture content.
Of the 5 cans analyzed for each batch, initial moisture content was more
inconsistent in those provided by Company Green and Company Yellow. Company
Red's RECORDED initial weights were consistently 0.3 to 0.4 grams lighter than
they were at the time we weighed them. It was assumed that the batch had
picked up the extra moisture (through the tape) while being stored prior to
our receiving them. Conversations with several canister manufacturers
confirmed our belief that they can indeed gain several tenths of grams on the
shelf if the storage environment is humid enough. One company reported he had
cans that had gained nearly 0.7 grams while being stored for nearly a year.
Examination of the carbon in these cans revealed differences as well.
Company Red and Company Green appeared to be Calgon carbon. Company Yellow
-------
was much less dusty than typical Calgon carbon, but was otherwise the same in
appearance. Company Pink carbon was different in density and appearance and
was apparently not Calgon. (An anonymous telephone conversation with an
unidentified representative from Company Pink did confirm that it was not
Calgon carbon, but he assured me that the "adjustment factor curves are the
same as published in the EERF Standard Operation Procedures Manual... the
same as our Calgon Cans.")
PREPARATION & EXPOSURE SEQUENCE
After selecting 5 cans from each batch of 100 for determining initial
carbon weight and moisture content, and selecting a background can from each
batch, 90 cans from each batch were weighed and initial weights (to 0.1 gram
accuracy) and serial numbers were logged on chamber data sheets. Chamber 1
was brought to 80% RH, Chamber 2 to 50% RH. Chamber 2 was adjusted to 20% Rft
the following week for the low humidity run. Radon concentration for the 6
days averaged 47 pCi/1 for the 80% run, 29 pCi/1 for the 50% run, and 25 pCi/^
for the 20% run. During exposure periods, continuous chamber monitors were
checked against daily grab radon samples. Chamber temperatures were
maintained between 72 and 74 degrees F.
When radon concentrations and humidity in the chamber were stabilized at
the desired conditions, a set of 30 canisters from each company were put into
the chamber and exposed in the following manner; a group of five each for
1,2,3,4,5, and 6 days, respectively. The canisters were opened inside the
chamber and the exact start times were recorded. This sequence was repeated
at each of the three humidities, to complete our calibration run.
REMOVAL AND COUNTING
At the end of each 1 day period specified above, a group of 5 canisters
were closed & taped inside the chamber and exact time was recorded. At least
3 hours were waited prior to counting to allow progeny to equilibrate with the
radon.
After removal from the chambers, each canister was weighed again,
recorded, and subsequently counted on our counting
system. Background canisters from each batch and our "standard" can were
counted prior to each counting period to determine our detector efficiency.
CALCULATION OF CALIBRATION FACTORS
A calibration factor was calculated for each canister using the following
equation published in the EERF Standard Operating Procedures for Radon-222
Measurement Using Charcoal Canisters:
-------
net cpm
CF » (Ts)(E)(Rn)(DF)
CF * Calibration factor, radon absorption rate (1/m), net
Net CPM » Gross counts per minute for the canister minus the background
CPM for the detector for that day,
Ts = Exposure time of the canister in the chamber (minutes),
Rn = The Radon concentration in the chamber for the exposure period
(pCi/1),
DF » Decay factor from the midpoint of exposure to the time of counting,
which is calculated from
• 693t
DF = e- T 1/2 Rn(tnin)
t * time in minutes from midpoint of exposure to the start of counting
and T 1/2 Rn(min) is the half-life of radon 222 (3.82 days) in
minutes. (1)
GENERATION OF CALIBRATION CURVES
Calibration factors were derived from the equation above to generate two
tables and two curves for each batch of canisters studied. One table relates
calibration factors to weight gain (water) for a two day exposure (the optimum
exposure time recommended by EPA for the EPA style 4" open face can). Data in
these tables were used to generate the two day weight gain vs. calibration
factor curves. NOTE: The weight gain curves generated for this study utilize
only the known data points from the 2-day calibration exposures and were not
extrapolated beyond known weight gains for two day exposures at 80% RH.
The second table relates exposure time to adjustment factors for the 20%,
50% and 80% humidity. The adjustment factors are used to modify the
calibration factors for exposure times different from the desired 2 day
exposure time. If a can is exposed for exactly 2 days, the adjustment factor
is 1. (1)
RESULTS
The first and most obvious discovery made upon completing the calibration
tables for all four batches was the Company Pink curves plotted considerably
below the EERF curves and those for the other three companies. Calibration
factors ran as much as 30% lower than EERF's at 20% RH. In fact, Company
Pink's 20% RH curve runs similar to EERF's published 50% RH curve. This is
not surprising when we consider that Company Pink's carbon was definitely not
Calgon, although the manufacturer of this can told us the calibration curves
would be " the same as his Calgon cans..., the same as the EERF curve." The
different brand of charcoal made a tremendous difference in the calibration.
The carbon used is not as sensitive as Calgon and is much more dense.
-------
Remember it took approximately 76 grams to fill up the can. The difference in
the carbon does not make it a bad can, however. The precision among detectors
was as good as the other three companies and our initial inspection of the
cans found them to be quite consistent in initial weight and moisture content-
With the correct calibration tables these cans perform admirably.
Curves for the other three companies ran surprisingly close to the EERF
curves through three days. This we found somewhat surprising because of the
variance in the initial moisture contents.
Deviations from EERF published curves were greatest beyond 72 hours for
Companies Red, Green, and Yellow. Calibration factors ran LOWER than EERF
published curves at 20% RH and HIGHER than EERF's published curves at 80% RH.
In other words beyond three day exposures the cans were somewhat less
efficient than EERF's at 20% RH and somewhat more efficient than EERF's at 80^:
RH. These differences were enough to cause accuracy problems beyond 3 days if
your software is utilizing the EERF tables. Using EERF data, these companies
would be reporting much too high at 4-6 days at 80% RH exposures and much to
low at 4-6 day 20% RH exposures.
A conversation with a representative of Company Yellow demonstrates this
well. Upon re-analyzing the canisters we exposed from her company, her system
utilizing EERF published curves computed a radon concentration for a can
exposed for 6 days at 80% RH to be 88 pCi/1. The actual exposure
concentration was 47.7 pCi/1. This is an error of over 80%. Upon changing
her adjustment factor for 6 days at 80% from .018 to her new AF of .027 she
read 58 pCi/1; much closer to the know concentration than before, but still
not 47,7 pCi/1.
The remaining difference can be explained when one closely examines the
EERF published 2 day weight gain curve. Our Company Yellow technician had not
adjusted this curve to her new data and that certainly would account for some
of her error. But the real problem lies in the extrapolated slope of the two
day curve beyond the known weight gains at 80% RH for 2 days.
Assuming you had a batch of cans that matched the EERF data perfectly,
you would know that a can exposed for 6 days at 80% RH and gaining 18 grams,
would have a CF of .018. This is a known CF determined in the calibration ru"
and is printed in the table. (1) But using EERF's tables and curves and the
formula for computing a final calibration factor from them, you would end up
with a final CF of .011, NOT .018. This is because the initial CF pulled frofl1
the 2 day weight gain table is .049. In order to compute to correct
concentration, the initial CF would have to be .078. As the EERF weight gain
curve slopes below .078 to .049 at 18 grams, the final CF drifts farther and
farther apart from the known CF for those cans exposed for 3-6 days at 80% RH
in the calibration run.
Simple logic would lead us to believe that herein lies the answer. If
the initial CF never drifted below .078, we ewould always end up with a final
CF equivalent to our known calibration factors. If we gained more weight tha®
what our calibration cans did at 3-6 days at 80% RH, we would know that they
-------
were exposed to a much greater humidity than 80% and we would simply throw
them out.
But while this solves one problem, it creates another by not allowing for
adjustment of CFs for those cans that perhaps gained only 10 grams in 6 days.
Such a can was exposed to something less than 80? RH and would have a CF of
s°oewhat higher than a can exposed for 6 days gaining 18 grams. If the weight
gain curve becomes a straight line after the maximum 2 day weight gain at 80%
the can gaining 10 grams in 6 days would have the same CF as the can
gaining 18 grams in 6 days.
We can only assume that this dilemma is one of the reasons EPA set their
°Ptimum exposure time at 2 days for the 4" open face cans.
CONCLUSIONS
Even though three of the four batches of EPk style canisters studied
followed the EERF published curves reasonably well through 3 days, the curves
were all somewhat different. Beyond 3 day, they were extremely different. We
certainly believe our study supports EERF's recommendation that calibration
curves be generated by every laboratory for THEIR detectors and that these be
checked regularly with exposures to known concentration.
The dilemma regarding the 2-day weight gain vs. CF curve vs. troublesome
and extremely important when analyzing cans with exposure times greater than 2
to 3 days. The curve in its present from "misses the mark" on KNOWN exposures
80* RH, even with the EERF's own data, for exposures beyond 3 days.
Changing it to a constant beyond the maximum weight gain for 2 day calibration
e*posures at 80% RH, would eliminate the error for 3 to 6 day known exposures
80% RH, but would no adjust for the can that gained only 10 grams at 6 days
gather than the 16 to 18 grams we know it would have gained at 80% RH.
£erhaps the simplest solution would be to discard any measurement made for
1)eyond 3 days and stay as close to the optimum 2 day exposure as possible.
The work described in this paper was not funded by the U.S.
¦tovironmental Protection Agency and therefore the contents do not necessarily
Reflect the views of the Agency and no official endorsement should be
xnferred.
REFERENCES
Gray, D.J. and Windham, S.T., EERF Standard Operating Procedures for
Radon-222 Measurement Using Charcoal Canisters, EPA 520/5-87-005, U.S.
Environmental Protection Agency, Montgomery, Alabama, 1987, 29 pp.
Gray, D.J. and Windham, S.T., The Overesponse of Open Faced Charcoal
adsorbers Used For Measurements Of Indoor Radon Concentrations, U.S.
Environmental Protection Agency, Montgomery, Alabama.
-------
TABLE 1. EXPOSURE TIME VSRSU8 ADJUSTMENT FACTORS FOR LOW, MEDIUM, AND HIGH HUMIDITY
COMPANY PINK VERSUS SERF PUBLISHED
XPTIMK
CO PINK
EERF PUd
CO PINK
BERF PUB
CO PINK
EERF PUB
LON RH
LOW RH
MED RH
MED RH
HIGH RH
HIGH RH
24
0.126
0.137
0.117
0.132
0.105
0.116
24
0.122
0.143
0.108
0.137
0.105
0.125
24
0.130
0.141
0.109
0.132
0.105
0.118
24
0.126
0.135
0.106
0.126
0.105
0.117
24
0.121
0.138
0.121
0.127
0.101
0.118
48
0.092
0.107
0.082
0.096
0.065
0.077
48
0.095
0.110
0.080
0.102
0.066
0.082
48
0.090
0.105
0.081
0.097
0.067
0.076
48
0.093
0.101
0.082
0.094
0.067
0.076
48
0.091
0.105
0.084
0.098
0.067
0.078
72
0.075
0.087
0.063
0.075
0.048
0.048
72
0.072
0.091
0.061
0.079
0.048
0.051
72
0.070
0.088
0.063
0.075
0.046
0.051
72
0.070
0.083
0.062
0.073
0.049
0.046
72
0.072
0.085
0.061
0.075
0.047
0.049
96
0.057
0.074
0.053
0.058
0.035
0.035
96
0.057
0.080
0.051
0.062
0.036
0.034
96
0.058
0.075
0.051
0.059
0.036
0.033
96
0.060
0.074
0.054
0.057
0.036
0.033
96
0.060
0.079
0.054
0.060
0.037
0.034
120
0.050
0.070
0.045
0.051
0.030
0.023
120
0.052
0.073
0.045
0.054
0.028
0.025
120
0.052
0.071
0.043
0.0S1
0.029
0.024
120
0.052
0.069
0.045
O.OSO
0.030
0.023
120
0.053
0.071
0.045
0.052
0.029
0.023
144
0.049
0.064
0.039
0.045
0.022
0.018
144
0.048
0.068
0.039
0.047
0.022
0.019
144
0.049
0.064
0.038
0.047
0.022
0.016
144
0.051
0.062
0.038
0.044
0.018
0.018
144
0.048
0.039
0.025
TABLE 2. UKIONT MIN VtMU* CMLXMNT10N FACTONS FOR A 8 DAY RUN
CBWflWY. PIWK VKWUB m* nWI TTT
BBMPflWy em mw puhbhbi
PERCENT
WATER
zr
HUMIDITY
BAIN (g)
LITER/WIN
20
-0.3
0.092
SO
-0.3
0.099
<0
-0.8
0.089
SO
-0.8
0.093
SO
-0.1
0.091
SO
1.6
0.0*8
90
I.ft
O.OftO
BO
1.7
0.0*1
SO
1.6
0.0*8
SO
1.7
O. Oft*
•O
7.S
0.063
to
7.6
0.066
00
0.4
0.067
•0
ft. 7
0.0*7
•0
ft. 4
0.0*7
uatCR CP
BAIN LITEB/MIN
0 0. 109
0 O. 101
0 O. 109
0 1.110
0 0.107
1.7 0.099
l.S 0.094
1.9 0.097
l.a o. ioa
l.S 0.096
7.7 0.077
7.9 0. OK
7. 7 O. 07*
7.9 0.076
7.9 0.078
-------
TABLE 3. EXPOSURE TIME VERSUS ADJUSTMENT FACTORS FOR LOU, MEDIUM, AND HIGH HUMIDITY
COMPANY GREEN VERSUS EERF PUBLISHED
bxptime
CO GREEN
EERF PUB
CO GREEN
EERF PUB
CO GREEN
EERF PUB
LOW RH
LOU RH
NED RH
NED RH
HIGH RH
HIGH RH
24
0.154
0.137
0.128
0.132
0.122
0.116
24
0.137
0.143
0.138
0.137
0.123
0.125
24
0.134
0.141
0.134
0.132
0.116
0.118
24
0.138
0.135
0.133
0.126
0.125
0.117
24
0.139
0.138
0.137
0.127
0.121
0.118
48
0.106
0.107
0.094
0.096
0.078
0.077
48
0.106
0.110
0.097
0.102
0.077
0.082
48
0.107
0.105
0.095
0.097
0.077
0.076
48
0.106
0.101
0.096
0.094
0.077
0.076
48
0.106
0.105
0.095
0.09 8
0.077
0.078
72
0.080
0.087
0.079
0.075
0.056
0.048
72
0.088
0.091
0.077
0.079
0.056
0.051
72
0.085
0.088
0.080
0.075
0.055
0.051
72
0.085
0.083
0.072
0.073
0.054
0.046
72
0.085
0.085
0.076
0.075
0.056
0.049
96
0.070
0.074
0.065
0.058
0.040
0.035
96
0.070
0.080
0.067
0.062
0.041
0.034
96
0.071
0.075
0.065
0.059
0.042
0.033
96
0.070
0.074
0.064
0.057
0.043
0.033
96
0.069
0.075
0.066
0.060
0.038
0.034
120
0.062
0.070
0.055
0.051
0.033
0.023
120
0.061
0.073
0.056
0.054
0.033
0.025
120
0.062
0.071
0.055
0.051
0.031
0.024
120
0.062
0.069
0.055
0.050
0.036
0.023
120
0.064
0.071
0.056
0.052
0.032
0.023
144
0.057
0.064
0.051
0.045
0.018
0.018
144
0.058
0.068
0.050
0.047
0.021
0.019
144
0.060
0.064
0.049
0.047
0.022
0.016
144
0.056
0.062
0.048
0.044
0.019
0.018
144
0.057
0.049
0.024
TABLI 4. MU«HT BAIN VERSUS CALIBRATION FACTORS FOR A • DAY RUN
PERCENT
WATER
cr
WATER
CP
HUMIDITY
BAIN
LITER/MIN
OAIN
LITER/MIN
20
0
0.10*
0
0. 109
20
0. 1
0.10*
0
0. 101
SO
-0.3
0. 107
0
0. lOS
SO
-0. 1
0.10*
0
1. 110
SO
-o. t
0. !0«
0
0. 107
SO
1
O. 054
1.7
O.OM
90
I
0.0*7
i.a
0.094
SO
1.4
O.OM
i.*
0.097
SO
1.4
0. ON
i.a
0.102
SO
1.4
O.OIS
i.a
O.OM
•0
7.3
o.o7a
7.7
0.077
*0
«. 3
0.077
7.S
O.OM
ao
a. 4
0.077
7.7
0.07k
BO
«
0.077
7.»
0.07*
«0
7
0.077
7.*
o.ora
-------
TABLE 5. EXPOSURE TIME VERSOS ADJUSTMENT FACTORS FOR LOW, MEDIUM, AMD HIGH HUMIDITY
COMPANY RED VERSUS EERF PUBLISHED
EXPTIME
CO RED
EERF PUB
CO RED
BERT PUB
CO RED
SERF PUB
LOW RH
LOW RH
MED RH
MED RH
HIGH RH
HIGH RH
24
0.155
0.137
0.139
0.132
0.117
0.116
24
0.155
0.143
0.141
0.137
0.116
0.125
24
0.158
0.141
0.143
0.132
0.119
0.118
24
0.154
0.135
0.139
0.126
0.122
0.117
24
0.156
0.138
0.145
0.127
0.116
0.118
48
0.110
0.107
0.100
0.096
0.071
0.077
48
0.110
0.110
0.099
0.102
0.071
0.082
48
0.111
0.105
0.102
0.097
0.068
0.076
48
0.112
0.101
0.100
0.094
0.069
0.076
48
0.111
0.105
0.099
0.098
0.071
0.078
72
0.087
0.087
0.077
0.075
0.052
0.048
72
0.086
0.091
0.078
0.079
0.048
0.051
72
0.087
0.088
0.077
0.075
0.049
0.051
72
0.088
0.083
0.079
0.073
0.054
0.046
72
0.087
0.085
0.080
0.075
0.050
0.049
96
0.070
0.074
0.063
0.058
0.035
0.035
96
0.068
0.080
*0.065
0.062
0.035
0.034
96
0.068
0.075
0.064
0.059
0.035
0.033
96
0.068
0,074
0.065
0.057
0.040
0.033
96
0.072
0.075
0.064
0.060
0.034
120
0.066
0.070
0.054
0.051
0.028
0.023
120
0.064
0.073
0.054
0.054
0.028
0.025
120
0.064
0.071
0.0S6
0.051
0.029
0.024
120
0.065
0.069
0.054
0.050
0.031
0.023
120
0.068
0.071
0.054
0.052
0.028
0.023
144
0.055
0.064
0.046
0.045
0.021
0.018
144
0.056
0.066
0.047
0.047
0.022
0.019
144
0.055
0.064
0.048
0.047
0.022
0.016
144
0.056
0.062
0.046
0.044
0.021
0.018
144
0.054
0.046
0.022
TABLE 6. uejbht SAIN VERSUS CALIBRATION FACTORS FOR A S DAY RUN
CQHPAMV RED "fTTUffi EEHE PUBLISHED
PERCENT
HUMIDITY
SO
20
£0
£0
20
SO
90
SO
SO
SO
SO
•O
SO
SO
SO
WATER
CF '
HATER
CF
SAIN (g)
LITER/MIN
SAIN (g)
LITER/MIN
-0.7
0.110
O
O. 108
-O. S
O. 110
O
O. 101
-O. S
O. Ill
O
O. 103
-O. S
O. Ill
O
1. 110
-O. S
O. 113
O
O. 107
i. 1
0.099
1.7
O. 093
1
0.099
l.S
0.094
0.9
0.100
1.9
O. 097
1. 1
0.099
l.S
o. ioa
1
0.099
1.8
0.096
a. 4
0.071
7.7
O. 077
S.3
0.071
7.S
O.OSS
B. t
O.OSS
7.7
0.076
S.S
0.0S9
7.9
0.07S
A* 4
0.071
7.9
O. 07S
-------
TABLE 7. EXPOSURE TIME VERSUS ADJUSTMENT FACTORS FOR LOW, MEDIUM, AND HIGH HUMIDITY
COMPANY YELLOW VERSUS KERF PUBLISHED
EXPTIME
CO YELLOW
EERF PUB
CO YELLOW
EERF PUB
CO YELLOW
EERF PUB
LOW RH
LOW RH
MED RH
MED RH
HIGH RH
HIGH RH
24
0.149
0.137
0.137
0.132
0.119
0.116
24
0.143
0.143
0.122
0.137
0.119
0.125
24
0.149
0.141
0.125
0.132
0.125
0.118
24
0.151
0.135
0.125
0.126
0.125
0.117
24
0.143
0.138
0.141
0.127
0.113
0.118
48
0.107
0.107
0.091
0.096
0.077
0.077
48
0.105
0.110
0.094
0.102
0.076
0.082
48
0.106
0.105
0.091
0.097
0.076
0.076
48
0.017
0.101
0.096
0.094
0.078
0.076
48
0.105
0.105
0.093
0.098
0.074
0.078
72
0.083
0.087
0.072
0.075
0.054
0.048
72
0.085
0.091
0.073
0.079
0.057
0.051
72
0.084
0.088
0.071
0.075
0.056
0.051
72
0.084
0.083
0.073
0.073
0.057
0.046
72
0.083
0.085
0.077
0.075
0.056
0.049
96
0.067
0.074
0.062
0.058
0.043
0.035
96
0.067
0.080
0.064
0.062
0.038
0.034
96
0.069
0.075
0.063
0.059
0.043
0.033
96
0.070
0.074
0.060
0.057
0.043
0.033
96
0.067
0.075
0.064
0.060
0.041
0.034
120
0.061
0.070
0.053
0.051
0.034
0.023
120
0.057
0.073
0.053
0.054
0.035
0.025
120
0.059
0.071
0.053
0.051
0.033
0.024
120
0.058
0.069
0.051
0.050
0.036
0.023
120
0.061
0.071
0.050
0.052
0.033
0.023
144
0.055
0.064
0.047
0.045
0.026
0.018
144
0.054
0.068
0.045
0.047
0.026
0.019
144
0.057
0.064
0.047
0.047
0.029
0.016
144
0.058
0.062
0.047
0.044
0.027
0.018
144
0.056
0.046
0.028
TABLE 8. UEIGHT BAIN VERSUS CALIBRATION FACTORS FOR A t DAY RUN
aaeam. Yin i nw vfwmib aac pvbuihp
company xnxfU —RF ni« TBiTT
PERCENT
MATER
CP
WATER
CF
HUMIDITY
GAIN
LITER/MIN
BAIN
LITER/MIN
80
0
O. 107
0
0. 10S
SO
O. 1
O. 103
0
0. 101
SO
O
O. 106
0
0. 105
SO
0.8
O. 107
0
1. 110
SO
0.8
O. 10S
0
O. 107
SO
1.8
0.0*1
1.7
0. 090
90
1. 1
0.094
1.8
O. 094
SO
1.8
0.0*1
1.9
0.097
SO
1
0.096
1.6
0. 108
BO
1. 1
0,093
1.6
0.096
SO
6.8
0.077
7.7
0.077
80
7.8
0.076
7.S
0.068
ao
7
0.076
7.7
0.076
ao
6
0,078
7.9
0.076
80
6.8
0.074
7.9
0.076
-------
2
V
L.
<
t
o
Ik
o
EXPOSURE TIME VERSUS ADJUSTMENT FACTORS
PIslK CO (symbols) VS. EERF PUBLISHED
20* M
201 KB
SM Ul
SOU Mi
80S «H
801 IB
144
HOURS OF EXPOSURE
Flgur* 1.
WATER GAIN VS. CF (FOR 2 DAY EXPOSURE)
PNK 00 (symbol*) VS EERF PUBLISHED
2 4 •
WKtfTt QMN fa) (M40WN Cfflft TOMS)
Fi|ur» I,
-------
N.
v
k.
!
h.
o
EXPOSURE TIME VS. ADJUSTMENT FACTORS
GREEN CO (aymbota) VS. EERF PUBLISHED
80% Ml
HOURS OF EXPOSURE
Ftgur* 3.
WATER GAIN VS. CF (2 DAY EXPOSURE)
GREEN CO tombola) US. EERF PUBLISHED
WTO* OMN («) (M40MN 0*1* P0MI3)
U|ura 4.
-------
EXPOSURE TIME VS. ADJUSTMENT FACTORS
RED CO (symbol*) VS. EERF PUBLISHED
HOURS OF EXPOSURE
ritur* 5.
WATER GAIN (g) VS. CF (2 DAY EXPOSURE)
WATtR (g) (MOWN OMR* P0WI3)
Fifur* 6.
-------
EXPOSURE TIME VS. ADJUSTMENT FACTORS
YELLOW CO (lymbo)*) VS. EERF PUBLISHED
s
It
•*
HOURS OF EXPOSURE
Pigor* 7.
WATER GAIN (g) VS. CF (2 DAY EXPOSURE)
YELLOW 00 («ymbo)«) VS. EERF PUBLISHED
1
V.
o
Wist OMN (a) (MMINN DMA PMNI3)
Fl|ur* 8.
-------
Ill—PI—6
CALIBRATION OP ALPHA-TRACK MONITORS
FOR MEASUREMENT OF THORON (RN-2201
by: Mark D. Pearson*
UNC Geotech
P.O. Box 14000
Grand Junction, CO 81502-5504
ABSTRACT
Several types of coMerclal alpha-track aonitors are supplied In a
standard configuration with diffusion barriers that are perneable to both
radon (Rn-222) and thoron (Rn-220). These Monitors can be Modified to exclude
thoron by the choice of a diffusion barrier so that the diffusion tine across
the barrier Is on the order of several half-lives of thoron (55.6 sec).
Devices with and without thoron diffusion barriers can be simultaneously
exposed to determine an appropriate calibration factor for thoron.
Terradex Track Etch* Type M and Type SM Monitors, supplied by Terradex
with a thoron barrier, and Terradex Radtrak* Monitors, Modified with a slMilar
thoron barrier, were exposed slMUltaneously with Track Etch* Type F, Track
Etch* Type SF, and Radtrak* Monitors in a thoron chaMber at the U.S.
OepartMent of Energy Grand Junction Projects Office. The thoron barriers
prevented thoron froM entering the Monitors. A calibration equation was
determined that perMits calculation of integrated thoron exposures froa the
reported track densities and radon exposures of the slMUltaneously exposed
devices. The thoron calibration factor was established with the greatest
precision for the Radtrak* Monitors. Study results Indicate that Radtrak*
Monitors provide the greatest sensitivity to thoron.
•Work perforMed for the U.S. Department of Energy under Contract No. DB-AC07-
861012584.
-------
INTRODUCTION
The U.S. Department of Energy Grand Junction Projects Office (DOE/GJPO)
Radon Laboratory has conducted a nuaber of studies (1,2,3) evaluating the
precision and accuracy of alpha-track monitors for the Measurement of radon
(Rn-222). The GJPO studies and other studies have demonstrated the usefulness
of using alpha-track monitors for measurement of integrated radon exposures.
Alpha-track devices have also been proposed for estimating exposures of thoron
(Rn-220), though few if any studies exist to demonstrate the validity of this
approach.
Thoron and thoron-decay progeny are believed to be a significant component
of indoor radioactivity. Some researchers have estimated that average dose
equivalents due to thoron daughters are 20 to 30 percent of those due to radon
daughters (4,5). Several methods exist for the measurement of thoron
daughters, but there are no proven, easily applied methods for measuring
integrated thoron exposures. Alpha-track devices could provide a simple,
inexpensive method for measuring integrated thoron exposures, under
appropriate conditions, if a valid procedure for doing so could be
demonstrated.
SCOPE
Two type8 of alpha-track devices are marketed commercially with filters
that the manufacturer claims excludes thoron. These devices are the Terradex*
Track Etch* Type M and Type SN monitors. Theoretically, these monitors could
be used in tandem with the standard Track Etch* Type F and Type SF monitors,
which admit both radon and thoron, to determine the net thoron exposure. The
net thoron exposure would be calculated from the difference in response
between the Type M/type SM and Type F/Type SF monitors.
Track Etch* Type M, Type F, Type SF, and Type SM alpha-track devices were
exposed to a range of thoron and
-------
Type SM monitors make use of this short half-life to exclude thoron by
employing a membrane with a mean diffusion length that precludes infiltration
by thoron. Because of thoron's short half-life, there is a low probability
that thoron atoms will cross the membrane. The diffusion of radon into the
alpha-track device is only slightly reduced by the increased diffusion length.
The standard filters provided by the manufacturer in Track Etch* Type F and
Type SF devices and Radtrak® devices readily admit both thoron and radon.
A thoron chamber was constructed at the GJPO Radon Laboratory to conduct
exposures of alpha-track devices at constant thoron concentrations and mixed
thoron-radon concentrations. The short half-life of thoron requires minimum
transit time to produce a relatively constant concentration inside the
chamber. If transit time is lengthy, a significant fraction of the thoron
entering the chamber will decay before exiting the chamber and a noticeable
concentration gradient will exist Inside the chamber. The thoron chamber was
designed with high air-flow rates and a small chamber volume to overcome this
difficulty.
The GJPO thoron chamber is a 20-liter stainless steel cylinder fabricated
from a commercial pressure cooker. The chamber dimensions are approximately
30 cm in diameter by 30 cm in height. The top of the chamber is removable,
allowing placement of devices in the chamber. Thoron is sampled from one of
two ports in the top of the chamber. Room air is pushed through a Pylon Model
Th-1025 flow-through thoron source and thence into the chamber. The
temperature and relative humidity of the chamber air are continuously
monitored but not controlled. The chamber was operated at flow rates ranging
from 90 to 500 Lpm for exposures during this study, which produced transit
times ranging from 2 to 13 seconds.
The primary thoron measurement is made using a two-filter tube (6).
Thoron concentrations are also monitored with two continuous flow-through
scintillation monitors. The first monitor samples air from the chamber with
as short a path length as possible to minimize the loss of thoron to
radioactive decay. The second monitor samples the exhaust of the first
¦onltor with a plumbing system that allows 10 minutes for 99.9 percent of
the thoron present to decay prior to entering the second scintillation
chamber. This second monitor, therefore, measures only radon.
The thoron concentration at steady-state conditions can be calculated by
subtracting the contribution of radon from the counts recorded by the first
monitor, and then applying a calibration equation. Hourly thoron
concentrations are recorded by a microcomputer-based data acquisition system.
The coefficient of variation of hourly thoron concentration readings measured
by this method never exceeded 10 percent during any exposure.
Seven exposures of sets of alpha-track devices were conducted in the
thoron chamber. The environmental conditions for each exposure are given in
Table 1. Six of the exposures were 4 days in duration and one exposure was
-------
TABLE 1. CHAMBER ENVIRONMENTAL CONDITIONS FOR
THORON EXPOSURES
Temp
RH
Duration
ExDOSure
(«>Ci-d/L)
Set
CC)
(*)
(davs)
Thoron
Radon
1
36
23
4
1,640
9.2
2
28
50
4
3,516
10.4
3
29
55
4
6,296
9.6
4
32
58
4
5,224
276.0
5
33
46
4
2,916
163.6
6
32
63
4
1,356
164.8
7
31
41
17
21,607
651.1
8
31
41
17
0
651.1
17 days in length. The first three exposures were nade in atmospheres
containing primarily thoron with little radon. The remaining four exposures
were performed in a mixed thoron-radon environment. The 17-day exposure set
included a group of monitors that were exposed to the exhaust air from the
chamber, with a sufficient time delay to ensure that all the thoron but
essentially none of the radon had decayed. This group of monitors, with the
time-delay exposure, is Set 8 in the data.
Pive Type F and Type M monitors and approximately six Type SF, Type SM,
Radtrak*, and modified (equipped with a thoron-barrier) Radtrak® devices were
exposed for each test. In most cases, two unexposed monitors were returned
for processing along with the exposed devices from each test.
THEORY FOR CALCULATING THORON EXPOSURE
The radon exposure reported by an alpha-track monitor without a thoron
barrier cannot be taken at face value as a thoron-exposure measurement because
the amount of reported exposure due tfo thoron and the amount due to radon are
unknown. The monitor also has different calibration factors for radon and
thoron. Taking the reported exposure at face value as the true thoron
exposure would underestimate the actual thoron exposure by 30 to 40 percent in
the study's chamber exposures, and would lead to even greater discrepancies as
the thoron and radon concentrations became more equal.
A method was developed for calculating thoron exposure from the side-by-
side exposure of alpha-track monitors with and without thoron barriers. For
monitors without thoron barriers, the method permits the separate calculation
of a distinct calibration factor for radon and thoron.
To apply the method, alpha-track monitors with and without thoron-barrlers
are exposed side-by-side, If the reported radon exposures from each type of
-------
monitor are significantly different, thoron is assumed to be present. The
following steps permit calculation of the thoron exposure.
The total track density registered by a monitor without a thoron barrier
is the sum of tracks due to the radon and the thoron decay-chain alpha
particles. The thoron track density is the difference between the total track
density on this monitor and the equivalent radon track density on the monitor.
The equivalent radon track density on the monitor without the thoron barrier
is determined from the radon exposure reported by the monitor with the thoron
barrier. Once the radon and thoron calibration factors have been determined
from known exposures, pairs of monitors with and without thoron-barriers can
be exposed to unknown mixed concentrations of radon and thoron.
For mixed exposures, the equivalent radon track density on the monitor
without the barrier is equal to the radon exposure reported by the monitor
with the thoron barrier multiplied by its radon calibration factor.
This value for the equivalent radon track density is then subtracted from
the total track density on the monitor without a thoron barrier, to give a
value for thoron track density. Finally, the thoron track density is divided
by the thoron calibration factor for the non-barrier device, which results in
a calculated thoron exposure.
This approach to calculating the thoron exposure can be expressed
mathematically as follows. The track density reported on an alpha-track
monitor is equal to the sum of the track densities due to radon and thoron.
For monitors without thoron barriers (denoted by "f"), and for monitors with
thoron barriers (denoted by "m"), the equations are:
Pf " RfBRn + TfETn»
pm " RmERn + TmETn»
where
pf or Pm • track density for f or m device (tracks/mm2),
Rf or Rh - radon calibration factor for f or m device (tracks/mm^/pCi*d/L),
ERn - radon exposure (pCi*d/L),
Tf or TB - thoron calibration factor for f or m device (track8/mm2/pCi-d/L),
ETn ¦ thoron exposure (pCl'd/L)
These two equations may be simultaneously solved for Efn:
RfP* * RmPf
ETn "
RfTm " R«Tf
-------
If the infiltration of thoron through the thoron-barrier is assumed to be
zero (which will be shown below to be a reasonable assumption) such that
Tn « 0, then the equation for the thoron exposure reduces to:
ETn " (Pf " RfERn) / Tf- ^
The term RfERn is simply the equivalent radon track density on the device
without a thoron barrier, and the first term is the total track density on the
same device; the difference is the net track density due to thoron. The net
track density is divided by the thoron calibration factor for the non-barrier
device.
A test was performed to evaluate the validity of the assumption that the
infiltration of thoron through the thoron-barrier is minimal. Radtrak®
monitors with thoron barriers were exposed to a mixed thoron-radon atmosphere,
and were also exposed to the same atmosphere after the air had transited a
delay volume so that all of the thoron present had decayed prior to reaching
the second group of monitors. This second group of Radtrak* monitors with
thoron barriers was therefore given the same radon exposure as the mixed
exposure group. The reported track densities for both groups should be equal
if the thoron barrier absolutely excludes thoron.
The six thoron-barrier monitors exposed to the mixed atmosphere recorded a
mean track density of 21.5 tracks/mm2, with a standard deviation of 1.1. The
four thoron-barrier monitors exposed to the radon-only atmosphere recorded a
mean track density of 16.5 tracks/mm2, with a standard deviation of 0.5.
Applying a t-test to the comparison of the two means yields a t-statistic of
7.44, which leads to the conclusion that the two means are not equivalent.
This result Implies that devices exposed to both thoron and radon recorded
more alpha tracks than the devices exposed solely to radon. An explanation
for this result is that the thoron barrier does not completely exclude thoron
from entering the monitor and that the admitted thoron is responsible for the
additional tracks.
The total thoron exposure for thi% test, however, was 21,607 pCi-d/L.
Assuming that the 5.0 tracks/mm2 difference in the two sets of monitors is due
to the thoron exposure, the thoron calibration factor for the thoron-barrier
devices would be on the order of 5.0 tracks/mm2 divided by the thoron exposure
of 21,607 pCl'd/L, or 0.0002 tracks/mm2/pCl-d/L. This calibration factor is
only 1 percent of the thoron calibration factor for the devices without thoron
barriers, or of the radon calibration factor for either type of Radtrak*
monitor. Thoron diffusion through the thoron barrier likely appeared in this
test due to the large thoron exposure the devices received. In lower thoron
exposures that are expected to be more typical, thoron diffusion through the
barrier can be ignored for computational purposes with minimal Impact.
-------
RESULTS
ANALYTICAL METHOD
Results were analyzed using the track densities reported by Terradex
and the Measured radon and thoron concentrations in the thoron chamber. A
calibration factor for the exposure of the various types of monitors without
thoron barriers (Radtrak*, Track Etch* Type F, and Track Etch*
Type SF) to a radon-only environment was determined from a least-squares fit
through the origin of a set of six or seven exposures. These exposures were
conducted separately from the exposures in the thoron chamber and covered a
range of 71 to 729 pCl-d/L. The calibration factor determined from these
exposures is the radon calibration factor for the monitors without thoron
barriers.
Applying the theory, the track density due to radon reported by the
Monitors without thoron barriers was determined using the radon calibration
factor and the measured radon exposure. This track density is the equivalent
radon track density for the monitors without thoron barriers. The track
density on these monitors due to thoron was then calculated as the difference
between the reported track density and the equivalent radon track density.
The calibration factor for the exposure to thoron of monitors without
thoron barriers was calculated from a least-squares fit through the origin of
the thoron track density versus thoron exposure. The uncertainty in the
calibration factor was estimated at the 95 percent confidence level by
applying the appropriate factor from the t-dlstribution table to the error in
the slope of the least-squares fit line.
There was an additional uncertainty inherent in the thoron measurement
relating to the calibration of the measurement. The "true" thoron
concentration value was determined fro* thoron grab samples taken with a two-
filter tube. The two-filter tube measurement was cross-checked with an
expected value from the thoron source, based on the stated activity of the
source and the flow rate through the source. As a result, there was
potentially a systematic uncertainty that would have a bearing on the accuracy
of the calibration factor but not on the precision of the calibration factor.
RADTRAK* RESULTS
The exposures of Radtrak* monitors provided results with the greatest
precision and the greatest potential for accurate calculation of thoron
exposures. The radon calibration factor for Radtrak* Monitors without thoron
barriers was 0.0266 tracks/mm2/pCl'd/L. The calculations of average thoron
track density for the Radtrak* monitors are summarized in Table 2. Radtrak*
¦onitors were not exposed in Sets 1 and 2.
-------
TABLE 2. AVERAGE TRACK DENSITY CALCULATION FOR RADTRAK® MONITORS
Thoron Radon Total Equivalent Radon Net (thoron)
Exposure Exposure Track Density Track Density Track Density
Set (pCl-d/L) (pCl-d/L) (tracks/mm2) (tracks/ma2) (tracks/mm2)
3
6,296
9.6
125.1
0.3
124.8
4
5,224
276.0
101.0
7.3
93.7
5
2,916
163.6
57.5
4.4
53.1
6
1,356
164.8
30.6
4.4
26.2
7
21.607
651.1
456.6
17.3
439.3
The calibration factor for the exposure to thoron of Radtrak* monitors
without thoron barriers, calculated from a least-squares fit through the
origin of the average thoron track density versus thoron exposure, was
0.0201 track8/mm2/pCi,d/L. Thoron track density is plotted against thoron
exposure In Figure 1. At the 95 percent confidence level, the uncertainty in
the thoron calibration factor for the Radtrak* monitors was 0.0010 tracks/
mm2/pCi-d/L.
0 4000 8000 12000 16000 20000 24000
Thoron Exposure (pCi • d/L)
Figure 1. Thoron track density recorded on Radtrak* monitors as a function
of thoron exposure.
-------
TRACK ETCH® TYPE SP RESULTS
The radon calibration factor for Track Etch® Type SF monitors (without
thoron barriers) was 0.0152 tracks/mm2/pCi-d/L. The calculations of average
thoron track density for the Type SF Monitors are summarized in Table 3. The
Type SF monitors were not exposed in Set 7.
The calibration factor for the exposure of Type SF monitors to thoron,
calculated from a least-squares fit through the origin of the average thoron
track density versus thoron exposure, was 0.00940 tracks/mm2/pCi-d/L. Thoron
track density for the Type SF monitors is plotted against thoron exposure in
Figure 2. The uncertainty in the least-squares fit at the 95 percent
confidence level was 0.00163 tracks/mm2/pCi-d/L.
TABLE 3. AVERAGE TRACK DENSITY CALCULATIONS FOR TRACK ETCH* TYPE SF MONITORS
Thoron Radon Total Equivalent Radon Net (thoron)
Exposure Exposure Track Density Track Density Track Density
Set (pCl-d/L) (pCld/L) (tracks/mm2) (tracks/mm2) (tracks/mm2)
1
1,640
9.2
16.7
0.1
16.6
z
3,516
10.4
41.2
0.2
41.0
3
6,296
9.6
61.2
0.1
61.1
4
5,224
276.0
46.1
4.2
41.9
5
2,916
163.6
29.1
2.5
26.6
6
1.356
164.8
13.8
2.5
11.3
Thoron Exposure (pCi • d/l)
Figure 2. Thoron track density recorded on Track Etch* Type SF monitors as
a function of thoron exposure.
-------
TRACK ETCH® TYPE F RESULTS
The radon calibration factor for Track Etch® Type F monitors without
thoron barriers was 0.0461 tracks/mm2/pCi'd/L. The calculations of average
thoron track density for the Type P monitors are summarized in Table 4.
The calibration factor for the exposure of Type P monitors to thoron,
calculated from a least-squares fit through the origin of the average thoron
track density versus thoron exposure, was 0.0133 tracks/m*2/pCi*d/L. Thoron
track density for the Type F monitors is plotted against thoron exposure in
Figure 3. At the 95 percent confidence level, the uncertainty in the least-
squares fit was 0.0031 tracks/mm2/pCi-d/L.
TABLE 4. AVERAGE TRACK DENSITY CALCULATIONS FOR TRACK ETCH* TYPE F MONITORS
Thoron Radon Total Equivalent Radon Net (thoron)
Exposure Exposure Track Density Track Density Track Density
Set (pCl'd/H (pel-d/L) (tracks /mm2) (tracks/mm2) (tracks /mm21
1
1,640
9.2
41.6
0.4
41.2
2
3,516
10.4
87.3
0.5
86.8
3
6,296
9.6
87.9
0.4
87.5
4
5,224
276.0
101.7
12.7
89.0
5
2,916
163.6
70.9
7.5
63.4
6
1,356
164.8
40.5
7.6
32.9
7
21.607
651.1
298.9
30.0
268.9
Figure 3. Thoron track density recorded on Track Etch* Type F monitors as a
function of thoron exposure.
-------
SAMPLE THORON CALCULATION
Once the calibration factors for radon and thoron for an alpha-track
¦onitor are known, the thoron exposure can be calculated from side-by-slde
exposures of devices equipped with and without thoron barriers. As an
example, consider the Radtrak* monitors with and without the thoron barrier
that were exposed in Set 6.
A comparison is made of the reported results from the side-by-side
exposure of the monitors. The Radtrak® monitors with a thoron barrier
reported 193 pCi*d/L while the monitors without the thoron barrier reported
1,072 pCi-d/L. Since the two types of monitors report significantly
different results, the assumption can be made that thoron is present.
Equation (1) is applied:
(pf - RfERn)
ETn
Tf
In this case,
pf = 30.6 tracks/mm2 (as reported by the manufacturer),
Rf - 0.0266 tracks/mmVpCi*d/L,
Tf » 0.0201 tracks/mm2/pCi'd/L,
ERn = 193 PCi'd/L.
The thoron exposure calculated with Equation (1) is 1,267 pCi'd/L. The
measured thoron exposure for Set 6 was 1,356 pCi'd/L. The calculated exposure
is within approximately 6 percent of the measured value.
PRECAUTIONS
Alpha-track monitors are known to give occasionally spurious results. In
some instances, a monitor receiving a low exposure will report a high result,
and the converse can also occur. Since the calculation procedure begins with
the comparison between a monitor with a thoron barrier and a monitor without a
thoron-barrier, an incorrectly high or low reading will lead to an incorrect
conclusion regarding the presence of thoron. To help overcome this inherent
difficulty, it is recommended that alpha-track monitors of each type be
exposed in duplicates, at a minimum.
While the best results were seen for the Radtrak* monitor with and without
the thoron barrier, it should be emphasized that the Radtrak* monitor with a
thoron barrier is not commercially available. The devices used for this study
were converted by removing the factory-supplied filter fro* the Radtrak*
-------
monitor and replacing it with a membrane cut from Material supplied for use
with the Track Etch* Type F Monitors.
The Measurement of thoron in a structure entails a variety of probleMS
unique to thoron and its short half-life. There may be little thoron present
in a structure, or the measured thoron concentration in a room may vary
dramatically depending on the location of the monitor and the location of
thoron entry into the room due to the short half-life of thoron.
CONCLUSIONS
Alpha-track monitors used in tandem with and without thoron-excluding
barriers can provide a reasonable estimate of thoron exposure. The
commercially available thoron-barrier does exclude thoron to a significant
extent. Thoron exposure can be readily calculated using the track densities
and radon exposures reported by alpha-track monitor manufacturers and by
applying an empirically determined calibration factor for thoron. Alpha-track
monitors are less sensitive to.thoron than they are to radon, probably becaus*
of thoron's short half-life.
Of the devices evaluated in this study, the Radtrak* monitors provided th«
greatest precision and sensitivity in estimating thoron exposure. The Track
Etch* Type SF monitor provided the least sensitivity for measuring thoron
exposure. The Track Etch* Type F monitor provided an intermediate sensitivity
for thoron. The diminished sensitivity of the Type F monitor may be due to
its large volume that reduces the efficiency by which the short-lived thoron
species registers alpha damage tracks on the alpha-track filM.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorseMent should be inferred.
-------
REFERENCES
1. George, J.L. and Langner, G.H., Jr. Field Study of Indoor Average
Radon-Daughter Estimation Methods. GJ/TMC-26, U.S. Department of
Energy, Grand Junction Projects Office, Grand Junction, Colorado, 1986.
2. George, J.L. and Langner, G.H., Jr. Validation of the Alpha-Track Method.
UNC/TMC-33(TMC), U.S. Department of Energy, Grand Junction Projects
Office, Grand Junction, Colorado, 1987.
3. Pearson, M.D. Evaluation of the Performance Characteristics of Radon and
Radon-Daughter Concentration Measurement Devices Under Controlled
Environmental Conditions. UNC/GJ-44(TMC), U.S. Department of Bnergy,
Grand Junction Projects Office, Grand Junction, Colorado, 1989.
4. Schery, S.D. Measurements of Airborne 212Pb and 220Rn at Varied Indoor
Locations Within the United States. Health Physics. Vol. 49, No. 6,
pp. 1061-1067, 1985.
5. Martz, D., Falco, R.J., and Langner, G.H., Jr. Time-Averaged Exposures
to 220Rn Progeny and to 222Rn Progeny in Colorado Homes. In publication.
U.S. Department of Energy, Grand Junction Projects Office, Grand Junction,
Colorado, 1989,
6. Mayya, Y.S. and Kotrappa, P. Modified Double Filter Method for the
Measurement of Radon or Thoron in Air. Ann. Occup. Hyg. Vol. 21,
pp. 169-176, 1978.
-------
III-P1-7
Quality Assurance Procedures for Home Radon Testing
by: Richard W. Tucker
Radonics Inc.
McLean, VA 22102
John G. McGreevey
Radonics Inc.
McLean, VA 22102
ABSTRACT
Quality Assurance is the critical issue for radon tests that are being
made in homes. There are many factors that must be addressed in a positive and
methodical manner. These issues include assuring that EPA closed house
conditions are maintained; that extraordinary weather conditions did not
adversely affect the test; that the equipment is not tampered with or relocated
during the test period; that test results are not confused or lost; and, that
correct measurement equipment operation was maintained throughout the test.
This paper addresses the above issues as well as proposing methods and
procedures to assure that reported test results are an accurate indication of
the conditions that existed during the test period.
INTRODUCTION AND SCOPE
It is common to read and hear that "radon measurements are easy to make".
On the surface this statement may appear to be true. What could be easier than
picking-up a test kit while at the grocery store, hanging it in the basement
for a few days and sending it off to a lab? If, however, one inquires into the
confidence that can be placed in any particular measurement when making health
or mitigation decisions based on that measurement, there are many questions
that must be answered before the result has any meaning. The elements that
surround the process of conducting a radon measurement and reporting the
results of the measurement in a manner in which meaning and confidence can be
assigned to the results is the process of quality assurance (QA).
Radonics is an engineering firm specializing in radon, especially the
making of radon measurements. This paper describes the QA procedure* and
methods that have evolved over the last three years at Radonics. These
procedures have been driven by actual problems encountered in the field and
have evolved through and been refined by actual field experience. In fact
refinements to the overall QA process are a continuing process which is driven
by problems that are still encountered and by the desire to conduct business
operations with the highest quality and in the most efficient manner possible.
-------
Radonics perforins radon screening tests in homes and commercial building®
using sophisticated equipment installed by trained field engineers and
technicians. Radon tests can of course be performed in other manners, with
other test methods and for other purposes. Radonics, for example, also
performs long term radon tests. The differences in the test technology and
procedures for the long term tests result in the need for different Qk
procedures for those tests. This paper is oriented to screening measurements
and the QA variations required for other test types are not discussed in this
paper. The procedures outlined here, however, should provide a good insight
into the types of activities, equipments, procedures and methods that at®
required to perform a meaningful radon test.
MAJOR ISSUES TO ADDRESS
The following list contains some of the issues with which the QA process
must ensure conformity during a test cycles
o Do not confuse, interchange or lose test information (information
integrity)
o Insure that EPA closed house conditions are maintained prior to and
during the test
o insure that all other EPA protocols (eg. equipment placement, test
times, weather conditions, etc.) are met
o Verify that equipment used is still in calibration period
o Check that equipment used is operating properly
o Analyze all critical test results by multiple sources before
reports are released
o Keep all test data for future review and analysis
Fortunately, for Radonics, most of these issues were identified early in
the history of the company when test volumes were low enough that each test
could receive the personal and intensive attention of one of the principals of
the company and mistakes were quickly corrected (although sometimes requiring
a re-test). It was easily recognized, however, that if the approach to
addressing these above stated issues in a consistent, orderly and as automated
a manner as possible was not established, the potential for major problem#
would grow faster than new business would grow. It was also recognized that,
while each of these above issues could be separately treated, many of the
issues or at least approaches to addressing these issues were interrelated*
A total systems approach would be necessary.
The remainder of this paper discusses each of the above issues and approaches
taken to address each of those issues separately and then presents the
integrated QA process currently being used that incorporates these approaches •
In other words, we will first examine each of the trees and then back-up to
examine the forest.
-------
INFORMATION INTEGRITY
Information integrity is the issue dealing with the handling of all of the
information associated with a test, from the time that an order is placed and
continuing through the entire sequence of events that normally culminates in
the issuing of a written report to the purchaser. There are three major areas
of mistakes that can degenerate the information content of a test: lost data,
data errors and interchanged data. A separate but somewhat related problem is
that of protecting access to the data and the QA process.
Lost data may occur because of a number of problems with the data transfer
and storage process. Lost data may occur because of manual data handling
problems or due to computer failure at a critical time. Lost data problems are
alleviated through the elimination of manual data handling, except at the point
of order taking, the use of un-interruptable power sources (UPS's) on critical
office automation equipment, office data backups every three business hours and
storage of data in multiple locations.
The test data collected in the test equipment remains in the test
equipment even after the test is complete. The test equipment contains storage
media that can hold up to about 75 tests. Each test remains in the test
equipment until the storage media fills up with data. At the end of a test,
data is transferred from the test equipment to a laptop computer. At the end
of each day, data is transmitted from the laptop computer to the office, but
the data is not deleted from the laptop computer until several days after the
completion of the job. When data arrives in the office over phone lines, two
separate copies are placed on two separate servers in the office. On the day
after the completion of a job, there are no less than four copies of the test
data in existence. If data is lost at any step, it is only required to back-up
one step to retrieve the test data.
When the test cycle is totally finished, all of the information files
associated with the particular job are compressed into an efficient data
storage element and put into a special section of on-line storage and backed-up
on tape.
Data Errors occur in the data transfer process when individual pieces of
data are incorrectly transferred.
The test sequence usually begins with what remains as the most error prone
portion of the test process: the taking of the order. This is usually done
over the phone with the person placing the order for the radon test giving data
such as address, name of owner, name of person responsible for property access,
etc. to a Radonics office worker who must enter this data into a computer
terminal. There is a lot of data to be taken and there are many chances of
making an error. Critical items are read back to the person placing the order
to help minimize errors. Both the person placing the order and Radonics
personnel are responsible for errors at this stage.
Once this manual data transfer process in complete, all data transfers are
automatic. Those that occur outside of the office computer system (such as
machine to laptop and laptop to office exchanges) make use of some form of
continuous error checking. Data is transferred over a serial interface from
the test equipment to the laptop computer. Every byte that is sent to the
laptop is re-transmitted back to the test equipment where it is compared
against the data byte just sent. Phone transmissions from laptops in remote
locations to the office use cyclic redundancy checks on each 256 byte block to
insure error free data transmission.
-------
Interchanged data is perhaps the most dangerous of all possible data
problems because the data may appear to have perfect integrity to the QA person
reviewing the data. Interchanged data would occur if a field test engineer
sent in data that appeared to be taken at one location but in actuality was
collected at another.
This problem is guarded against in several ways. Each job is assigned a
job number by the office automation system at the time that the order is
received in the office. All portions of the test information are tagged with
this job number. This number is never assigned to a piece of data through
manual means. For example, when an engineer is installing a test system in a
home, he does not enter the address or job number with which that data will be
tagged. Rather he had received, via modem the previous night, the list of jobs
and addresses that he would perform the next day. This information resides on
his laptop computer. When he arrives at a job site, the laptop presents him
with the list of locations that he is to be at that day. He needs only to use
his cursor keys to highlight the current address. This address is double-
checked when he returns to pick-up the machine.
The access to the test data and results is protected. First, it is
protected against certain access by unauthorized office personnel. Only
certain individuals are allowed to participate in the in-office QA process.
Proper names and passwords must be entered before data can be reviewed and
accepted. Results are not available to anyone until the QA process is complete
and the results of a test are treated as confidential information. These are
given only to the person paying for the test or others designated by the person
paying for the test.
These procedures insure that data is correctly transmitted. There is no
chance of confusing collected test data with the wrong order. If data is lost
at any stage it is backed-up at all previous stages. Access to data is
protected. Results are unable to be reported until the QA process is complete,
and then only to authorized persons.
EPA CLOSED HOUSE CONDITIONS
Ensuring EPA closed house conditions is perhaps the most important element
in conducting screening measurements. The EPA specified conditions provide
consistency of measurement results. Without them, screening measurements would
be subjected to wide variations in response to changes in living conditions and
the effects of the outside environment. Radonics data for 1989 shows that
seasonal variations are limited to 1 sigma points that are only 16% from the
mean of the annual test average (mean = 2.90pCi/l, 1 Sigma = 0.47pCi/l).
Actual variations in a dwelling not under closed house conditions can be over
a thousand percent as living conditions and seasons vary.
Other than leaving personnel in a house to monitor the closed house
conditions, there is no absolute way to know if EPA closed house conditions
have been maintained for the entirety of the test. There are however several
means that can be used to test for compliance with closed house conditions.
The best and most reliable of these is an analysis of the actual test data.
The measurement equipment used by Radonics logs measurement data, certain
environmental parameters and system performance parameters every fifteen
minutes. Other than normal diurnal variations and extraordinary weather
conditions, variations in the measurement data collected every 15 minutes
should remain within certain statistical norms when closed house conditions are
maintained. With the absence of EPA recommendations in this area each testing
company must define its own limits of acceptability. When test data fall out
of the statistical norms the test should be repeated. One of the factors that
Radonics considers in the test evaluation part of the QA process is how "well
behaved" the test was. Figures 1 through 6 illustrate graphs of data from
actual tests.
-------
Figure 1 - Test A Figure 3 - Test B
Figure 5 - Test E
Figure 6 - Test F
-------
Three of the key measurements are ramp-up time, maximum variation betwe®"
the highest and lowest sample and the "tightness" of the sample data *¦
indicated by the variance of the 1 sigma points from the mean of the dat#'
Table I shows these statistical measures for the test data graphed on
previous page. The last entry in the table gives the typical maximum value®
for good tests. Exceeding any one of these limits is cause for concern abo°~
the validity of the test. Exceeding any of these limits by more than a factor
of two should be enough cause to designate the test as invalid and to rep®a
the test.
Table I - Test Statistics
Test
Mean
1 Sigma
Max
Ramp
Data
(Av)
Error
Ratio
Up
Test A
0.4 pCi
81
%
11.8
22 hrs
Test B
1.1 pCi
91
%
13.8
2 hrs
Test C
10.1 pCi
19
%
2.3
2 hrs
Test D
2.0 pCi
35
%
7.5
8 hrs
Test E
4.2 pCi
13
%
1.8
2 hrs
Test F
1.3 pCi
109
%
73.8
2 hrs
Typical Max
—
33
%
3.0
2 hrs
The tests labelled Test A, Test B, Test D and Test P all exceed the stated
maximum limits. The tests labeled Test C and Test E fall within the acceptabl®
limits. Test F in figure 6 on the previous page is of particular interest*
Test F represents a two day test. The house was closed-up prior to the arrival
of our test technician. The equipment quickly ramped-up to almost 7 pCi/*
during the first two hours of the test. The home owner then opened the do of*
and windows of the house and the reading fell to outdoor levels (about
pci/l) and stayed low until two hours before the scheduled piok-up time th®
following day. At the appointed pick-up time we were unable to get back to th®
property to pick-up the measurement equipment. The home owner then re-open®"
the house until two hours before our pick-up the next day.
The reasons for the variations in test data are many. They may b®
occupant interference with the EPA closed house conditions/ extreme weath®r
conditions, unusual activity within the dwelling or the phase of the moon. W*
are usually able to associate variances with an external cause, but not; in all
cases. The most common problem, by far however, is some sort of interference
with the maintenance of the EPA closed house conditions on the part of th«
occupant. Whatever the reason though, the calculated measured radon
-------
8pa protocols
In addition to closed house conditions, the EPA specifies other test
Parameters for screening tests. TheBe include directions as to the placement
f the measurement device, length of time for which the test must be performed
nd cautions for unusual weather conditions.
Test personnel should be thoroughly trained in these protocols. To make
t"r® that the equipment has been properly placed, the test engineer or
wh? sketches on his laptop computer screen, the level of the house on
"ich he has placed the CWLM. He must show walls, windows, crawl spaces, sump
Pump holes and other pertinent information. On this drawing, he marks the
Placement of the CWLM and the locations at which grab samples are taken. He
8 also asked a number of questions by the laptop pertaining to the house
onstruction and conditions. This information is available to the QA personnel
c°r their review.
Equipment calibration
.It ^-8 not left up to the field engineer to make sure that equipment is
eoi < ° cal*-bratl-on period. In fact, it is usually necessary to pry
HUipment, that appears to be in good working order, out of the hands of field
ngineers and technicians.
i The last calibration date for each piece of measurement equipment is kept
ua ^ data base in the office. The calibration period for most of the equipment
eoff ky Radonics is six months. A weekly report is generated that flags
quipments approaching the calibration expiration date. These systems are
thf d before the expiration date to ensure that no equipment is in the field
at has exceeded the calibration date.
correct equipment operation
thi Insuri»9 that machines are within their calibration cycle is a fairly easy
De ?g to do* Just because a measurement system is within its' calibration
th*+°d however' is no guarantee that it is working properly. It is mandatory
no « QA process ensure that constant monitoring of equipment operation is
k In home radon testing, measurement equipment is typically moved from a
BUM t0 a motor vehicle and back to a home every day or two. This process can
80m ??*" t*le measurement equipment to shock/ vibration and temperature changes,
ometimes to extreme levels. This constant movement of equipment can sometimes
suh6 <*ama9e* Host precision radon measurement equipment contains components
com as aolid state alpha detectors and photo-multiplier tubes. These
ar?P°nenta are the most sensitive of the measurement system components to shock
a "ave the biggest impact on measurement accuracy.
re .Alao air sampling process required in making precision measurements
the use of pumps and air paths that must be absolutely air-tight.
mP8 and seal are mechanical devices that can fail at any time.
ev Because of these types of problems, equipment checks every six months or
3i®ry three months or even every month are inadequate. It is unacceptable to
eff?°Ver three days after test results are reported that a major change in the
thon°y o£ the ^ or 9as detector has occurred sometime in the last week on
0£® machine that made that test. The test could be repeated, but mitigation
other contractual events may have already occurred.
-------
Because of the possible deleterious effects that can result from afl
inaccurate test report, the test equipment must be tested immediately prior to
and following each test. In fact, it is best if the equipment can test itsel*
during the test. Each type of measurement equipment has its own peculiar
weaknesses that must be examined for each test. Radonics primarily use*
continuous working level monitors (CWLMs) and the procedures and methodologie'
that are discussed in this paper address the particular concerns of that typ*
of equipment.
A CWLM has three primary areas that must be monitored to ensure a good
test. These are filter integrity, flow and detector efficiency. The filte*
must not leak. Even pin hole leaks that are invisible or barely visible to th®
naked eye can cause a large error in the test results and in the direction ox.
yielding a false negative. Air flow and detector efficiency must be exactly
known in order to perform an accurate working level (WL) calculation.
measurement errors are directly proportional to flow errors and detecto*
efficiency errors.
The CWLM equipment used by Radonics tests the filter pressure drop during
the equipment power-up built-in-test and then continuously monitors the filter
pressure during the test. If the pressure drops below the 8% limit at any
time, error flags are set in the data file. The filter pressure is recorded
in the data file with the raw measurement data each 15 minute period.
One of the first actions in the CWLM power-up test is a detector check*
The equipment operator is requested to insert a radioactive source into the
filter holder in place of a piece of filter paper. A two minute count is then
performed on the activity of the source. The activity of this source and th®
number of counts measured in the two minute period are compared against similar
data taken at the time of the equipment calibration and stored in the internal
system calibration tables. The activity ratio must compare favorably with that
of the reference source made at the time of calibration.
Finally, the flow rate must be precisely known. This is typically tested
with either a bubble tube device or with a mass flow meter. Care must be taken
with a mass flow meter to correct for the altitude, barometric pressure and
temperature effects on the mass flow meter. The equipment used by Radonic*
includes a mass flow sensor, absolute pressure sensor and temperature sensor.
The absolute pressure sensor measures the cumulative effect of the altitude and
barometric pressure.
These sensors allow for accurate volumetric flow measurements to be made
at the initial power-up built-in-test and continually during the test period'
The results of the flow measurements are stored with the raw measurement data
for each 15 minute period. If internal flow measurement capability is not
available, then flow measurements should be made both before and after the test
and frequent checks should be performed to demonstrate that the flow for the
device is stable and does not "wander" during tests.
In addition to these three critical parameters, there are other system
parameters that are Important to system operation. These parameters should be
measured frequently in the field to ensure that proper system operation
continues. An example of another class of parameter that should be closely
watched are internal equipment reference, bias and threshold voltages. The
Radonics equipment measures these and records the results with the raw
measurement data every 15 minutes.
Also parameters that can indicate wear of parts in a measurement system
and alert the users to possible up-coming failures prior to the actual failure
are important in insuring that tests are properly performed. For example, the
equipment used by Radonics measures and records motor current. Motor current
has been shown to be a good indicator of pending motor bearing failure.
-------
. Even with all of these tests before and during a measurement period, there
8 still sometimes doubt as to whether or not a test result was correct. It
j-8 especially disconcerting to think that a test result might have been a false
**e9ative, especially a false negative that was not a few percents off of the
a^k but rather was substantially lower than the true (high) radon level.
To remove this final concern, the standard QA procedure of making
e^ur*dant measurements with independent equipment is used. At the initiation
nd completion of each test at least two WL grab samples are taken. While the
ccuracy of the grab samples may be substantially less than the accuracy of the
/WLM equipment used by Radonics, it does give the QA personnel the assurance
hat an undetected CWLM error has not occurred which would result in the
u«8tantial under-valuation of the actual radon problem.
All of this data must be recorded and transmitted back to the office for
review. The CWLM equipment used by Radonics stores all of the data
iscussed in this section (including the grab WL data which is manually
ntered) and a computer analysis of this data is available to the person
^®rforming the in-office QA function.
Multiple INDEPENDENT REVIEW
In any health or safety related decision making process there should be
review process that consists of persons close to the situation and those
®tached from the situation. There are four principle reviewers of the
®asurement data in the Radonics QA procedure.
The first review of the measurement data is performed by the measurement
-------
PUTTING IT ALL TOGETHER
The elements described in the first portion of this paper are only
major elements of the QA process. There are of course dozens of email**
details that have to be addressed to keep tight control over the QA process*
A brief summary of the job flow and the QA part in that flow is illustrated W
Figure 7 below.
The office system is the glue that ties all of the surrounding aspects
this system together. While the arrows show the flow going from one of tl>*
outer blocks to the next for simplicity, each step in this cycle actualW
begins and ends with an interface to the office system.
7 6 5
Figure 7.
1. An order is placed. The office personnel must read back critical
information to the person placing order to ensure that data has gotte#
into the computer system correctly.
2. Equipment installation and pickup date/time is established with dwelling
occupant. The EPA closed house conditions are explained to the occupant*
The 12 hours of closed house conditions necessary prior to the test ar*
especially emphasized.
3. Schedule and job information are sent by modem to test engineer*
Automatic error detection and correction techniques are used during
computer to computer phone transfer.
4. The test engineer performs equipment installation. The house is checked
on arrival for closed house conditions. CWLM equipment is checked and
installed. The engineer selects the current address from a menu on hi®
laptop. This address is associated with a job number with which all dat*
for this test will be tagged. Grab samples are taken. Questions ar«
answered and sketch of dwelling made. closed house conditions ar*
explained and written instructions are left concerning those conditions*
A toll-free phone number is left in case of questions.
5. The QA process continues after the engineer leaves as the measurement
equipment continues to log radon levels, environmental conditions and
equipment operational parameters every fifteen minutes.
-------
The test engineer performs equipment pickup. The house is checked on
arrival for closed house conditions. The engineer selects the current
address from a menu on his laptop. The matching job number is checked
against the number selected at installation. At least two more grab
samples are taken. The occupant is asked if EPA closed house conditions
were maintained and they are asked to sign a document stating that they
were maintained. The test data from the CWLM equipment is transferred
electronically to the laptop. Data echoing is used to insure 100% data
transmission accuracy.
All data is sent via modem from the engineers laptop computer to the
office. Automatic error detection and correction procedures are used.
The data is analyzed by the office computer and any problems are flagged.
A QA person having the proper access codes to the QA system, reviews the
raw test data and the results of the computer analysis of the test data.
If the results are good, a code is selected that allows the release of
the report to authorized persons.
CONCLUSION
It is clear from the preceding discussions that while the making of an
Actual radon measurement may not be difficult, the process surrounding that
Measurement is very involved. There are many areas that must be given careful
&nd detailed scrutiny. The primary considerations that must be addressed are
the maintenance of EPA closed house conditions, the careful monitoring and
Cross-checking of measurement equipment operation and the process of insuring
that results are not confused. Critical pieces of the process that must be
present, as a minimum, are a high degree of automation, measurement equipment
that records data samples every fifteen to thirty minutes and multiple levels
®f independent review. All elements of the system must be designed to work
together.
The work described in this paper was not funded by the U.S. Environmental
protection Agency and therefore the contents do not necessarily reflect the
Vi«ws of the Agency and no official endorsement should be inferred.
-------
Session III-P2:
Panel Session on Short-/Long-term Radon
Measurements
-------
III-P2-1
OVERVIEW OF SHORT-/LONG-TERM RADON MEASUREMENTS
by: Richard Sextro
Lawrence Berkeley Laboratory
Berkeley, California
-------
III-P2-2
THE RELATIONSHIP BETWEEN
WINTERTIME SCREENING AND ANNUAL AVERAGE RADON LEVELS IN U. S. HOMES
by: Melinda Ronca-Battista
Scientific and Commercial Systems Corporation
Falls Church, Virginia
Bonnie Chiles
Dennis Wagner
U. S. Environmental Protection Agency
ABSTRACT
There are a total of about 3700 homes in the country where both short-
term, wintertime (screening) and annual average measurements have been made in
the same home, and for which the results are available in the published
literature. This paper investigates the relationship between these
measurements, and presents the percent of homes with different ratios between
the screening and annual average concentrations.
The studies vary from a rigorously designed survey of 2000 homes, to
small studies of several dozen homes. The methods of measurement include
charcoal canisters and alpha-track detectors. Only that data which was from
EPA studies was available in non-summary form. An overall pooled frequency
distribution was derived by weighting the statistics from each study, and
pooling the statistics to form a new distribution. The data are from a large
number of homes, from different areas of the country, and when the parameters
from the different datasets are compared, they agree surprisingly well with
one another.
-------
III-P2-3
TEMPORAL PATTERNS OF TNDOOR RADON IN NORTH CENTRAL FLORIDA
AND COMPARISON OF SHORT-TERM MONITORING TO LONG-TERM AVERAGES
C.E. Roessler, J.W. Revell, and M.J. Wen
University of Florida
Gainesville, FL 32611
ABSTRACT
Both 12-month alpha track (AT) and monthly were
charcoal canister (CC) measurements of lndoo averacre
made in 37 Gainesville, pCi/L. Considerable
concentrations by AT ranged from 0.5 to ¦32 p / ements;
month-to-month variation was observed in v>00ied average
this pattern varied from house to house. ^ with a winter
monthly concentrations showed a_seasonal1 p f some houses,
maximum; a secondary summer maximum wasob correlated
The annual geometric means of monthly CC Saluls
well with the year-long AT measurements, b p
about 12% higher than the AT results.
This study indicates that there are gather large
uncertainties associated with estimating lon^ _ t por
concentrations from this form of short-term measurement^^For
example, a single CC measurement predicted the annu
with a 95% confidence interval of about a *for season
confidence interval is narrower if correction averaqe of
of measurement and if the estimation is based o
more than one short-term measurement.
The data were also evaluated in terms of
of various long-term Rn levels as a functio
short-term measurement.
This paper has been reviewed in accordance
with the U.S. Environmental Protection
Agency's peer and administrative review
policies and approved for presentation and
publication.
-------
INTRODUCTION
Since the health risk associated with indoor radon-222 (Rn)
is related to the long-term cumulative exposure to the airborne
radon decay products, the best way to evaluate indoor radon in a
structure is by long-term monitoring under lived-in conditions.
On the other hand, there are situations, such as in real estate
transactions and in post-construction demonstration of compliance
with building codes, with compelling reasons for assessments on
the basis of short-term testing.
The precision of a short-term test result as an estimation
of the long-term average concentration is affected by systematic
variations such as seasonal effects and by random variations
associated with the particular device and protocol. Seasonal
effects can be induced by environmental variables such as weathe*
and climate as well as by house variables such as construction
type and heating, air-conditioning, and ventilation type and
operational schedule. These variables in turn may be regionally
specific.
This study was prompted by such considerations. The Florida
Statewide Radiation Study (1), conducted in 1986-87, indicated
evidence for elevated indoor radon in parts of North Central
Florida. A winter 1986-87 single-sample screening survey of 67
houses in Gainesville, Florida, found Rn concen-trations
exceeding 20 pCi/L in 12% of the houses tested and levels in the
range of 4 to 20 pCi/L in an additional 31%.1 The State of
Florida is currently developing a Rn-resistant building code that
will have prescriptive (i.e., construction practice) provisions
and likely also will have the option for alternative construction
if performance criteria are met (i.e., Rn concentrations are
acceptable).
In this study, both long- and short-term measurements of
indoor Rn concentrations were performed in 37 Gainesville,
Florida, houses over a 12-month period from June 1987 through May
1988. The frequency distributions of the measurements were
examined, the averages of the short-term measurements were
compared to the long-term measurement in each house, and the data
were examined for temporal patterns. In addition, an evaluation
was made of the ability of various short-term measurement
programs to predict annual average concentrations.
'Unpublished, University of Florida.
-------
METHODS
SELECTION OF HOUSES
The houses studied were single-family, detached dwellings.
The primary considerations for selection were perceived depen-
dability of the participant, intent to occupy the house for at
least a year with no plans for major modification, and
construction type. About one-third of the participants were from
the participants in the earlier single-sampling screening study,
about one-third were city employees, and the balance were other
recruits and volunteers. This group of houses is believed to
have a bias toward middle income housing. Thirty (81.1%) of the
houses were built over concrete slabs, five (13.5%) were over
crawlspaces, and two (5.4%) had combination slab/crawlspace
substructures. This roughly represents the distribution of
substructure types in this region.
MEASUREMENT PROCEDURES
Long-term measurements were made with alpha track (AT)
detectors1 deployed in each house for 12 months and then
retrieved and returned to the vendor for processing.
Short-term measurements were made under "lived-in"
conditions with charcoal canisters (CCs) deployed for 4 days each
month in each house. The participants completed monthly
questionnaires in which they recorded whether the house was
closed prior to monitoring and also the conditions of occupancy,
window status, heating, ventilating, and air conditioning each
day during the monitoring period. This information was
subsequently used to identify the subset of "closed house"
measurements.
The CCs were the 4-in. (10 cm), open-face type.' CCs were
reused and were regenerated before deployment by overnight
heating at 100° C. Counting was performed at the University of
Florida using a multichannel gamma spectrometer with a 4 x 4 in.
(10 x 10 cm) Nal(Tl) scintillation detector. Radon on the
canister was determined by comparison to a standard canister
containing a known amount of radium-226. Airborne Rn
concentration was calculated by correcting for decay between
sampling and counting and by employing an effective sampling rate
based on humidity (inferred from weight gain) and length of
deployment. Duplicate CCs were deployed in at least three houses
each month.
'RADTRAK, Tech/Ops Landauer, Inc., Glenwood, IL.
JF S. J, Miami, FL.
-------
RESULTS
Of the 37 houses selected, one underwent partial mitigation
during the study and ATs from two were reported as damaged. As a
result, AT data were available from 35 houses (34 of which were
unmitigated for the entire 12 months). CC data were available
from 37 houses (36 unmitigated) and CC and AT data pairs were
available from 35 houses (34 unmitigated).
A total of 393 CC measurements were recorded; the "closed
house" subset consisted of 265 measurements.
Individual short-term concentrations ranged from <0.1 to 110
pCi/L. Arithmetic annual averages of the short-term CC
measurements ranged from 0.3 to 43.4 pCi/L for the various houses
and geometric means ranged from 0.3 to 35.1 pCi/L. The 12-month
AT measurements ranged from 0.5 to 31.9 pCi/L.
FREQUENCY DISTRIBUTIONS
Frequency distributions for the data as collected were
highly skewed to the lower concentrations. After making a log
transformation, the AT results, the CC data pool, the arithmetic
annual house means by CC, and the geometric annual house means by
CC all had frequency distributions approximating the normal
distribution. Consequently, it was concluded that the data were
reasonably described by a log-normal distribution. Thus, data
sets were described by geometric means, and statistical testing
was performed on log-transformed data. Since there were 12 or
fewer CC observations within any individual house, the sample
size was too small to determine the frequency distribution of
within house values, but it was assumed that these data would
also be more closely approximated by a log-normal distribution.
COMPARISON OF AT AND CC ANNUAL AVERAGE CONCENTRATIONS
The CC annual averages by house and AT measurements are
compared in the scatter diagram of Figure 1. Comparison of the
two data sets by linear regression yielded:
- Using arithmetic mean CC:
Y = 0.19 + 1.34 AT, R* = 0.95; and
- Using geometric mean CC:
Y = 0.22 + 1.12 AT, R2 = 0.93.
The CC results appear to be biased high relative to the AT values
by about 0.2 pCi/L. Both forms of CC annual mean correlate well
with the AT values (R* >0.9), but estimate higher values than the
AT method. In addition to the 0.2 pCi/L inter-
-------
Alpha Track (pCi/L)
Figure 1. Indoor Radon - Comparison of Geometric
Mean of Monthly Canister Measurements with
1-year AT Detectors.
-------
cept, the arithmetic and geometric CC means are higher than the
AT values by about 34% and 12%, respectively. This may indicate
a difference between the two methods. On the other hand, about a
third of the houses were part of an earlier, screening
measurement study and the participants may have consciously or
subconsciously employed closed-house conditions during CC
deployment at times when they would otherwise have had open-house
conditions.
TEMPORAL PATTERNS
General Pattern
For each house, the monthly CC values were normalized to the
respective CC annual mean (AM). Monthly minimums, geometric
means, and maximums of these normalized values are plotted in
Figure 2. This figure indicates a general seasonal pattern for
the monthly means (MM):
- Winter (Nov-Feb): maximum at MM = 1.6 AM,
- Spring (Apr-May): minimum at MM = 0.8 AM,
- Summer (Jun-Aug): MM = AM, and
- Other months (Mar, Sep-Oct): transitional.
The figure indicates considerable variation between houses.
Normalized values as low as 0.6 or less and as high as 2.0 were
seen in every month. The maximum normalized concentrations, in
addition to having a winter maximum, had a secondary summer
maximum; this suggests the possibility of a subset of houses with
a summer maximum.
Testing For Seasonal Effect
Months were grouped into equal 3-month quarters for
statistical testing. Since it was felt that the use of heating
may have a significant effect on pressure differentials and hence
Rn entry, and since the heating season in this part of the
country generally falls in the months December through February,
the seasons were designated:
- Summer: Jun-Aug, - Fall: Sep-Nov,
- Winter: Dec-Feb, and - Spring: Mar-May.
The seasonal effect was significant; the following were indi-
cated:
Summer < Spring < Fall < Winter
Significant
P = 0.11 € 0.1 level P = 0.38
Not significant Not significant
@ 0.1 level
-------
5
4-
\
3-
'n Max
\ ^ «s /
V y
¦J". TV-
\ '
J J 1 . 1 I 4tl <
M,n *.
I
Jun Jul
—i—
Nov
—i—
Dae
I
«lan
—i 1—
Fab Mar
1088
Apr Mty
Figure 2. Relative Indoor Radon Concentration By Month (Monthly CC
Normalized to Geometric Annual Mean)
-------
The statistical testing confirms the winter maximum suggested by
Figure 2. The 3-month spring and fall seasons rank slightly
differently in the statistical testing than they did in the
Figure 2; months were assigned intutively to equal-length
quarters in the former case and based on experience to variable-
length seasons in the latter case.
Patterns Within Individual Houses
Inspection of data for individual houses indicates at least
four patterns:
1. Some houses had relatively level concentrations with little
variation,
2. Some houses had a distinct winter maximum,
3. Some houses had a maximum in the July-August time frame or
had a second maximum in the summer, and
4. Some houses had highly variable results without any
systematic pattern.
Further inspection of the data for individual houses
suggests that some of the seasonal effect may be confounded with
ventilation effect (i.e., whether the house was open or closed
during the CC deployment). However, this does not appear to be
consistently the case; and in analysis of variance, interaction
between ventilation effect and season effect was not
statistically significant.
ESTIMATION OF THE ANNUAL MEAN FROM CC MEASUREMENTS UNDER LIVED-lN
CONDITIONS
The within-house root mean square error for 393 log-trans-
formed CC measurements in 36 houses was 0.825. This translates
into a geometric standard deviation of 2.28. If a small number
of CC measurements, uncorrected for season of year, were used to
estimate CC-derived annual mean Rn concentrations, the 95%
confidence intervals would involve multipliers/divisors of 5.0,
3.1, and 2.5 for estimates based on the averages of one, two, or
three short-term CC measurements, respectively. Stated
another way, the 95% confidence intervals would be:
- Single measurement:
Lower bound = 0.20 CC, Upper bound = 5.0 CC;
- Average of two measurements:
Lower bound = 0.32 CC, Upper bound = 3.1 CC; and
- Average of three measurements:
Lower bound = 0.39 CC, Upper bound = 2.5 CC.
-------
INFERENCES FROM THE CLOSED-HOUSE DATA SUBSET
ESTIMATIONS OF THE LONG-TERM MEAN
The data were analyzed to estimate the constraints as-
sociated with the use of a small number of short-term measure-
ments to estimate the long-term average in future testing. The
closed-house subset of CC measurements were taken as a represen-
tation of what might occur in future short-term testing in this
geographic region under closed house protocols. The AT results
were taken as the best estimate of the long-term averages in
these houses.
For this purpose, the statistic (Z} - Y) was defined where:
Zj = Ln CC for the jth measurement1 and
Y = Ln AT.
The data were fit to the model:
Zu - Yt = u + a4 + ei3,
where:
ZtJ = Ln CC for the jth measurement in the ith house, Yt
= Ln AT for the ith house,
u = the mean difference between Ln CC and Ln AT
measurements,
= the house effect for the ith house, and
etl = the random error.
For a future short-term measurement, the estimated long-term mean
¦"Quid be:
YiA = Zi} - uA, where uA = the estimator of u.
In terms of concentration, pCi/L, this back-transforms to:
LT(pCi/L) = C(pCi/L),
where LT = the estimated long-term mean and
c = the measured short-term value.
The data were analyzed for the case of the season not
specified and also for each of the four seasons. Confidence
intervals were calculated for estimations based on a single
Bhort-term measurement and also for estimations based on the
^erages of two and three short-term measurements. The resulting
Estimation equations and the associated 95% confidence intervals
presented in Table 1.
Note: To avoid having zero values for the log transformation, CC
taken to be the observed value + 0.01 for this analysis.
-------
Table 1.
Predictors for Long-term Indoor Rn Concentrations from Short-term, Closed-house, Charcoal Canister
Measurements
Season of
Measurement
Predicted
LT Avp.
No. of
Measurements
95% Confidence Interval of Predicted LT Ave.
Lower Bound Upper Bound
Not
0.93
1
0.21
LT
or
0.20
C
4.72
LT
or
4.39
C
Specified
2
0.33
LT
or
0.31
C
3.00
LT
or
2.79
C
3
0.41
LT
or
0.38
C
2.45
LT
or
2.28
C
Spring
1.11
1
0.23
LT
or
0.26
C
4.34
LT
or
4.82
C
(Mar,Apr,May)
2
0.35
LT
or
0.39
C
2.86
LT
or
3.17
C
3
0.42
LT
or
0.47
C
2.37
LT
or
2.64
C
Slimmer
1.22
1
0.23
LT
or
0.28
C
4.22
LT
or
5.27
C
(Jun,Jul.Aug)
2
0.35
LT
or
0.43
C
2.84
LT
or
3.46
C
3
0.42
LT
or
0.51
C
2.36
LT
or
2.88
C
Fall
0.78
1
0.23
LT
or
0.18
C
4.35
LT
or
3.39
C
(Sep,Oct,Nov)
2
0.35
LT
or
0.27
C
2.86
LT
or
2.23
C
3
0.42
LT
or
0.33
C
2.38
LT
or
1.86
C
Winter
0.62
1
0.23
LT
or
0.14
C
4.33
LT
or
2.68
C
(Dec,Jan,Feb)
2
0.35
LT
or
0.22
C
2.84
LT
or
2.76
C
3
0.42
LT
or
0.26
C
2.36
LT
or
1.46
C
LT - predicted long-term average c - measured CC value; either single measuremet
or geometric mean
Models:
LT - a, C where a. - multiplier for predicting long-term average.
95Z Confidence Intervals Lower bound - LT / b, - ( a( / b, ) C where b, = multiplier / divisor
Upper bound • LT x b, - af b, C for 952 Conf. Interval
-------
PROBABILITIES OF ELEVATED Rn AS A FUNCTION OF OBSERVED
MEASUREMENTS
Calculation of Probabilities
Calculations were made of the probabilities of long-term
wean indoor Rn concentrations exceeding various reference values
^LTr). These calculations employed the "t" statistic:
t = [Ln(LT„/C) - uA]/[Var(Zj) + Var(u*)]05.
The probability, P, of a long-term mean concentration, LT,
e*ceeding a stated reference value, LT„, is the one-tailed
Probability associated with the "tM statistic; i.e.
P(LT > LTR) = P(-t).
The quantity, t, was calculated for observed concentrations and
vaiues of LT„ in the range of 1 to 20 pCi/L for measurements made
without regard to season and for each of the four specified
®easons. The associated probabilities are tabulated in Table 2
f°r a single short-term measurement. Figure 3 presents a family
°f curves for the example case of a single short-term measurement
with the season not specified. As part of work in progress,
flgures are being developed for other cases, and probabilities
ar® also being calculated for observations based on the mean of
^ore than one measurement.
^EBlications Of The Probability Data
These data indicate the consequences of decisions based on
Yarious criteria. For example, consider the case where 4 pCi/L
ls used as the criterion for elevated Rn.
" A program might allow a 25% margin and set the acceptance point
f°r a single screening measurement at 5 pCi/L or less. Table 2
shows that the probability of the true long-term average exceed-
ing 4 pci/L ranges from 35.5% to 71.6%, depending upon the season
0f measurement.
" Policy might be considered in which a certain risk of exceeding
primary criterion would be accepted if the probability of
Exceeding a second, higher level is sufficiently low. For
Sample, for a measurement of 5 pCi/L, the probability of the
a°tual long-term value being >20 pCi/L ranges from <1% to about
^i depending upon the season.
" Another approach would be to reject a house unless there is
s
-------
Table 2. Probability of Long-term Average Radon Concentration
Exceeding Various Levels as a Function of Observed Concentration
¦JW11.T II T — — -UL _rr - — — n — **
Observed Probability of LT Average Exceeding Indicated Value
Value
pCi/L
1 pCi/L
2 pCi/L
4 pCi/L
8 pCi/L
20 pCi
A. Season
Not Specified
1
0.458
0 .164
0.032
0.003
<0.001
2
0.777
0.455
0.162
0.032
0.001
3
0.898
0.654
0.317
0.089
0.006
4
0.948
0 .776
0.454
0.161
0.016
5
0.972
0.850
0.566
0.239
0.031
6
0 .984
0.897
0.653
0.316
0.052
7
0. 990
0.928
0.722
0. 388
0.075
8
0.994
0 .948
0.775
0.453
0.102
9
0.996
0.962
0.817
0.512
0.131
10
0.997
0.972
0.850
0.565
0.161
15
0.999
0.992
0.939
0.750
0.315
20
>0.999
0.997
0.972
0.850
0.453
1
0.561
0.221
0.046
0.005
<0.001
2
0.857
0.558
0.219
0.045
0.002
3
0.946
0.753
I 0.406
0.123
0.009
4
0.976
0.857
[ n.557
0.218
0.023
5
0.989
0.913
0.670
0.314
0.045
6
0.994
0.945
0.752
0.405
0.072
7
0.997
0.964
0.812
0.486
0.105
8
0.998
0.976
0.856
0.556
0.140
9
0.999
0.984
0.889
0.617
0.178
10
0.999
0.988
0.913
0.670
0.217
15
>0.999
0.997
0.971
0.836
0.404
20
>0.999
0.999
0.988
0.913
0.556
C. Summer
1
0.662
0.260
0.059
0.007
<0.001
2
0.885
0.609
0.258
0. 058
0.003
3
0.959
0.793
0.456
0.150
0,012
4
0.983
0.885
0.608
0.257
0.031
5
0.992
0.933
0.716
0.361
0.058
6
0.996
0.959
0.792
0.456
0.091
7
0.998
0.974
0.846
0.538
0.129
8
0.999
0.983
0.884
0.607
0.171
9
0.999
0.988
0.912
0.666
0.213
10
>0.999
0.992
0.932
0,716
0.256
15
>0.999
0.998
0.979
0.867
0.455
20
>0.999
>0.999
0.992
0.932
0.607
Continued
l • •
-------
^able 2. Continued
Observed
Value
PCi/L
—
D- Fall
1
2
3
4
5
6
7
8
9
10
15
20
E« Winter
1
2
3
4
5
6
7
8
9
10
15
20
Probability of LT Average Exceeding Indicated Value
1 pCi/L
0.373
0.723
0.871
0.934
0.964
0.979
0.988
0.992
0.995
0.997
0.999
>0.999
2 pCi/L
0.107
0.371
0.582
0.722
0.812
0.870
0.908
0.934
0.952
0.964
0.990
0.997
4 pCi/L
0.015
0.106
0.237
0.370
0.485
0.582
0.659
0.722
0.772
0.812
0.922
0.964
8 pCi/L
0.001
0.015
0.051
0.105
0.169
0.237
0.304
0.369
0.429
0.485
0.692
0.812
20 pCi/L
<0.001
<0.001
0.002
0.007
0.015
0.027
0.042
0.060
0.081
0.105
0.236
0,369
0,364
0.060
0.007
0.004
0.000
0.613
I
0.059
0.007
<0.001
0.796
0.461
0.153
0.026
<0.001
0.886
0.612
0,361
0.059
0.003
0.934
0.719
0.365
0.102
0.007
0.959
0.795
0.460
0.152
0.013
0.974
0.848
0.542
0.206
0.021
0.983
0.886
0.611
0.260
0.031
0.989
0.914
0.670
0.313
0.044
0.992
0.933
0.719
0.365
0.059
0.998
0.979
0.869
0.577
0.152
>0.999
0.992
0.933
0.719
9»2
-------
Observed Concentration, pCi/L
Figure 3. Probability of Long-term Average Rn Concentration Exceeding
Various Levels as a Function of Observed Short-term Measurement -
Single CC Measurement, Season not Specified.
/
-------
exceeds 4 pCi/L, the acceptance level must be set at 2.5 pCi/L
for winter season measurements.
- An approach at the other extreme would be to pass a house
unless there is strong evidence that the long-term average is
likely to exceed the criterion (i.e., minimize false positives).
For example, to ensure that there is no more than a 10% chance of
falsely rejecting a house when the average long-term _
concentration is actually less than 4 pCi/L, the decision level
must be set at 13.5 pCi/L for winter season measurements.
- Another alternative would be a graded approach. For single^
measurements in the winter season, example criteria might De:
<2.5 pCi/L: Accept as <4 pCi/L;
2.5 - 13.5 pCi/L: Perform further testing; ana
> 13.5 pCi/L: Reject as >4 pCi/L.
-------
SUMMARY AND CONCLUSIONS
1. Frequency Distributions of Indoor Rn Concentrations - The AT,
cc data pool, and house CC annual mean data sets all had
approximately log-normal frequency distributions. Consequently,
the data were log-transformed for statistical summarization and
testing, and the geometric mean was selected as the most
appropriate statistic for summarizing these data.
2. Comparison of CC and AT Measurements - The averages of monthly
short-term CC measurements collected over a 12-month period under
lived-in conditions correlated well with AT measurements over the
same period. The annual geometric means of these CC measurements
were generally higher than the AT results - about 12% plus 0.2
pCi/L. The reason for this difference has not yet been
determined.
3. Temporal Patterns - In general, for short-term measurements
made under lived-in conditions, there was an overall seasonal
pattern highest levels in the winter (Nov-Feb), lowest levels in
the spring (Apr-May), and intermediate in other months.
However, there was considerable variation between houses. Other
patterns observed for individual houses included a) a relatively
constant concentration, regardless of season, b) a primary or
secondary concentration peak in the summer, and c) variability
without a discernible pattern.
4. Variability Under Lived-in Conditions - Due to the seasonal
effect and other sources of variation, there was a large variance
associated with the short-term measurements. The within-house
geometric standard deviation for measurement without adjusting
for season was 2.28. As a result, the 95% confidence interval on
the long-term average concentration as estimated by a single
short-term measurement (season not specified) under lived-in
conditions is the measured value multiplied and divided by a
factor of 51
5. Estimating Long-term Levels From Short-term. Closed-house
Measurements - if the year-long AT measurement is taken as a
valid representation of the long-term average Rn concentration, a
short term (4-day) closed-house CC test overestimates the LT
value by about 7.5% on the average when season of the year is not
specified, When season is specified, the average CC measurement
ranges from an 18% underestimate during the summer to a 60%
overestimate for winter measurements. However, there was
considerable variability between houses: in some cases summer
measurements overestimated the long-term average.
The single, short-term, closed-house measurement is not a
very precise estimate of the long-term average. The 95%
confidence interval on this estimate is the estimated value
multiplied by and divided by factors on the order of 4 to 5,
depending upon the season of measurement.
-------
6. Testing for Compliance With a PerfQCTance Standard - JJata
presented in this paper provide a means for evaluating the
uncertainties associated with decisions based on short-term
testing and various decision levels.
7- Further Needs - In specifying pass/fail criteria for testing,
serious thought will have to be given to determining the
acceptable level of error. It also is important to determine the
extent to which the results of this study are generally
applicable to other years, other geographic areas, and other
short-term methods. There is an urgent need for measurement
methods and protocols with lower variances than observed in
study.
ACKNOWLEDGEMENTS
Partial support and assistance for data collection were provided
by Gainesville Regional Utilities and by the Florida Department
of Health and Rehabilitative Services. Partial funding for data
analysis was provided by the Florida Department of Community
Affairs and by the U.S. Environmental Protection Agency.
REFERENCES
1. Nagda, N.L., Koontz, M.D., Fortmann, R.C., Schoenborn, W.A.,
and Mehegan, L.L. Florida Statewide Radiation Study.
Publication No. 05-029-057, Florida Institute of Phosphate
Research, Bartow, Florida, 1987.
-------
Ill - P2-4
STUDY ON THE RELIABILITY OF SHORT-TERM MEASUREMENTS
TO PREDICT LONG-TERM BASEMENT RADON LEVELS IN A RESIDENCE
David A. Hull and T. Agami Reddy
Center for Energy and Environmental Studies
Princeton University
Princeton, NJ 08544-5263
ABSTRACT
This paper addresses the statistical issues relating to the reliability
short-term measurements (during 1 -, 2 -, 4 -, and 7-day measurement
Periods) for predicting long-term radon levels in a residence. The study is
based on measured continuous daily-averaged basement radon data for an
Entire year, for a residence in Princeton, New Jersey. Day-to-day
variability, both during individual months and over the different seasons of
the year, is quantified statistically. The practical implication of the
faulting summary statistics is that, in the absence of detailed seasonal
Knowledge, it seems Impossible to make accurate long range predictions by
"Beans of observations in a single season. Quantile-quantile plots for each
°f the four different averaging periods and for different seasons of the
year are generated in order to test the probability distributions for both
normality and log-normality. Finally, a short-term measurement strategy
which involves performing measurements during each of the four seasons of
the year is studied and its superiority over a single short-term measurement
quantified in probabilistic terms.
This paper has been reviewed in accordance vith the U.S. Environmental
Protection Agency's peer and administrative review policies and approved for
Presentation and publication.*
1. STATEMENT OF PROBLEM
There is increasing concern over the quality and validity of short-term
*adon measurements in residences as accurate predictors of long-term
Radiation exposure. It is veil known (for example [1)2]) that indoor radon
concentrations vary widely from day to day and that they often show strong
••asonal patterns. Unfortunately, these patterns My be considerably
different from house to house, depending on the soil type and permeability,
the climatic parameters, the house construction, the house dynamics, and the
occupant behavior which can be to some extent stochastic. Therefore, before
*This work was funded by the U.S. Environmental Protection Agency under
Cooperative Agreement No. CR-814673.
-------
any concrete standards on radon exposure can be formulated, it is important
to develop short-term screening measurement techniques which would enable
reliable estimation of the long-term average. Unfortunately, such a study
would require taking accurate long-term measurements in a large number of
different homes all over the country, an expensive and time-consuming
undertaking. Consequently, it is not surprising that there are only a few
studies of this type in the published literature [3-6]. The objectives of
the present study, though limited to a single residence and to a single
year, will be to quantify and study the variability of indoor radon over the
year and to evaluate a new screening technique involving four seasonal
short-term measurements in terms of its ability to yield predictions closer
to the long-term average.
2. DESCRIPTION OF DATA
Princeton University has collected continuous radon measurements during
a full year in an unmitigated home in New Jersey (which we shall refer to as
H22). The radon concentration in the basement of H22 was measured by a
Wrenn chamber every 6 seconds and the data were stored as 1/2-hour averages-
For the purpose of this analysis, the data set was reduced further by
creating daily averages. The overall average radon concentration for the
year (March '88 to Feb '89) was 41.5 pCi/L with a standard deviation of
18.7 pCi/L. Thus, the daily mean basement radon levels in H22 show
significant variation on a day-to-day basis. In order to measure the
variation over the year, the data were initially broken down into monthly
averages (see Table 1 and Fig. 1). They reveal a definite seasonal patterni
with the basement radon concentration dropping to roughly half of the yearly
average for 2 months during the summer (July and August) and climbing well
above the 40 pCi/L mark during the fall. We also note that it is precisely
these 2 months in summer which have coefficient of variation (CV) values
twice or even three times those of other months. We attribute this fall in
average radon values (and thereby an increase in CV values) to increased
air-conditioner (HAC) operation due to an unusually hot summer. It must be
stated that this is a house-specific phenomenon since the opposite was
observed in another house [7].
From Table 1, we note that the winter months (December to April) have
monthly mean radon values closest to the yearly mean, while the CV values
are lowest. In fact, the standard deviations during December and January
are seen to be about half of those during the summer months and the fall
seasons. This suggests that, If single short*term measurements have to be
done, it is advisable to do so during these months (especially December and
January). Since this observation is house specific, the single most
important implication of the summary statistics given in Table 1 is that, i*1
the absence of detailed seasonal dynamics of the particular house, it seem*
impossible to make accurate long-term radon predictions by means of
observations in a single season.
Next, we chose to divide the year into five distinct seasonal groups
(Early Spring, Late Spring, Summer, Fall, and Winter) which correspond
roughly, but not exactly, to the seasons after which they are named (see
Table 2). The clear seasonal difference indicates that it would be more
meaningful to compare various short-term vs. long-term averages for each of
-------
these five periods instead of simply considering the entire year. By
proceeding thus, which is akin to a stratified sampling approach, one can
hope to decrease the error bounds. We note again that Winter and Spring
seen to be the seasons with mean basement radon levels closest to the annual
average and with lowest CV.
3. ANALYSIS AND DISCUSSION
3.1 BOX AND WHISKER PLOTS
The goal will be to compare the ability of daily-average basement radon
®easurements and those of 2-day, 4-day, and 7-day moving averages to predict
the overall seasonal average. Box and whisker plots of the data are
presented in Fig. 2. The dashed line represents the seasonal average, while
the median is shown as a solid line. The ends of the box are referred to as
the "hinges" and designate the lower and upper quantiles. Points that are
1.5 times the interquantile range are called the "inner fences." The
"whiskers" drawn from each hinge extend to the most extreme observation
Inside the Inner fence. Outlier points are shown individually. The
standard deviations for each period are given in Table 3. As expected, the
standard deviation decreases with a longer measurement period. However, the
number of outliers often does not, although their magnitude certainly
decreases. Since radon often follows periodic cycles that are several days
in length (i.e., a positive autocorrelation coefficient is present 17]),
2- and 4-day averages often result in less smoothing of the data than would be
expected. In other words, if the radon value is high on a specific date, it is
quite likely that the concentration would also be high on the previous day and
on the day after, resulting in relatively high values even when 2- or 4-day
averages are taken.
3.2 STATIONARY DISTRIBUTIONS
The next topic which has been addressed relates to the shape of the
stationary distribution of these averages. Larsen [8] found that
concentrations of air pollutants are log-normally distributed for all
averaging times. However, a recent study 16] stated that the radon
probability distributions looked more normal than log-normal. In order to
test the above claims, the data were plotted against both a normal and a
log-normal distribution using quantile-quantile plots (Q-Q plots) (see any
appropriate statistics book, for example [9]). The sorted values of the
daily averaged radon data (or their natural logarithms) are plotted on the
Y-axis of the Q-Q plots, while the corresponding quantiles of the normal
distribution (or of the log-normal distribution) are plotted on the X-axis.
Thus these Q-Q plots give good insight into the form of the distribution
since the more linear the point spread, the mors closely do the data
Approximate the corresponding probability distribution.
These plots are shown in Figs. 3 (a and b) for suaner and winter
seasons and for the 1 2 4 - and 7-day averages. For the 1-day
averages, it appears that the log-normal provides a better approximation
than the normal distribution. However, for Late Spring and Summer, it
appears that the true distribution of the data might be somewhere in-
-------
between. As the averaging time becomes longer the linearity of the data
seems to deteriorate, indicating that longer-term moving averages deviate
from both the normal and the log-normal distributions. From the quantile
plots, it seems reasonable to fit all the data using the log-normal
distribution. Although it fits some seasons and averaging periods better
than others, it is probably better to use only one distribution in order to
maintain consistency. Given the limited nature of this study, we have
refrained from performing stricter statistical tests on these probability
distributions. Future studies involving more houses and locations should
address this aspect in a more detailed manner.
3.3 QUANTIFYING SHORT-TERM MEASUREMENT RELIABILITY
The final stage in the analysis will be to quantify the reliability of
the short-term measurements in a particular season relative to the long-term
annual average. The measure of reliability will be quantified by the
probability that a random observation falls within a certain range about the
true annual average. Based on the true approximate yearly average of
40 pCi/L, the following ranges have been chosen in the subsequent analysis:
The results for each season and each averaging period are presented in
Table 4. The numbers in the table represent the percentage probability that
an observation taken from the theoretical log-normal distribution and fitted
to the actual data will be within the specified range. In other words, the
fractional area of the theoretical log-normal probability distribution of
that particular season and averaging period contained between the specified
radon level range is equal to the number cited In Table 4. For example,
a basement radon level chosen from the log-normal distribution fitted to the
Late Spring 1-day averages would have a 44% chance of being between 30 and
50 pCi/L. For better visualization of the numbers in Table 4, we present
Fig. 4 to illustrate variation of probability values at the 25% uncertainty
level with averaging period during different seasons.
From Table 4 and Fig. 4, we note that during Winter and Spring, the
chances of accurate prediction Improve as the measurement period increases
in length. For the Summer and Fall, since the seasonal average Is very
different from the yearly average, the probability of a short-term test
result falling within the allowable range 1* poor, regardless of the
averaging Interval. Therefore, a einele short.tern. ^Muren.»nfc in lllcelv tfl
be accurate only If it la measured during the right Even In the
best of situations (I.e., during Early and Late Spring and Winter) the
probability of a 7-day average test result being within 12.5% of the yearly
average is still only about 45-60%. Clearly, in H22, it is Impossible to
use a single short-term measurement to estimate the annual long-term average
to within 12.5% of the actual true value.
Radon Range (pCi/L)
35-45
30-50
25-55
Magnitude of Error (%)
12.5
25
37.5
-------
3.4 AN ALTERNATIVE SHORT-TERM MEASUREMENT STRATEGY
A previous study [4] suggested that an average of measurements taken
during two different seasons (Summer and Winter, for example) would provide
a more satisfactory estimate. In H22, this method would probably provide
better results, but it still has the problem that, unless the proper two
seasons are chosen, the results may still be poor. For example, an average
winter measurement will be quite close to the yearly average, 39.9 vs. 41.5
pCi/L. However, if this were combined with an average summer measurement,
the resulting estimate would be (39.9+24.1)/2 - 32.0 vs 41.5 pCi/L, which is
considerably worse.
In this study, ve have chosen to investigate an alternative strategy
vhereby one takes one short-term measurement during each season of the year.
In order to test this strategy, four random samples of 1000 observations
were generated from each of the theoretical distributions of the Late
Spring, Summer, Fall, and Winter seasons. Sampling without replacement was
then done from each of these four seasonal samples to form 1000 sets of four
different short-term estimates. An analysis of this data produced the
results in Table 5 which are illustrated in Fig. 5. This prediction method
clearly outperforms a single short-term average as is obvious from Figs. 4
and S. It is important to note that, even using 7-day averages, the chance
of getting a measurement within 10-15% of the yearly average is only 68%.
However, if a measurement that is accurate to within 25% is acceptable, then
the probability leaps upward to 96%. This means that, from four seasonal 7-
day short-term estimates, the true annual average radon value of this
particular house can be predicted to within bounds of ± 25% at a confidence
level of 96%.
4. CONCLUDING REMARKS
It is important to realize that the. above conclusions are based on the
analysis of only a single residence. However, some of these conclusions can
certainly be generalized. The combination of high day-to-day variations
vith strong seasonal fluctuations in radon concentrations makes it very
difficult to reliably estimate the yearly average using a single short-term
measurement. Increasing the measurement length improves the accuracy of a
seasonal estimate, but it does not guarantee that the seasonal average
accurately reflects the yearly average. Taking a measurement from each
season appears to provide the best results. Though this strategy is perhaps
not practical, the analysis based on such a strategy does provide us vith an
upper bound to the prediction accuracy one could realistically achieve from
random short-term radon measurements made in a house: i.e., viChout explicit
recognition being given to the physical forces which influence indoor radon
levels,
In addition, the calculations in this study did not include instrument
error, which ranges from 10 to 15% for the Vretm chamber used for collecting
data. Other short-term Instruments lika charcoal canisters are likely to be
•ven lass reliable. This means that, realistically, the highest accuracy
on* could hope to achiave in terms of predicting annual average Indoor radon
levels from short-term measurements is likely to be only about 50%. Not only
should similar studios be conducted in mors houses and in other
geographic locations in order to verify the conclusions presented here, but
-------
year to year variation in the same house should equally be addressed in the
framework of future studies.
5. ACKNOWLEDGEMENTS
Discussion by R. Sextro, A. Cavallo, and K. Gadsby is acknowledged.
Critical comments by R. Leadbetter and R. Mosley of U.S. EPA/AEERL are
appreciated.
6. REFERENCES
1. Nazaroff, W.W. and Nero, A.V. (Eds.), Radon and Its Decay Products in
Indoor Air, John Wiley and Sons, NY, 1988.
2. Hopke, P.K., (Ed.), Radon and Its Decay Products, American Chemical
Society, 1987.
3. Cohen, B.L. and Gromicko, N., Surveys of radon levels in U.S. houses,
Submitted to Health Physics. 1987.
4. Harley, N. and Terilli, T., Predicting annual average 222 Rn exposure,
Health Phvsics. June 1987.
5. Russell, P.A., Recommendation for Estimating the Annual Average Radon
Concentration from Short-Term Measurements, PU/CEES Working Paper
No. 109, Princeton University, May 1989.
6. Yuill, G.K. and Associates Ltd., A study of the statistics of radon
measurements in houses, March, 1989.
7. Reddy, T.A., Molineaux, F.B., Hull, D.A., Gadsby, K.J., and Socolow, R.H.,
Statistical Analyses of Radon Levels in Residences Using Weekly and Daily
Averaged Data, FU/CEES Report, Nov. 1989.
8* Larsen, R.I,, A Mathematical Model for Relating Air Quality Measurements
to Air Quality Standards, Report AP-89 US-EPA, Nov. 1971.
9. Sachs, L., Applied Statistics. 2nd Ed. Springer-Verlag, New York, 1986.
-------
Table 1 - Monthly basement radon statistics for H22
Month,
1988-1989
March
April
May
June
July
August
September
October
November
December
January
February
Days of
missing
data
10
7
0
0
0
0
0
0
0
7
4
0
Monthly
radon average
(pCi/L)
48.7
39.8
46.9
42.3
21.0
27.2
57.1
56.8
37.9
40.0
40.7
41.3
Standard
deviation
12.2
10.4
20.1
16.2
14.9
19.9
17.8
20.2
13.3
9.3
9.6
15.5
Monthly
Yearly
1.17
0.96
1.13
1.02
0.51
0.66
1.38
1.37
0.91
0.96
0.98
1.00
Coefficient
of variation
0.25
0.26
0.43
0.38
0.71
0.73
0.31
0.36
0.35
0.23
0.24
0.38
Year
28 of 365
41.5
18.7
1.00
0.45
Table 2 - Seasonal basement radon statistics
Season
Months
of year
Days during
which data
are available
Seasonal
radon average
(pCi/L)
Standard
deviation
Monthly
Yearly
Early Spring
3-4
44
of
61
44.0
12.0
1.06
Late Spring
5-6
61
of
61
44.6
18.3
1.07
Summer
7-8
62
of
62
21.0
17.7
0.58
Fall
9-10
61
of
61
57.0
18.9
1.37
Winter 11
-12, 1-2
109
of
120
39.9
12.2
0.96
Year 3
-12, 1-2
337
of
365
41.5
18.7
1.00
Table 3 - Standard deviations of basement radon
averages for different averaging periods
Averaging
Period
Early Spring
Late Spring
Summer
Fall
Winter
1-day
2-day
4-day
7-day
12.0
9.8
7.6
6.1
18.3
16.2
13.4
10.1
17.7
15.2
13.1
10.8
18.9
15.8
12.8
10.7
12.2
11.2
10.1
9.4
-------
Table 4 - Reliability of short-term vs. long-term measurements for
different averaging periods and different seasons for H22 whose
yearly basement radon level is 41.5 pCi/L. Numbers in the table
represent the percentage probability that an observation taken
from the theoretical log-normal distribution and fitted to the
actual data will be within the radon range cited.
Range of
average Early Spring Late Spring Summer
radon
(pCi/L)
Id
2d
4d
7d
Id
2d
4d
7d
Id
2d
4d
7d
35-45
36
43
50
57
22
27
33
45
07
07
07
05
30-50
64
73
81
87
44
52
61
76
15
16
15
12
25-55
82
88
93
97
62
71
80
91
25
25
25
23
Range of
average Fall Winter
radon
(pCi/L)
Id
2d
4d
7d
Id
2d
4d
7d
35-45
19
20
18
15
32
34
37
39
30-50
37
37
36
33
60
64
67
71
25-55
51
52
53
54
81
84
87
90
Table 5 - Reliability of short-term vs. long-term measurements
for different averaging periods when a mean value of four seasonal
measurements is taken. (For H22, the yearly basement radon level is
41.5 pCi/L.)
Range of
Averaging
average
Magnitude
period
radon
of
(pCi/L)
uncertainty (%)
Id 2d 4d
7d
35-45
12.5
41 51 57
68
30-50
25
74 82 89
96
25-55
37.5
89 95 98
99
-------
—i t i , p
S 4 • • 7 • • 10 11 12 1 2
1988 Month of Year 1989
Figure 1 Month by month variation of the monthly mean dally
average basement radon for H22. The dashed lines
represent + 1 standard deviation range.
-------
o
a
c
o
*o
<§
Late Spring
I
I
1
T
I
I
I
1
I
T
J
o
a
e
o
•o
<2
100
80
60
40
•H
u
a
§
•o
Figure 2 Box and whisker plots of 1 - , 2 - , 4 -, and 7 - day averages of
basement radon levels for the five different seasons
chosen. Dashed lines represent the seasonal average,
while the median is shown by the solid line.
Outlier data points falling outside the inner fences
are also shown.
-------
Norma!
Log-Normal
•r1 «o
o
o.
§
T3
A
40
20 •
0
•2 -1
O
O.
a
o
•o
•0
so
40
30
20
10
2-day avg
/
. . ••
u
a.
•o
*
-2 -1
O
e.
•o
4.0
15
3.0
25
2.0
13
4J0
M
3.0
M
2.0
1J
2-day avg
r
/
•2
4-day avg
*
-2 -1 0
Figure 3a Quantile-quantile plots of basement radon levels for different
averaging periods during Summer.
-------
Normal
Log-Normal
•J 80
U
-------
N
w
*
¦s
•8
u
a*
100
•0
•0
70
00
¦0
40
»
SO
10
0
25% Uncertainty Level
Early-Spring —1
•
Late Spr-t"ff
""" Winter
•
( Pall
-
. Summer
Averaging Period (days)
Figure 4 Reliability of single short-term measurements done
during different seasons for H22,
N
v_/
4J
tI
•§
s
£
Averaging Period (days)
Figure 5 Reliability of a protocol involving four seasonal
short-term measurements to predict long-term average
of H22.
-------
III-P2-5
BESUKCS QE SHQKEr. AMD J&HGblERM EACQN MEASUEEMEEES IE SOIL AND
DWELLINGS EI ALPHA IEACK DEXEC1QRS..
L. Tommasino and G.Torri
Laborat-orio di Misure, ENEA-DISP,
Via V. Brancati 48, 00144, Rome, ITALY
ABSTRACT
Passive radon monitoring devices, based on damage track
detectors, present attractive characteristics not only for long-
term (months to years) meastirements but also for short-term (days
to weeks) exposures in dwellings.
A gas radon permeation-type sampler based on a heat-sealed
Polyethylene bag has been extensively used in Italy for large
scale surveys of both soil radon and long-term measurements in
dwellings.
For short-term radon measurements, the radon sampling device
is obtained by using a bare cellulose nitrate detector. Any
desired sensitivity can be achieved simply by scanning a
sufficiently large detector area by the the spark counter.
Several different sets of data have been gathered so far
Using both the techniques mentioned above, which data are very
useful to illustrate the weekly, monthly and yearly variation of
the radon concentration respectively in soil and in dwellings.
INTRODUCTION
There are a wide variety of well established techniques for
the measurement of radon and its daughters (1). In the following
we will describe the radon monitoring by alpha track detection
with the specific applications in mind of large scale surveys,
i.e. those applications where damage track detectors present
unique characteristics. Large scale surveys for the assessment of
the radon concentration in dwellings are needed respectively for:
- The assessment of long-term radon exposures (months to
years) for determining the population doses.
- The assessment of short-term radon exposures (days to
weeks) for screening purpose or for investigating the
geographical variation of the radon concentration.
-------
ASSESSMENT OF LONG-TERM EXPOSURES
THE PLASTIC-BAG RADON SAMPLER.
Damage track detectors present unique characteristics for
the assessment of long-term radon exposures for their ability to
integrate over long periods of time (months to years). Radon
monitoring devices based on damage track detectors can be formed
respectively by diffusion and permeation samplers (2) .
A novel radon monitor device has been recently developed
(3,4) which is a permeation sampler formed by a heat-sealed
plastic bag made of polyethylene. Enclosed in the bag are two
LR-115 detectors held one cm apart by a plastic frame, the cross-
section of which is shown in figure 1. Aluminized polycarbonate
degraders are facing the detectors both to optimize the detector
responses and to make their surfaces conductive. Of the two LR-
115 foils, only one is used for the assessment of the radon
concentration while the other is used as the back-up detector.
The back-up detector can be made of any type of detectors,
such as chemically and/or electrochemically etched polycarbonate
or CR-39 detectors.The polyethylene bag is needed to protect the
detector from humidity, dust, thoron and radon daughters,
allowing the preferential entry of radon gas(2,3). This bag is
very important for long exposure periods both in indoor and
outdoor environments and presents the following advantages:
- low cost, simple heat-sealing;
- small size and fast sampling time;
- high permeability to radon with small permeability
to water vapor;
- no formation of glass-like or brittle like bag in
very cold weather.
SOIL AND INDOOR RADON MEASUREMENTS WITH THE PLASTC BAG SAMPLERS.
The most important contribute to indoor radon is the radon
of the soil beneath the house. Extensive investigations of soil
radon have been made in Italy in the past few years mainly using
the plastic bag sampling device (5,6). For soil radon
measurements, detectors have been placed at a depth of 0.6 m
below the ground surface. Figure 2 shows how important is the
protection of the polyethylene bag under large humidity
conditions. In this figure two replicas are shown, which have
been obtained by spark counting LR-115 foils (7) irradiated
respectively with (right-hand aide) and without (left-hand side)
protection of the polyethylene bag. The water droplets on the
detector surface can be easily seen, which droplets greatly
affect the response of the radon sampler.
The survey of soil radon in Italy commenced in November 1982
-------
by establishing hundreds of different monitoring stations mainly
in Latium and Campania, which are two regions of Italy with
rather large content of uranium in soil. Some of these field
stations are still in use after an interruption of two years
(1984, 1985). The seasonal variations of soil radon in these
stations are shown in figure 3, which reports the mean radon
concentration versus the month period for four stations in the
year of 1983, 1986, and 1987 (6). The radon concentrations have
been normalized to a month arbitrarily chosen (August 1983). It
appears clear from this figure that there is a constant trend for
the three different years. The radon concentration results always
higher in the summer and in the fall. It can be also noted that
the total radon concentration relative to the first 9 months of
1987 is 3 times higher than that of 1983. Long term variations of
soil radon indicate that indoor radon concentrations may be known
with a limited accuracy even when radon measurements are
integrated over an entire year (6).
The plastic bag radon sampler is now being used for the
Italian national survey of indoor radon, which is carried out
with the contribute of different regional laboratories. The
simplicity, the low cost of both the radon sampler and the track
counting automation-system have greatly facilitated the use of
this technique by different Italian laboratories.
Finally, radon monitoring in a roman residential home has
been made continuously since 1987. Figure 4 shows the monthly
variation of indoor radon concentration in the children bedroom
of this test house. During the cold weather registered on January
1987 (when Rome was covered by almost one meter of snow) the
radon concentration was relatively high. In the holiday period of
August 1987, the bedroom was kept closed and the radon
concentration increased as a result of a decreased air exchange
rate. Since then the radon concentration was never so high
because of the exceptional good weather, registered in Rome both
in 1988 and 1989.
SHORT-TERM RADON MEASUREMENTS
LIMITATIONS OF EXISTING TECHNIQUES FOR SHORT-TERM MEASUREMENTS.
Measurements of short-term exposures (from a few days to one
week) are particularly useful for screening surveys to identify
houses with high radon concentrations and to investigate
geographical variations. Unfortunately all the above radon gas
samplers are not sufficiently sensitive for short exposure-
periods. For these applications the Diffusion Barrier Charcoal
Adsorption Collector-DBCA is used. Cohen (8) has reported an
exercise based on the assumption of the validity of the Poisson
statistics in order to compare uncertainties (one standard
-------
deviation) between SSNTD samplers (used for three months
exposures) and the diffusion barrier charcoal adsorption method -
DBCA (one week exposure). This comparison results in the
following:
Bq/m3 DBCA Microscope
(one week) (three months)
18.5 46% 89%
These conclusions are misleading since they are based on
less than 10 mma of microscope counting area. On the other hand,
areas of hundreds of cm8 can be easily scanned in a fraction of
a second using the spark counter (7). Since it was suggested
that there was no alternative to the charcoal adsorption
technique, this latter technique has been widely used for
screening surveys in spite of the following limitations (9):
- Response with undue weight to the final day or two
of the exposure;
- Batch to batch variations;
- Response highly dependent on temperature and
humidity.
In the following it will be illustrated how simple it is to
measure short-term radon exposures in dwellings by using the
spark counter for the scanning of large detector-areas of
cellulose nitrate foils.
LOWER LEVEL OF DETECTION-LLD
Assuming valid the Poisson statistics, the Lower Level of
Detection - LLD can be related to the area of detector counted,
S, as in the following (10):
wr
Ll»- "J* «
where j is the background track density and € is the track
registration efficiency per unit exposure.
According to equation (1), if the detector area is increased
by a factor of one hundred, the LLD decreases by a factor of ten.
Since detector areas up to hundreds of cm* can be easily counted
with the spark replica counter (7), there should be no problem of
measuring short—term radon exposures by damage track detectors.
-------
THE ENVELOPE-TYPE RADON SAMPLER
To be able to use large detector areas, a radon monitor
based on bare cellulose nitrate foils appears the most convenient
(2,11). This detector presents a response directly proportional
to the radon gas concentration with little dependence on the
plate-out and the equilibrium factor (12,14).
To exploit these attractive characteristics a new radon
monitor has been developed using a large cellulose nitrate film
(LR-115 Type II with an area of 9cmxl2cm) enclosed in an
envelope (2,11). When the envelope is closed the LR-115 detector
does not register alpha-particle tracks and the radon monitor is
in the off-state. To turn it on, the envelope must be opened as
shown in Figure 5. Half of the strippable LR-115 foil is
protected by a plastic cover thick enough to stop all alpha
particles from radon and its daughters. This part of the LR-115
foil represents the "blank" detector, while the remaining bare
part is used for the radon gas measurement. After exposure, the
entire detector is etched with 2.5 N NaOH solution at 60 °C for
110 minutes. Once etched and washed, the thin cellulose nitrate
film is stripped off from the backing and both the "blank" and
the bare detector surfaces are counted with a 43 cm* electrode at
500 volts after a pre-sparking at 900 volts. This counting
procedure takes less than one second and can be repeated up to
hundreds of times with a reproducibility better than 2 % .
In spite of the very different etching and counting
procedures, the registration efficiency, €, has resulted in a
value similar to that evaluated by other investigators (13,15)
and precisely:
€ = 2.0 (Tracks . m3.h)/(cm". kBq)
Comparison of the advantages and the disadvantages of the
SSNTD and the DBCA methods for indoor radon measurements have
been reported (8) under the assumption of the validity of Poisson
statistics. When similar comparison is made between the DCBA and
the spark counted envelope detector, the statistical
uncertainties (one standard deviation) of measurements for one
week exposure in indoor air result as in the following:
Bq/m3 DBCA Spark Counter
18.5 46 % 12 %
Incidentally the small uncertainties predicted on the basis
of the Poisson statistics have been confirmed by preliminary
empirical data.
To conclude, the envelope-type radon sampler presents the
following advantages:
- Possibility of scanning large detector areas in a
fraction of second achieving very low detection limits.
-------
- Possibility to turn off and on the detector simply
by closing and opening an envelope.
- Simplicity of delivering the detector by mail.
- High sensitivity and accuracy for short term exposures.
- Unique possibility to have a "blank" detector.
- No diffusion time required for the radon to enter
the sampler volume.
Figure 6 shows the weekly variations of the radon
concentrations in the bedroom of the roman test house in the
spring 1988. Variations greater than a factor of two are
measured, which variations can not be monitored with the plastic
bag radon sampler.
REFERENCES
1. OECD/NEA Expert report "Metrology and monitoring of radon gas
and their daughters products". OECD/NEA Paris, ISBN 92-64-
12767- 4, 1985.
2. Tommasino L. Assessment of natural and man-made alpha emitting
radionuclides. Nucl. Tracks and Rad. Meas. 15: 555, 1988.
3. Tommasino L., Cherouati D. E., Seidel J. L. and Monnin M. A
plastic bag sampler for passive radon monitoring. Nuclear
Tracks 12: 661, 1986.
4. Torri G. The plastic bag radon monitor and survey results.
Paper presented at the International Workshop on Radon
Monitoring in Radioprotection, Environmental Monitoring and
Earth Sciences, Trieste, Italy, April 4-13, 1989.
5. Azimi-Garakani D., Flores B., Piermattei S., Susanna A. F.,
Seidel J.L., Tommasino L., and Torri G. Radon gas sampler for
indoor and soil measurements and its applications. Radiat.
Prot. Dos. 24: 269, 1988.
6. Torri G., Azimi-Garakani D., Oppon O.C., Piermattei S.,
Susanna A. F., Seidel J. L., Tommasino L., and Ardanese L.
Measurements of soil and indoor radon in Italy. Nucl. Tracks
and Rad, Meas. 15: 637, 1988.
7. Tommasino L., Cherouati D. E., and Raponi F. Improvements in
spark replica counter and the breakdown counter. Nuclear
Tracks 12: 275, 1986.
8. Cohen B.L. Comparison of nuclear track and diffusion barrier
adsorption methods for measurement of Rn-222 Levels in indoor
air. Health Physics 50: 828, 1988.
-------
9. Ronca-Battista and Gray D. The influence of changing exposure
conditions on measurements of radon concentrations with the
charcoal adsorption technique. Paper presented at the Thirty-
second Annual Meeting of the Health Physics Society. Salt Lake
City, Utah, USA, July 5-9, 1987.
10.Nelson R.A. Measurement uncertainties of long-term Rn-222
averages at environmental levels using alpha track detectors.
Health Physics 40:693, 1987. ,
ll.Oppon O.C., Azimi-Garakani D., Tommasino L., Torri G. and Aziz
S. Radon monitoring for short term exposures in indoor air.
Nuclear Tracks and Rad. Meas. 15, 633, 1988.
12.Rannou A., Jeanmaire L. , Tymen G. Mouden A., Naour E.,
Parmentier N. and Renouard H. Use of cellulose nitrate as
radon and radon daughters detectors for indoor measurements.
Nuclear Tracks 12: 747, 1986.
13.Makelaimen L. Calibration of bare LR-115 films for radon.
Measurements in dwellings. Rad. Prot. Dos. 7: 195, 1984.
14.Miles J.C., Stares E.J., Cliff K.D., Sinnaeve J. Results from
an international intercomparison of techniques for measuring
radon and radon decay products. Rad. Prot. Dos. 7:169, 1984.
15.Segovia H. and Cejoud J. Radon measurements in the interior of
household dwellings. Nuclear Tracks 8: 407, 1984.
wal not funded by the U. S. Environmental
The work described in do not necessarily reflect the
Protection Agency and therefor dorsejBent shouid be inferred,
views of the Agency and no official btow
-------
POLYETHYLENE-BAG DETECTOR
FIG. 1 THE PLASTIC-BAG RADON-GAS SAMPLER
-------
FIG. 2 SPARK
COUNTER REPLICA OF LR-115 DETECTORS WITH AND WITHOUT PLASTIC BAG.
-------
2.6
z
D
V
t£
<
(T
t
CD
tr
<
2.4
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1983
TT
n
n
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
I
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
rnr
1986
i—r
\
\
\
\
\
\
\
\
\
\
—
pa
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
s
\
\
\
\
\
\
\
\
\
\
\
I r
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
T
1987
X
\
\
N
\
\
\
\
\
\
\
r~\—r
J FMAMJ JASOND
J FMAMJ JASOND
MONTHS
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
S
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
I
J
l
F
I
M
I
A
1
M
1
J
1
J
1
A
1
S
1
o
1
N
FIG.3 SEASONAL VARIATIONS OF SOIL RADON
-------
1000
900
800
700
600
500
400
300
200
100
0
87 FE MA AP MY JU JL AU SE OC NV DE JA 88 FE MA AP MY JU JL
Children's Room
FIG. 4 MONTHLY VARIATION OF THE RADON CONCENTRATION IN THE TEST-HOUSE.
-------
ENVELOPE-TYPE RADON MONITOR
FIG. 5 ENVELOPE-TYPE RADON GAS DETECTOR
-------
360
340
320
300
280
260
240
220
200
180
160
140
120
100
mi inn nil mm mil m i mil i iii nm niu m tn i ii ii in ii mn i in it in mi hi n ii 1111111111111 m 1111 n ii 11 n i m i n
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1
NUMBER OF WEEKS (FEB.-MAY 1988)
FIG. 6 WEEKLY VARIATION OF THE RADON CONCENTRATION IN THE TEST HOUSE.
-------
III-P2-6
A STATISTICAL ANALYSIS:
PREDICTING ANNUAL M3Rn CONCENTRATIONS FROM 2-DAY SCREENING TESTS
S.B. White, CA. Clayton and B.V. Alexander
Research Triangle Institute
Research Triangle Park, N.C. 27709
M.A. Clifford
U.S. Environmental Protection Agency
Office of Radiation Programs, Washington, D.C.
ABSTRACT
Radon screening tests are short-term tests used to determine whether additional testing for longer
duration (usually one year) is needed to more accurately characterize health risks to Rn exposure. The
extent to which these tests are useful in reaching appropriate decisions is governed by the degree to which
short- and long-term measurements are related. A subsample of participants in the EPA/State Indoor
Radon Surveys for 1988 provided Rn measurements from 2-day charcoal canisters located in the lowest
livable level and from 1-year alpha track detectors placed on each livable level. Data from 383 houses
located in six states and representing a broad range of geologic conditions, climates, and housing types were
used to derive the relationship between 2-day measurements and 1-year measurements averaged over all
levels to represent the household annual concentration. For given short-term concentrations, the ability
to predict annual average concentrations for an individual house and the likelihood of errors in
misclassUlcation are examined.
The relationship between short-term screening measurements and long-term annual measurements
Is an important parameter in developing program policy and public information materials. This statistical
analysis is one of a series of studies being undertaken by the EPA to look at this relationship.
This paper has been reviewed In accordance with the U.S. Environmental Protection Agency's peer
and administrative review policies and approved for presentation and publication.
-------
INTRODUCTION
Short-term screening tests for radon are used to determine if additional testing (usually one year
duration) is needed to more accurately characterize health risks to Rn exposure. The extent to which
screening tests can properly identify houses needing further testing is governed by the degree to which
short- and long-term measurements are related. The relationship between short- and long-term radon
measurements has not been studied extensively. In a recent review of the published literature, Ronca-
Battista (I) found only nine studies in which this issue was addressed and six of these had sample sizes
less than 100 houses. This statistical analysis is a part of a series of studies currently being undertaken
by the U.S. Environmental Protection Agency (EPA) to better understand this relationship.
A subsample of houses in the EPA/State Indoor Radon Surveys for 1988 was tested with a 2-day
charcoal canister placed in the lowest livable level and with 1-year alpha track detectors (ATDs) placed on
each livable level. A total of 383 houses located in six states provided both short- and long-term radon
measurements. Relationships between 2-day measurements and 1-year measurements were derived for four
different housing types and for all houses combined. The ability to predict an annual concentration for an
individual house is examined. Also, the likelihood of making incorrect decisions (i.e., false positive and false
negative errors) using screening measurements is investigated.
OBJECTIVES
The purposes of this study were (1) to examine the overall relationship between short-term (i.e.,
2-day) measurements and long-term (I.e., 1-year) measurements of indoor radon, (2) to determine if a short-
term measurement can be effectively used to predict a long-term measurement for an individual house, and
(3) to assess false positive and false negative error rates associated with a short-term screening test.
METHODOLOGY
Two indoor radon measurements (X, Y) were obtained on each house participating in this study.
X refers to a 2-day charcoal canister reading taken on the lowest livable level according to the EPA's
recommended protocol. Y refers to the house annual concentration obtained by averaging 1-year ATD
readings taken on each livable level (two ATDs were used in one-story nonbasement houses). There are
several other methods which could be used for characterizing the annual concentration. One would be to
use the first floor ATD measurement Another would be to use a weighted average of the ATD
measurements from each level, where the weights reflect the proportion of time spent on each level. This
paper has been limited to the average of the measurements taken on each livable level.
The houses selected for this study were of four types • one-story with basement, two-story with
basement, one-story without basement, and two-story without basement A relationship between X and Y
was derived for each type of house. This was done In anticipation that the relationships may differ
substantially from one house type to another and, If true, separate prediction equations would be required.
In all analyses, X Is the independent or predictor variable and Y is the dependent variable.
A scatter plot of the data Indicated that Y was linearly related to X and that the variation In Y
tended to increase as X increased. The relationship between X and Y was derived using a model which
-------
reflects these visual observations in the data. A specification of the model is given below.
The results in this paper employ a mathematical model that assumes that long-term measurements
of radon (I) are linearly related to short-term measurements and (2) have variances that are proportional
to their expected values. That is,
Y, » (a + 0Xi) + cZt(a + /(X, )112 (1)
where
Yi = ATD measurement on the ith house, (average over all floors),
X{ » canister measurement on the ith house (lowest livable level),
a^,o = parameters to be estimated, and
Z/ = random error for ith house, assumed to be normally distributed with mean 0 and variance
1.
In order to convert (1) to a model having a homogeneous error structure, we divide by (a + fiXj)112 and
substitute JYi for (a + fiXf)112 on the left hand side of (1):
ft * (« + fXi)m + oZj. (2)
The parameters in (2) were estimated using nonlinear least squares. The prediction equation
^ /s ^
JTJ «¦ (« + *X)M was squared to obtain predictions of the long-term concentrations for given short-term
measurements.
rs
Similarly, endpoints of the 95% confidence Interval for the expected value of 17 were squared to
obtain a corresponding interval estimate for die long-term concentration.
DATA
Data used In this study were obtained from a tubsampie of 531 homes surveyed in the EPA/State
Residential Radon Surveys In six states (Arizona, Massachusetts, Minnesota, Missouri, North Dakota, and
Tennessee). In December 1987, the states began placing one charcoal canister and from two to four alpha
track detectors In survey eligible homes. The homeowners were Instructed to expose the charcoal canister
for a two-day period on the lowest livable level of their home during closed-house, winter conditions. At the
same time, the homeowners were supplied with enough alpha track detectors to place at least one detector
on each livable level of their home. Homes with only one livable level were given two alpha track detectors
for exposure In different rooms on the one level. Homeowners were Instructed to expose the alpha track
devices for a one-year period and were contacted by the state at the end of the exposure period to remind
them to return die devices.
For purposes of this Investigation, the data were restricted to four types of homes: one-story with
basement, two-story with basement, one-stoiy without basement, and two-story without basement. After
examining the data, three homes were excluded as outliers (one each In Missouri, North Dakota, and
-------
Tennessee). In addition, all homes with a canister reading less than 1 pCi/L or greater than 35 pCi/L were
excluded from the analysis so that the prediction equations would not be influenced by values in this range-
After applying these restrictions, 383 homes were included in the analysis. Table 1 gives the number of
homes tested and the actual number included in the analysis from each of the six states. It should be
noted that 136 homes were excluded due to canister measurements less than I pCi/L, whereas only 9 homes
were excluded due to canister readings greater than 35 pCi/L.
Table 2 summarizes the placement of the charcoal canister and the alpha track detectors in the four
types of homes. For each house, a new variable was constructed from the mean of the two or three alpha
track measurements made on the appropriate levels of the house. This mean value represents the annual
household concentration, and it is this variable which was used as the long-term measurement in this study.
RESULTS
The results of fitting the model previously described to each of the four types of houses are shown
in Table 3. For each house tvpe, Table 3 gives the prediction equation, the correlation between X and Y,
and the standard deviation, a, from the fitted model.
Note the prediction equations in Table 3 are, for all practical purposes, the same for the four
house types. The estimated coefficients of X ranged from 0£ to 0.6. Also, the correlations between X nnd
Y are similar for the four house types • ranging from 0.79 to 0.82. Finally, the standard deviations of the
fitted models for the four house types do not vary enough to warrant separate prediction equations for the
four house types. Thus, it was decided to ignore house type and to use only one relationship for all houses.
The results of fitting the model described earlier to all 383 houses is shown at the bottom of Table
3. The prediction equation is
Y = 0.42 + 0.56X
where Y is the expected (or predicted) value of the annual concentration in a house that has a screening
measurement of X. For example, for a house that had a charcoal canister measurement of, say, 3.5 pCi/I*
the predicted house annual average concentration from 1-year ATDs placed on each level would be 2,4 pCl/l.
(= 0.42 + 0.56(3.5)).
A scatter plot of the actual test results from all 383 houses is given in Figure 1. Note that most
of the houses reflect a canister measurement under 6 pCi/L and a long-term measurement of the house
annual concentration under 4 pCi/L. Based on other survey results and research findings this was expected.
Note also that as the short-term measurements get larger the long-term measurements show greater
dispersion. This increase in variability In long-term measurements is taken into account by the model used
in data analysis.
Superimposed on the scatter plot of Figure 1 are three lines. The center line Is, as noted, the
prediction equation. The other two lines (designated as UCL and LCL) represent the estimated upper and
lower 95 percent confidence limits on the predicted value for an individual house. These should not be
confused with 95 percent confidence limits on the prediction equation which are, in this case, very close to
the center line (within 1 pCi/L of the prediction equation for screening measurements less than 10 pCi/L)
because of the large sample size. The interpretation of the confidence limits in Figure 1 is as follows. If
a 2-day charcoal canister reading is X for a given house, there is • 95 percent chance that the true house
annual average concentration would be covered by the interval felling between the upper and lower lines
corresponding to X. For instance, If X ¦ 14, we can be 95 percent confident that the Interval (3.7 -14.4)
-------
Will cover the true Iong*term concentration.
As an aid in utilizing the information shown in Figure 1, Table 4 gives the predicted concentration
for an individual house and associated confidence limits corresponding to various short-term screening
measurements ranging from 1 to 35 pCi/L.
Screening tests for radon should be used only for determining If additional testing (of longer
duration) is needed. EPA currently recommends additional testing if the screening test measurement
exceeds 4 pCi/L. In this case, a perfect screening test would correctly classify a house as to whether its
annual concentration would exceed 4 pCi/L. Although there is no perfect test, one can, however, assess the
performance of a screening test by characterizing the probability or likelihood of an incorrect decision. One
of two incorrect decisions can be made on the basis of a screening measurements a screening
measurement is <4 pCi/L, one may incorrectly conclude that the house annual concentration is <4 (false
negative); if a screening measurement exceeds 4 pCi/L, one may incorrectly conclude that the house annual
concentration is also greater than 4 pCi/L (false positive).
The probability that the house annual concentration will exceed 4 pCi/L, given a specified screening
measurement, X, is given by
P [2 < (afb]Q1/2 - 2 | X]
where Z is a standard normal deviate, a and b are the estimated model parameters, and o is the standard
deviation from the fitted model (bottom line in Table 3). This probability was calculated for screening
measurements ranging from 1 to 16 pCi/L; the results are shown in Figure 2. The regions of false positive
and false negative errors are noted and the probability of an error associated with a given screening
measurement can be determined directly from the plotted curve. For instance, for a screening measurement
just below 4 pCi/L the probability of a false negative error is about 0.2. The probability decreases to about
0.05 for screening measurements near 2 pCi/L. On the other hand, for a screening measurement just over
4 pCi/L, the probability of a false positive error is almost 0J (one minus the point on the curve at 4 pCi/L).
Although the probability of a false positive decreases as screening measurements increase, it is still rather
high (about 0-32) for a screening measurement of 8 pCi/L. A high false positive error rate implies that a
large number of houses may be needlessly subjected to farther testing or corrective action.
False positive and false negative error rates can be changed by changing the screening test criterion
for making additional measurements. If the current criterion Is changed from 4 to 6 pCi/L, the false
negative rate increases and the false positive rate decreases. The magnitude of the errors under such a
change can be determined from Figure 2 by shifting the vertical line two units to the right (i.e., centered
over 6 pCi/L). Under this rule change, the probability of a false negative error would be about 0.45 for
screening measurements just under 6 pCi/L and the probability of a false positive error would be about 0.55
for screening measurements just over 6 pCi/L.
The numbers of false positive and false negative errors that actually occur depend on the distribution
of concentrations In the houses being tested. It, for example, most houses have screening measurements
under 4 pCi/L then, clearly, the number of false negative errors will be very small. The number of false
Positive and false negative errors found In the houses Included In this study are shown in Table 5. The
felse negative error rate was 3% (11 out of 354) as contrasted to a false positive error rate of 48% (84 out
of 174).
-------
CONCLUSIONS
Two-day charcoal canisters and 1-year alpha track detectors were used to measure indoor radon In
383 houses located over a six state area. Results from this one-year study show there is a strong positive
relationship between short-term and long-term measurements - on average, the annual concentration is 56
percent of the screening measurement plus 0.42 pCi/L. The results also show that annual concentration
varies widely among houses having the same screening measurement Consequently, a 2-day screening
measurement may not be a very reliable measure of the annual concentration for an individual house. For
instance, for a house having a screening measurement of 6 pCl/L, we conclude (with 95% confidence) that
the interval (1.0 - 8.2) will cover the true house annual concentration. Finally, if a screening test uses 4
pCi/L as the critical value for determining the need for additional testing, the findings of this study show
the test procedure would have a relatively small false negative error rate (about 12 percent for screening
measurements near 3 pCi/L) and a relatively high false positive error rate (about 67 percent for screening
measurements near 5 pCi/L). These error rates can, however, be changed by changing the screening test
criterion for determining when additional testing is needed.
REFERENCE
11] Ronca-Battista, M. Radon in U.S. Homes: A Summary of the Available Literature on Annual and
Screening Radon Concentrations, Season Variations, and Differences Between Floors. Unpublished
report prepared for U.S. Environmental Protection Agency, Washington, D.C., September 1989.
-------
t
I
39 *
J
30 i . - - ^
• 9 * * * jo
-------
1.00
.80
| -60
4-»
c
©
1,0
.20
False Positive
False
Negative
» I I I I L
12 3 4
5 6 7 8 9 10 11 12 13 14 15 16
Screening Measurement, X
Figure 2. Probability that house concentration exceeds 4 pCI/L as a
function of screening measurement
-------
TABLE 1. NUMBER OF HOUSES TESTED BY STATE
Number of Houses
State
Tested Used in Analysis
Arizona
102 46
Massachusetts
68 51
Minnesota
58 52
Missouri
118 69
North Dakota
119 109
Tennessee
66 56
Total
531 383
TABLE 2.
PLACEMENT
OF CANISTER
AND ALPHA TRACK DEVICES
Number
of
Houses
Placement
of
Canister
Placement and Number of ATDs
Basement 1st Floor 2nd Floor
One-story
with Basement
182
Basement
1 1
Two-story
with Basement
71
Basement
111
One-story
without Basement
104
1st Floor
2
Two-story
without Basement
26
1st Floor
1 1
Total
383
-------
TABLE 3. EQUATIONS FOR PREDICTING ANNUAL MEAN CONCENTRATIONS
FOR DIFFERENT TYPES OF HOUSES
Type of House Prediction Equation Correlation (X, Y) 0
One-story Y » 0.40 + 0.60X 0.82 0.53
with Basement (0.19)* (0.04)
Two-story Y « 0.41 +¦ 0.50X 0.79 0.31
with Basement (0.16) (0.05)
One-story Y « 0.59 + 0.48X 0.82 0.41
without Basement (0.15) (0.05)
Two-story Y » 0.42 + 0.45X 0.79 0.54
without Basement (0.33) (0.09)
A
All Houses Y = 0.42 + 0.56X 0.82 0.47
(0.10) (0.02)
* (Standard error of parametric estimate.)
-------
TABLE 4. PREDICTED LONG-TERM CONCENTRATION AND ASSOCIATED 95 PERCENT
CONFIDENCE LIMITS FOR AN INDIVIDUAL HOUSE VITH A SPECIFIED SCREENING
MEASUREMENT
Screening Long-term 95 Percent Confidence Limits:
Measurement Predicted Value Lower Upper
1
1.0
0.0
3.7
2
1.5
0.1
4.7
3
2.1
0.3
5.6
4
2.6
0.5
6.5
5
3.2
0.7
7.4
6
3.8
1.0
8.2
7
4.3
1.3
9.0
8
4.9
1.6
9.8
9
5.4
2.0
10.6
10
6.0
2.3
11.4
12
7.1
3.0
12.9
14
8.2
3.7
14.4
16
9.3
4.5
15.8
18
10.4
5.3
17.3
20
11.5
6.1
18.7
25
14.3
a.i
22.2
30
17.1
10.2
25.7
35
19.9
12.4
29.1
Note: Predicted values are based on this limited data set and this specific
type of statistical analysis. It is one of a series of studies being undertaken
-------
TABLE 5. NUMBER OF FALSE POSITIVE AND FALSE NEGATIVE ERRORS IN THE
STUDY SAMPLE USING 4 PCI/L AS THE CRITICAL VALUE
<4 pCi/L
Number
of
1-Year
Measurements
24 pCi/L
Total
Number of 2-day Measurements
<4 pCi/L *4 pCi/L Total
427
101
354 174 528*
343
84
(false positive)
11
(false negative)
90
* This includes those houses with screening measurements <1 pCi/L and
£35 pCi/L which were excluded in the analysis of short-term versus
long-term relationships.
-------
III-P2-7
CORRELATION BETWEEN
SHORT-TERM AND LONG-TERM INDOOR RADON CONCENTRATIONS IN FLORIDA HOUSES
by: Ashley D. Williamson
Southern Research Institute
Birmingham, AL 35205
WITHDRAWN BY AUTHOR
-------
Session B-lll:
QA/QC of Measurement—POSTERS
-------
B-III-1
AN Fj-RPTfW (BTQff mm FFQgSNY IMHffATINS SMffLfflg TO)
A NEW mSTRUMEMT FOR MEASUREMENT OF RADON PROGENY OMCEMKftTICN IN AIR
by: P. Kdtrappa, J. C. Denpsey, L. R. Stlfiff and R. W. Ramsey
Rad Elec Inc.
5330 J Spectrum Drive
Frederick, MD 21701 USA
A RPISU (radon progeny integrating sanpling unit) is a generic nana
given to instruments which collect radon pcogeny on a filter paper and
register the alpha radiation from the deposited progeny during the entire
period of collection. An E-RPISU is a variation of this device with an
electret located in an electret ion chamber serving as a sensor fat
registering the ions created by the alpha radiation. A conventional low-flow
rate air sanpling puqp is used to collect the progeny on a one-inch diameter
filter which is mounted on the side of the electret ion chanter (220 mL) such
that the collected progeny lonlww the air inside the chariber. The negative
ions are collected by a positively charged electret (0.127 mm thick) causing
the electret voltage to drop. The electret voltage drop which occurs during
the sampling period is proportional to the tima integrated progeny
concentration. The calibration factor for E-RPISU ranged from 1.2 to 1.6 V
per nKL-day when sampled at 1 liter per minute. The calibration procedure
and the performance of these devices are discussed. These devices have the
sensitivity needed for indoor radon progeny measurements.
IMW i '<1 Itj.CN
It is known that the radiation dose to the lungs is almost entirely due
to the inhalation of the particulate radon daughter products. These daughter
products deposit in the airways of lungs and irradiate the basal cells of
tracheobronchial and pulmonary epithelia. These cells receive dose not only
from the deposited progeny but also frcm the alpha emitted by the daughter
products formed after the deposition. This led to the concept of relating
the inhalation hazard to the ultimate or the potential alpha energy
concentration in working level units (WL). One WL is defined as the
potential alpha energy concentration of the decay products of radon
equivalent to 135,000 MbV per liter of air. Evans (1) has shown that one ttL
also corresponds to 16844 alpha frcm progeny contained in one liter of air.
-------
The radon progeny integrating sampling unit (KPISU) is an instrument
to ireasure the radon progeny concentration in WL units. These units
collect the ractan progeny on an air sampling filter paper. The alpha
radiation is registered not only during the entire period of collection but
Alan three hours after cessation of sampling. Uiis data along with the
volume of tha sample collected can readily be converted into the progeny
concentration in WL units. RPISU units have been reported which use
thsrmoluminiscent dosimeters, alpha track detectors and solid state
detectors. Their sensitivities are restricted by the fact that (a) the
detectors are email and must be located close to the filter within the rancp
of alpha radiation and (b) the area of filter paper is limited by the size of
the detectors.
An electret is a piece of dielectric material carrying permanent
electrical charge. Its effective surface charge decreases only by the
collection of ions fran the ambient air. An electret ion chamber is simply an
electrically conducting plastic chamber containing a charged electret located
at the bottan of the chanter. Electret ion chanters and their successful use
as a passive environmental radon monitor has been described fully (2,3). An
E-RPISU uses an electret ion chamber configuration to register the ions
produced by the alpha radiation emitted by the progeny collected on the
filter paper. The restrictions mentioned in the earlier paragraphs do not
apply to these devices. The filter area could be large allowing a low cost
pump to be used without worrying about the pressure drop. The detector
(electret) need not be in the immediate vicinity of the filter. The chanter
size can be relatively large.
The scientific basis of this type of RPISU was first established by
Kbtrappa(4). His goal was far determining the progeny concentrations in mine
atmosphere during a short period of sampling. The key far the application of
the devices far long duration integration required far indoor monitoring was
the recent development of very stable electrets (2,3).
DESIGN FEATURES OF E-RPISU
Figure 1 shows the design features of an E-RPISU. A standard one-inch
filter head is fitted to the side of a standard electret ion chanter (2,3).
Small openings (2 ran diameter) are provided for air entry. A spring-loaded
electret cover is used to cover the electret whan the E-RPISU is either in
storage or in transit. The left half of the figure shows the electret closed
position and the right half shows the electret open position. The filter
head is connected to a pump with a flowmeter. An alternate way is to use a
pump with a calibrated critical orifice.
CPERKTMM OF E-KPIStf
TSie initial surface voltage of the electret is measured using an electret
surface potential voltmeter (2,3). It is fitted into the bottom of the
chanter. The device is placed at a location where measurement has to be
-------
done. Bring tha E-RPISU chamber to the open position by unscrewing the top
screw cap (Figure 1). Turn on the pcwer to the air sampling jxntp. Note dcwn
the time and date of the start of tha punp. Note down the air flow rate.
After the desired period of sampling, bring the E-RPISU chamber to a closed
position (Figure 1). Note down the final air flew rate, time and date. Stop
the punp. 3fte final surface voltage of the electzet is measured, This can
be done either at the place of measurement or after taking the device to the
laboratory. The data to be noted are: (a) initial electret voltage, (b)
final electret voltage, (c) air sampling duration, (d) the average flew rate.
E-RPISU™
Electret-Radon Progeny Integrating Sampling Unit (Schematic)
Figure 1. Schematic of tha electret radon progeny integrating sanpling unit.
TrfffTCTT
Tha E-RPISU unit has to be calibrated in a standard chanter where the
progeny concentration is precisely known. Most of the work was carried out
in a well calibrated QC chanter (5). A set of three E-RPISU units ware used,
all of them with a starting initial electret voltage of about 750 V. tt»ee
ware run far a pwiod of one day with a flow rate of 1 liter per minute at a
progeny concentration of about 0.075 WL or 75 nHL. The final voltages of the
electrets ware recorded after allowing the registration of ions for an
artrtitloml period of three hours after cessation of the sanpling. Experiments
ware continued for several days until the final voltages of the electrets
dropped to about 150 V. The calibration factors were determined using the
following equation (1) *
PC - (IV - FV) / (CF x F x D) (1)
vhare PC is the progeny concentration in nKL
IV is the initial voltage of tha electret
FV is tha final voltage of the electret
-------
F Is the flow rate in liter per minute
D is the duration of sanpling in units of days
CF is the calibration factor
Figure 2 shows such a calibration curve. The CF is on vertical axis and
the average of the initial and final voltages of each measurement (also
called the midpoint voltage or MFV) is on the horizontal axis. Equation (2)
is a linear regression equation relating CF and MFV. Die correlation
coefficient is 0.89.
CF = 1.0744 + 0.000685 x (MPV) (2)
¦o
Ii
?
E
>
z
u.
o
1.7
1.6
0E
O
J-
o
2 1.5
Z
o
5
ffi 1.4
<
O
1.3
1,2
1.1
|
cj
MJBRA1
HON LINE
rt
ri.
E-RI
jn
>ISL
y
u
J
<
o
o
<
>
o
-6 1
IO "
u
o
y
O '
1 0
o
>
0
<
r
o
100
200 300
400 500 600 700 800
-'MIDPOINT VOLTAGE (MPV) IN VOLTS
Figure 2. Calibration line for electret progeny integrating sanpling unit.
The air sampling rate is 1 liter per minute.
PROCEDURE for actual measurement
In practice it is not always convenient or possible to wait an
additional three hours after the cessation of the sampling for the
registration of the alpha radiation of the tail end. It is possible to apply
/
-------
a correction for this part. If DV Is the voltage drop over a sampling period
of dt days, then the average rate of change of voltage is DV/DT, which can be
taken as constant and hold good at the end of sampling. Often the toted
expected tail-end voltage drop (TD) during the tail-end portion is given by
equation (3).
TO = (DV)/(DT) x exp (- x t) dt
= (DV)/(OT) x 0.03 (3)
vihere is the decay constant for the deposited radon progeny is about
33.3 per day. This is arrived at by assuming the half-life of the
deposited progeny as about a half-hour or 0.0208 days.
Therefore, the total corrected voltage drop (CDV) is given by equation
(4).
CDV « Uf + (DV)/(OT) x 0.03 (4)
let us assume the following values for a typical measurement*
IV-700 V; FV®650 V; EH^l.25 days; F-1.00 liter per minute.
Calculated parameters are:
MPV-675 V; CF-1.537; DV-50 V; CDV*51.2 V.
Result is:
PC - 51.2/(1.537 x 1.25 x 1.00) - 26.6 nML
- 0.0266 WL
The average radon progeny concentration is 0.0266 WL.
Note that FV is the final voltage measured without waiting for the
registration of tail-end ionization.
FBKTOfflroCE ffiBST
The performance of this device has been tested by the U.S. Environmental
Protection Agarcy as part of the services provided by them far a nmr device.
A unit was sent by null to the EPA evaluation branch in las Vbgas and was
evaluated at different progeny concentrations, at different humidities and at
different conrtanflatixm nuclei counts. The report concluded that the
instrument gave a satisfactory performance and a measurement was never more
than 12% off from the chamber value (6).
Itoo radon measuring ccqpanies entered the U.S. EPA (RMP6) program with
these devices for single blind testing. The E-RPISU units were sent and
received by mail. Measuring companies sent their results to the EPA where
these results were compared to the chanter values. Table-1 gives these
-------
results. absolute value of relative errors ware in the range of about
5%, much better than the EPA requirement of 25%.
As part of thB international intarccrrparison study (blind test) conducted
by the Department of Energy Environmental Measurements Laboratory (DOE-EML),
four E-RPISU units were entered. Results are also shewn in Table-1. As can
be seen the absolute value of the relative errors were in the range of 2% to
3%.
TABLE-1. PERFORMANCE EVALUATION OF E-RPISU
E-RPISU RESULTS OF SINOE BLIND TEST
CONDUCTED BY EPA-RMP-RCXIND 6 (1989)*
Company
Code
Detector
Number
Measured
Value
(MI)
Target Absolute Value
Value of Relative Error
(TI) [ (MI-TI)/TI]
HJTLJ RP
HEHBH RP
793975
0.029
0.031
0.065
793976
0.031
0.031
0.000
793977
0.000
0.000
Blank
793978
0.033
0.034
0.029
793979
0.031
0.034
0.088
Mean
0.0455
883570
0.030
0.031
0.032
883571
0.045
0.047
0.043
883572
0.049
0.044
0.114
883573
0.000
0.000
Mean
0.063
SINGLE BUND TEST CONDUCTED EY DOE-EML (NEW YORK)
AS PART OF DUERNKTICNAL SfEERGOMPARISCN (1989)*
fti
571
610
547
603
0.136
0.144
0.135
0.134
0.137
0.137
0.137
0.137
0.007
0.051
0.015
0.021
Mean
0.024
* NOTEi Obese results are in the process of publication by the U.S. EPA and
DOE-EML, respectively.
-------
mm mwsjs
There are two sources of possible error. Cue Is the random error
associated with the dimensional variations, flow rate variations and other
system errors, As seen from the performance (Table-1), this appears to be
small and can be taken as about 5%. This was confirmed by running a large
number of units in the same location. Another error is in the measurement of
surface voltages. The measurement can be done only to within an accuracy of 1
V. When a difference between the two measurements is taken, as Is always the
case, the expected error Is 1.42 V (square root of 2). This error can be
substantial If the total measured voltage difference Is small. Let us
sample for one day at a concentration of 5 nWL at a flew rate of 1 liter per
minute. The expected voltage drop is about 7.5 V, and the error in voltage
measurement itself is 19%. This combined with the system error using the
principle of quadrature comes cut to be about 20%. This error can be reduced
by sampling for a longer period or by sampling at a higher flow rate.
Similarly at a concentration of 20 nHL, the error Is about 7%.
SHORT DURATION MEASUREMENT
If one wishes to make a short-duration measurement — less than cone day-
- then it is necessary to allow the tail-end portion of the Ionization to be
registered by allowing a delay of at least three hours. The correction
equation (4) becomes lees accurate.
It Is possible to increase the sensitivity of the device by using a
different electret of larger thickness (2,3). If another available 1.524 inn
thick electret is used, the sensitivity Is increased by a factor of 11. If a
short duration sampling Is desired at a lew concentration, then this electret
can be used in the same way. The CF will be 11 times the CF conputed by
using the equation (2).
DgMMIC RANGE m? fCTfftTTffi
The electret can be used down to a surface voltage of about 150 V. The
calibration factor does not hold good thereafter. The upper limit of the
surface voltage Is about 750 V (2,3) in the chanter used in this wade. This
leaves a hsadrocm of 600 V. Therefore, an E-RPISU goes out of range when the
total sample exceeds about 400 nKL-day. This corresponds to a saupling time
of 20 days at a concentration of 20 nHL. In other wards, one can use one
electret for about 20 measurements In the case considered.
DISCUSSION
Fran the results of blind tests, it can be seen that the E-RPISU can be
used to make an accurate measurement of radon progeny concentration. The
units can be used as a mailable unit similar to the other passive monitors
such as charcoal detectors or E-PERMb. Electret ion chambers are known to
give good performance even at extreme temperatures and humidities encountered
-------
In the indoor environment.
The electrets and the measuring device used with standard E-FERM radon
monitors (2,3) ace usable with the E-RPISU and hence, at a modest additional
cost, it can be a useful addition to the laboratories already using E-PERMb.
The electrets have to be handled with soma care which is described in our
earlier work (2,3). Proper air flow calibration has to ensured as is the
case with other KPISU units.
Kotrappa et al (4) have described how these devices can be used to
measure both radon and thoron progeny concentrations. Further, a method is
also presented (4) to measure the effective radon progeny concentration
taking into account the contribution frcm the thoron progeny. That
methodology is equally applicable to the present instrument.
-------
The \«ork described in this paper was not funded by the U.S. Environmental
Protection Agency, and therefore the contents do not reflect the views of the
Agency and no official endorsement should be assumed.
The work was performed under partial support under DOE Contract No. DE-
AC07-861D12584 far tte U.S. Department of Energy, Grand junction. A detailed
report (D0E/ID/12584-UNC/GJ-TMC-5, May 1989) on the developmental work is
available as a publication from the U.S. Department of Commerce. We are
thankful to Mrs. Lu Markland for editorial assistance.
BEEEEEEKSS
1. Evans, R. D. Engineers guide to the elementary behavior of radon
daughters. Health Plysics 17<229, 1969.
2. Kotrappa, p., Dempeey, J. C., Hickay, J. R. and Stieff, L. R. An
electret passive environmental 222-Rn monitor based on ionization
measurement. Health Physics 54i47, 1988.
3. Kotrappa, p., Denpsey, J. C., Ramsey, R. W and Stieff, L. R. A
practical passive environmental monitor for indoor radon measurement.
Accepted for publication in Health Physics 1990.
4. Kotrappa, P., Dua, S. K., Gupta, P. C., Pimpalle, N. S. and Khan, A. H..
Measurement of potential alpha energy concentration of radon and thoaxm
daughters using an electret dosimeter. Radiation Protection Dosimetry 5i49,
1983.
5. QC counter. Operated by Radon QC, Palmer, PA 18043 USA.
6. Personal canmnniration from Richard Hopper. Operational evaluation of
electret radon progeny integrating sampling unit. Fab 1989.
-------
B-III-2
A REVIEW OF THE DETECTION TECHNOLOGY IN THE AT EASE RADON MONITOR
by: William E. Simon, MS
Sun Nuclear Corporation
Melbourne, Florida
Michael B. Schell
Residential Division
Honeywell, Inc.
Golden Valley, Minnesota
ABSTRACT
The AT EASE radon monitor is a relatively low cost device which
utilities patented technology developed by Sun Nuclear Corporation under an
SBIR contract with the U. S. Environmental Protection Agency. The device has
been manufactured and distributed since August 1987. This paper is a
description of the design and detection technique of AT EASE. At the heart of
the device is a measurement circuit and pulse processing software which
records PO-218 and PO-214 alpha events on the diffused junction surface of an
inexpensive photovoltaic detector. The device has a nominal background
counting rate of 0.1 counts per hour. A description of the noise elimination
circuit is provided along with traces of alpha pulse shapes and pulse height
spectra. Also defined is the display parameters and calibration technique.
Test results on linearity, constancy, statistics, and diffusion time are
provided.
-------
B-III-3
A COMPREHENSIVE RADON ASSAY SYSTEM
USING CELLULOSE NITRATE FILMS
by: Anders Damkjaer & John L. Jorgensen
Department of Electrophysics
The Technical University of Denmark
DK-2800 Lyngby, Denmark
Milton E. McLain & Michael J. Shymanski
Department of Nuclear Engineering
Texas A&M University
College Station, Texas 77843-3133
ABSTRACT
A comprehensive radon assay system based on track produc-
tion in cellulose nitrate films for the measurements of radon
concentrations in air is described in detail. Specifically
addressed are: 1) optimization of detector size and packaging
Vs. performance and processing considerations; 2) details of
the chemical etching process for track visualization in exposed
films; 3) microcomputer-based image analysis as applied to
track counting; 4) effects of etch process parameters on track
size distribution; 5) method for correction of observed track
density for non-standard etch conditions; 6) radon-222 cali-
bration procedures; 7) precision vs. radon concentration and
Exposure time considerations; 8) recommended quality assurance
Program; and 9) overall assay time/cost factors.
-------
I INTRODUCTION
The aweu^iess towards the population exposure from the
presence of Rn progeny in indoor air has led to a demand for
simple and reliable radon assay systems. Most national surveys
have found a log-normal distribution of the radon concentra-
tions in dwellings (1),(2). This indicates that a small but not
negligible fraction of the dwellings have radon concentrations
so high that remedial actions should be taken. A large number
of dwellings must have their radon concentration measured
before the high level cases are found. Hence, the radon assay
system must be a low cost system and yet it must yield a
reliable estimate of the annual average radon concentration in
order to avoid both false positives and false negatives (3).
Radon assay systems based on Solid State Nuclear Track
Detectors are well suited for that purpose. A two month expo'
sure time yields a sufficient number of alpha particle tracks
for the radon concentration measurement. This exposure time i*
also necessary to average out the effect of changing meteoro-
logical conditions and daily variations in the radon concen-
tration.
The radon assay system described in this work is based on
cellulose nitrate film as the alpha particle track detector•
The radon assay system has been used since 1984 for rutins
measurements of radon in dwellings. A large number of measure-
ments have been carried out on requests from private hom®
owners and institutions and a survey, involving 200 radon
dosimeters, has been made in a local area in Denmark (4).
II DETECTOR SYSTEM
DESIGN
The radon assay system is based on the closed dosimeter
(detector) concept employing cellulose nitrate film,(Kodak £*
115 II), as the alpha track detector. Closed dosimeters hav*
been proven superior to open dosimeters due to better protec-
tion of the alpha track detector during extended exposur*
periods as well as to a more well defined air quality ins id®
the dosimeter (5),(6). The present assay system utilizes *
cylindrical metallic container (figure 1), which provide0
adequate mechanical strength and shields the alpha trac*
detector against infrared radiation. In addition, the metallic
can ensures that electric fields due to static charges will no*
be present inside the dosimeter. Random static fields coul^
affect the distribution of the radon progeny inside the dosi"
meter, and thus contribute to the observed data scatter.
-------
A rectangular piece of the cellulose nitrate film, 1 cm x
2 cm, is fixed to the underside of the dosimeter's circular
lid. Radon gas enters the dosimeter by passive diffusion via a
5 mm diameter hole in the lid, just beside the alpha track
detector.
A filter paper membrane covers the diffusion hole. This
ensures that radon daughters and pollutants in the form of
aerosols will not enter the dosimeter and subsequently produce
Unwanted tracks or undesirable chemical reactions during the
®tch process.
Optimum dimensions for a cylindrical dosimeter depend on
the efficiency function for the alpha track detector, i.e. the
track forming efficiency as a function of alpha energy and
angle of incidence (7). The dosimeter depicted in figure 1
Yields a maximum, efficiency for alpha particle detection from
the decay of Rn and Po in the dosimeter volume, while
still registering a larga fraction of the alpha decays from
Slate out of Po and Po on the inside walls of the dosi-
meter.
The membrane covering the diffusion entrance hole of the
dosimeter introduces a time constant for the establishment of
Jadon concentration equilibrium between the interior of the
dosimeter and the outside air. Standard laboratory filter paper
***§ a diffusion constant for radon gas of the order of 0.002
/s. For the present dosimeter this corresponds to a time
Constant of 1.2 hours and an equilibrium radon concentration
inside the dosimeter can which is 99 % of the concentration
°utside the dosimeter.
5 m hole Alpha track
with datcctor
filtarpapar inaida lid
Figure 1. Radon dosimeter
with Kodak LR 115 II as
the alpha track detector.
Figure 2. Etch cell with
alpha track detector. The
cell is submerged into a
constant temperature
water bath.
-------
OPERATION
The detector material LR 115 II, is delivered from Kodak
in boxes containing 25 sheets, each measuring 9 cm x 12 c»
(Ref). Sheets of film are sandwiched between paper of low alpha
activity in order to prevent damage to the sensitive layer and
to maintain low background counts. For long time storage in the
laboratory, (i.e. 1 - 2 years), the material is kept in its
original box and stored in a refrigerator. Detector preparation
starts with cutting the sheets into rectangles, i.e. 1 cm x 2
cm, without removing the protective paper. Again, this protects
the sensitive surface against contamination which may affect
the subsequent etch process. For mass production of similarly
sized detectors, a machined punch die of the desired dimensions
could be fabricated. However, for the laboratory-scale studies
described here, a small cutting board designed for photografic
paper is convenient. Next a 1.5 mm hole is punched in one end
of the detector which will be used for holding the detecto*
strip during the etch process. Then, the detector is mounted on
the inside of the metal can lid with an asymmetric double-sided
adhesive tape strip, which attachs firmly to the lid and less
firmly to the detector. Finally, an identification number i*
scribed into the alpha track detector surface and the dosimete?
can is closed.
If the dosimeter is to be exposed immediately afte*
assembly, a can with a press-on friction-fit lid which is not
completely radon-tight, is sufficient. However, if the dosi'
meter is to be stored for some period before use, a different
type can is used which is sealed with a conventional can lid
using a machine designed for that purpose. For these latte*
dosimeter cans, the membrane opening is covered with a radon'
tight, aluminized Mylar pull-tape which is removed at the start
of the exposure.
Public survey measurements, using the radon assay syste*
described, can be carried out by sending the dosimeters through
the postal system. A standard 20 cm long cardboard mailing tub0
will contain up to three dosimeters including instruction
papers and return labels.
ETCHING
Following exposure and return to the laboratory, alpft®
track detectors are etched for 140 minutes in 2.5 N (10 %)'
analytical grade NaOH, at 60 degrees centigrade. This etch tin1'
is appropriate for LR 115 II which has been stored in tftj
laboratory for more than half a year. However, it has be*p
observed that freshly delivered detector material often
quires up to 20 % longer etch time. A homogeneous etching
obtained when the etch bath is absolutely quiet i.e. witho^*
stirring and in absence pf convection currents. In order t°
-------
achieve this condition, each detector is etched in an indi-
vidual 10 ml glass vial submerged in a temperature controlled
Water bath. Figure 2 shows a vial with an alpha track detector
in position for etching. The Teflon stopper fits loosely into
the vial and the detector hangs freely on a stainless steel
hook attached to the inside of the stopper. A small permanent
magnet is fixed to the top of the stopper for easy handling of
many detectors at one time, or if preferred, on an individual
basis. After etching, the detctors are washed for 30 minutes in
a stirred bath of demineralized water at 30 degrees centigrade,
followed by a 2 minute, non-stirred bath of 50 % water/iso-
propanol. After a few minutes of drying in the air the detctors
are ready for track counting.
track counting
The alpha track detector LR 115 II is well suited for
automatic track counting with image analysis systems. The dark
red dyed cellulose nitrate material comprises a 12 micrometer
thick layer bound to a 100 micrometer thick, colorless trans-
parent polyester film backing. This provides desirable mechan-
ical strength for handling. After etching, a large fraction of
the alpha tracks have penetrated completely the red dyed
cellulose nitrate sensitive layer. When viewed under a trans-
mission optical microscope, these tracks appear as bright spots
on a red background with diameters up to 20 micrometers.
The track counting equipment consists of a microscope with
a CCD video camera, a "framegrabber" board (video image digi-
tizer) and a personal computer. The camera is a Philips type
56472 based on the CCD chip NXA 1011 which has 576 picture
lines with 604 pixels each. Other CCD cameras with less reso-
lution have also been employed. Several models of PC-based
framegrabbers are now commercially available. The present radon
assay system uses the "PC Vision - Plus** framegrabber from
Imaging Technology, Inc. or a multipurpose framegrabber de-
signed and built by one of our laboratories.
A low magnification microscope is used as the optical
setup. The objective is a Will Wetzlar 4/0.10. No eyepiece is
Used and the effective magnification is adjusted by selecting
the proper distance between the objective lens and the CCD chip
in the camera. The alpha track detector is held flat in a brass
frame placed in the focal plane of the microscope. An ordinary
incandescent lamp is sufficient as the light source for the
Microscope. However, an infrared filter excluding wavelength
above 735 nm must be used to avoid saturation of the CCD -
camera which is rather sensitive in the infrared region of the
spectrum. In addition, the optical system is fitted with a
green filter with transmission in the range of 515 nm to 565
ha. This filter, combined with the red dyed alpha track detec-
tor, renders the background area essentially black, creating
-------
extraordinarily good contrast for the tracks in the video
image.
The combined effect of a homogeneously etched alpha track
detector and a very high contrast between the tracks and the
background allows for a simple and fast track counting algo-
rithm. First the digitized image is binarized with a fixed
threshold. This is followed by a segmentation procedure where
the tracks are found and the track sizes determined. The
algorithm, which is written in the C language, counts the
tracks in the field of view within a few seconds.
ETCH VARIATION CORRECTION
In spite of a well controlled etch procedure some fluctu-
ations in the bulk etch rate within a batch of detectors are
regularly observed. Such inconsistent etching can adversely
affect the results of the radon assay due to induction of
Nnon-standardM track hole size distributions. The degree of
etch can be quantified by measurements of the thickness of the
remaining layer of cellulose nitrate on each detector after
etching. The thickness is determined by the transmission of
light of short wavelength through the red detector film. The
present radon assay system uses a photometer consisting of a
mercury vapor light (Philips TL 4W/33 Xm8) as the light source
and a photo resistor (CdS, type T 906021) as the light detec-
tor. Also, taking into account the previously mentioned need
for adjustment of etch time with increasing detector film
storage time, it is clear that the etch time itself is not a
reliable reference parameter. Hence, for the present radon
assay system, a reference photometer reading is chosen in order
to define the standard etching. The radon exposure calibration
factor for the assay system is determined as a function of the
degree of etch, (i.e. the photometer reading), and individual
calibration factors are used for the alpha track detectors
according to their photometer reading.
CALIBRATION
Calibration of the alpha track radon assay system is
accomplished with reference to a set of calibrated scintilla-
tion cells (Lucas flasks). The calibration system consists of:
1. A radon reservoir made of a 200 liter closed steel drum
containing 26 kg. of finely ground uranium ore (400
ppm). The equilibrium radon concentration in the drum is
approximately 100 kBq/m .
2. An exposure chamber, 5.8 liter cylindrical steel vessel
with a radon tight lid (0-ring sealing) and two Swagelok
self-closing quick-disconnect fittings for attachment to
the radon reservoir drum.
-------
3. A set of calibrated scintillation cells.
A batch of up to 12 dosimeter cans is placed in the
exposure chamber and the lid is tightly closed. The exposure
chamber is connected to the radon reservoir and radon rich air
is circulated for a few seconds using a small-capacity, diap-
hram-type air pump. The radon reservoir is then disconnected
and a scintillation cell is connected to the exposure chamber.
The radon-laden air is then circulated between the cell and the
exposure chamber for about 3 minutes in order to homogenize the
radon concentration. Finally, the exposure chamber is discon-
nected and the scintillation cell is counted to determine the
initial radon concentration in the exposure chamber. During the
hext 4 hours a diffusion equilibrium is established in the
exposure chamber between the radon concentration inside and
outside the dosimeter cans. This makes the true start concen-
tration for the exposure lower than that initially measured,
and the appropriate correction for radon dilution by the can
air volume must be calculated. After an exposure time of
typically 3 days the vessel is opened and the dosimeters are
again allowed to reach equilibrium before the alpha track
detectors are removed for etching. The procedure described here
typically leads to an exposure in the range of 500 kBq hours
per cubic meter.
Ill RESULTS
Figure 3 shows the results of a calibration run involving
12 alpha track dosimeters. All dosimeters received a radon
exposure of 1359 kBqhm . The alpha track detectors were
deliberately etched at progressively longer etch times ranging
from 140 to 195 minutes in order to determine the relationship
between the calibration factor K and the photometer reading
(PMR).
PMR, photoMtar reading [*icrow»p«r«»J
Figure 3. Calibration factor K versus etching (PMR)
-------
Track counting was performed with a resampled picture
field of 288 lines, each with 400 pixels^ The area of the field
viewed by the microscope was 0.0129 cm and up to 10 fields
were counted on each detector in order to obtain a minimum of
300 tracks observed per detector. The PMRs for the 11 detectors
shown (one detector was over-etched) fall in the range from 60
to 81 microamperes. The measured calibration factor K as a
function of PMR is approximately linear in the low etch region
from 60 to 70 mA. In the high etch region from 70 to 80 fih, the
slope increases, and above 80 nA the detectors begin to show
signs of over-etching, i.e. the red dyed layer of cellulose
nitrate may completely dissolve or separate from the colorless
substrate film.
The reference etching for the alpha track detectors for
the radon assay system is defined as the etching which gives a
PMR of 64 mA. The current value for the calibration.factor for
the present setup is K = 2.16 [(tracks/cm )/(kBqhm ) ] and the
sensitivty, (i.e. the slope of the calibration curve at PMR *
64 nA), is found to be S » 3.4 [%/juA]. As evident from figure 3
the calibration curve has no plateau on which the calibration
factor is independent of the degree of etch. This is further
illustrated in figure 4 which shows the track hole size dis-
tribution as measured by the present setup. The distribution
for the group of 5 detectors in the low etch region (60 - 70
/iA) has been compared with the distribution for the similar
group in the high etch region (70 - 80 /iA). within each group
the data have been averaged in order to obtain sufficient
statistics.
tracXs/ca2
JcBq h m-3
K
<
k 0.6
il
«
*4
s 0,4
o
m _ _
a 0.2
H
S
• 200 400 <00 800
tj track «r«a (pa3]
I I— high «tch d*t*ctor«
1^1 — low «tch datactors
Figure 4. Track size distribution for low etch
detectors compared with the distribution for
high etch detectors.
-------
The results depicted in figure 4 are expressed as the
contribution to the calibration factor from tracks in the
specified size region. The two distributions have similar
shapes with indications of a local maximum around the average
track size. The high etch distribution is shifted towards
larger track sizes compared to the low etch distribution and in
both cases the largest contribution comes from the smallest
tracks. Hence, a prolonged etching not only increases the
track sizes, it also continues to develop embryonic tracks into
countable tracks.
Fast exposure facility
The calibration procedure involving the exposure of
dosimeter cans to radon gas in a closed vessel is a time
Consuming process. It is obviously more convenient to place the
alpha track films in contact with an alpha active material in
order to obtain an exposure of the track detectors. This
Exposure technique can not replace a radon calibration, how-
ever, it can serve as a fast and reliable check of the detector
Material and the etch process. In the present work we have used
ft piece of depleted metallic uranium for this purpose. The
energies of the alpha particles emerging from metallic uranium
ftre lower than those energies from radon and its progeny. This
Accounts for small discrepancies in the recorded track size
distribution for the alpha track detector. Hence, the track
forming efficiency as a function of the degree of etch must be
Separately established for the "uranium exposure" of the alpha
track detectors. At the reference etching, (PMR - 64 fih), the
track detector records alpha particles from metallic uranium at
ft rate similar to that of a track detector housed in a current
dosimeter package placed in an ambient radon concentration of
$4000 kBqm .
iNTERCOMPARISONS
The present radon assay system has participated in the CEC
feadon Intercomparisons since 1984 (5),(6). This includes the
Second (1984) and the Third (1987) intercomparison and "The
Urst small scale" (1989) intercomparison. The radon assay
System has also passed the US-EPA National Radon Measurement
Proficiency Program, Test Round 6 (1989) • The CEC Radon Int-
trcomparisons have been carried out by The UK National Radio-
logical Protection Board (NRPB) who has a 43 m hermetically
Sealed environmental chamber at its disposal* In this chamber
the ventilation rate, the humidity, the temperature andL.the
Aerosol composition of the air can be controlled. The Rn
Concentration in the chamber can be maintained at constant
levels ranging from 4 kBqm to 15 kBqm" . In addition the NRPB
has a "Fast Radon Exposure Device" (FRED) consisting
a-0.2 m closed vessel where a radon concentration of 60
kBqm"' can be maintained.
-------
A total of 24 laboratories participated in the last two
CEC Radon Intercomparisons of alpha track radon dosimeters. All
participants mailed their dosimeters to NRPB for exposure and
future return to their laboratories for evaluation of the NRPB
exposure. Subsequently, the participating laboratories for-
warded their results to NRPB who eventually revealed the radofl
exposures given to the dosimeters.
NRPB intercomparison results for the present radon assay
system are presented in table 1. Several different exposure
conditions were employed and equilibrium factors (F) from 0.2'
to 0.88 were obtained. In addition to the test chamber expo"
sures, two long term exposures in occupied dwellings were
included in the intercomparison studies.
Evident from table 1, is the good agreement between
exposures stated by the NRPB and those found by the radon assay
system. Reproducibility of results by the current system wa*
also found to be independent of the exposure conditions. Thi*
is further illustrated in figure 5 where the exposures a*
determined by the radon assay system are plotted versus tl)*
NRPB exposures. The deviations are of the order of 10 % with '
maximum and minimum of 20 % and 3 %, respectively.
r>
'¦
a
1
J
3
e
I
1
• /
2000
/ •
1000
-
300
•
100
IP 1
1
1
100
300
1000
2000
-1
NRPB •xpoaura [kBq ha4]
Figure 5. CEC Radon Intercomparisons.
The results for the radon assay system
is shown versus the NRPB exposures.
-------
1984 NRPB radon exposure
Radon Assay System
Run
time
location
F
exposure
exposure
N
st.dev.
#
[hours]
[kBqhm ]
[kBqhm ]
[%]
1
1344
occp.house
0.41
147
118
10
11
2
41.25
env.chamb.
0.81
198
176
10
11
3
21.9
-/-
56
47
10
19
4a
2.0
FRED
82
84
1
4b
3.75
-/-
171
140
1
4c
7.75
360
303
1
4d
16.33
771
674
1
4e
31.75
W j <¦*
1803
1742
1
1987 NRPB radon exposure
Radon Assay System
Run
*
time
[hours]
location
F
exposure
[kBqhm ]
exposure
[kBqhm" ]
N
st.dev.
[%]
1
2
3
4
1512
3.42
16.25
163
occp.house
env.chamb.
-/-
0.44
0.24
0.37
0.88
197
37
208
2120
210
41
186
1874
10
10
10
10
13
24
15
5
1989 NRPB radon exposure
Radon Assay System
Run
*
time
[hours]
location
F
exposure
[kBqhm" ]
exposure
[kBqhm" ]
N
st.dev.
[*]
1
96.8
env.chamb.
0.4
233
243
10
5
Table 1. The performance of the radon assay system
in the CEC radon intercomparisons 1984 - 1989.
env.chamb. : NRPB's 43 m -radon environmental chamber.
FRED : NRPB's 0.2 m Fast Radon Exposure Facility.
F : Equilibrium factor.
N : Number of dosimeters in the batch
-------
PRECISION
The CEC Radon Intercomparisons include 8 cases where a
batch of 10 dosimeters has received the sane radon exposure.
This allows for an experimental determination of the standard
deviation for the radon assay system at the various levels of
exposure. Final results are shown in the last column of table
An analysis of the effect of the photometer correction has
been carried out on the data from the 1987 and the 1989 CEC
Intercalibrations. All 10 dosimeters in each of the 5 groups
received the same exposure and have all been maintained under
equivalent etch conditions.
The exposures for all alpha track detectors were re-cal-
culated using the single value for the calibration factor valid
for the reference degree of etch (PMR = 64 ph). The experimen-
tal standard deviation within each group was then calculated
and compared to the standard deviation shown in table 1 where
individual, (PMR - corrected), calibration factors were used.
The results, shown in figure 6, demonstrate the effect of the
photometer correction. All the standard deviations are reduced
to sizes comparable to the standard deviations expected from
the track counting statistics alone (dashed line).
* *
• c
(-4 «
• •
U b
n
100 200
2120
NRPB •xposur# [kBq h ¦ ]
Figure 6. The effect of the photometer correction.
The relative standard deviations for groups of 10
detectors are calculated with ( • ) and without ( o )
the photometer correction.
-------
IV DISCUSSION AND CONCLUSIONS
The etch process is essential for the quality of the radon
assay system. A homogeneous etching of the alpha track detector
films is only obtained when the glass vials are kept thermally
isolated so that convection currents in the etch bath are
avoided. It is vital for the reproducibility of the etch
process that analytical grade NaOH and demineralized water is
used for the etch bath. All items in contact with the etch
liquid must be kept clean. Also the track detector films must
be kept free from fingerprints, dust or any unwanted chemicals.
With these precautions, however, the radon assay system de-
scribed in this work is simple to operate.
A standard radon measurement consists of a two months
exposure. This is sufficient to average out the effect of
meteorological changes as well as the daily variations in the
radon concentration. A reliable estimate of the annual average
of the radon concentration can then be made with a proper
correction for the seasonal variation.
The precision (one standard deviation) obtained in a two
months measurement is read from figure 6. At the level of 150
Bqm , (i.e. 4 pC/1), the exposure will be 216 kBqhm" and the
radon assay system yields a precision better,.than 15 %. Simi-
larly# a radon concentration level at 50 Bqm" can be measured
with a precision better than 20 %.
__The detection limit in a two months measurement is 10
jjcpn~ . This limit is determined as twice the standard deviation
on the background measurements which include both the natural
background of tracks on the alpha track detector and the radon
exposure received during the transit through the postal system.
We estimate the cost in US$ for the etch equipment and the
framegrabber to be $ 3500. The cost for each radon dosimeter,
including materials and labor for both fabrication and pro-
cessing, is estimated to be about $ 10.00. This is not based on
a truly "mass production" scale of dosimeter processing and the
individual cost could be reduced significantly when large
groups of detectors are processed simultaneously.
REFERENCES
1, Ulbak, K., Stenum, B., Serensen, A., Majborn, B.,
B0tter-Jensen, L. and Nielsen, S.P. Results from the Danish
indoor radiation survey. Rad. Protection Dosimetry. 24:
-------
401, 1988.
2. Ronca-Battista, M., Moon, M., Bergsten, J., White, S.B.,
Alexander, B. and Holt, N. Radon 222 concentrations in the
United States - results of sample surveys in five states.
Rad. Protection Dosimetry. 24: 307, 1988.
222
3. Fleischer, R.L. Protocol for Indoor Rn Measurements.
Health Physics. 57: 842, 1989.
4. Damkjaer, A. and Korsbech, U. A search for correlation
between local geology and indoor radon concentration. Rad.
Protection Dosimetry. 24: 51, 1988.
5. Miles, J.C.H. and Sinnaeve J. Results of the second CEC
intercomparison of active and passive detectors for the
measurement of radon and radon decay products. EUR 10403
EN, commission of the European Communities, Brussels, 1986.
59 pp.
6. Miles, J.C.H. and Sinnaeve J. Results of the third CEC
intercomparison of active and passive detectors for the
measurement of radon and radon decay products. Private
communication.
7. Damkjaer, A. The efficiency of cellulose nitrate LR 115 li
for alpha particle detection. Nuclear Tracks. 12: 295,
1986.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
-------
B-III-4
TWO-FILTER CONTINUOUS MONITOR FOR LOW LEVELS OF 220RN AND 222RN
by: D. Grumm and S. Schery
New Mexico Institute of Mining and Technology
Socorro, New Mexico
S. Whittlestone
Environmental Science Division
Australian Nuclear Science and Technology Organisation
Menai, New South Wales, Australia
ABSTRACT
A two-filter monitor system has been developed as a research instrument
for measuring low-level atmospheric 220Rn and 222Rn. There is growing
interest in thoron as a potential indoor air pollutant and measurement of
both radon and thoron at low levels. The high sensitivities (typically 33.5
(a counts/hr)/(Bq/m3) radon and 2.2 (a counts/hr)/(Bq/m3) thoron for a two hour
sample time) enable accurate time dependent measurements. We report results for
collection and counting intervals of 2 hours and a sampling rate of 65 1pm into
the 79 liter decay chamber. Continuous data taken 15 weeks at a site near
Socorro, NM show the outdoor thoron concentration at Ira varying between 2 Bq/m3
and 90 Bq/m3 with the high typically in the early morning. Preliminary data
suggest a significant correlation between outdoor thoron and radon, although
there was measurable difference in response to temperature and rainfall.
-------
INTRODUCTION
Measurement of low levels of radon and thoron is necessary for several
reasons. It has been recommended that indoor levels of radon be reduced to its
level outdoors, so concentrations of lpCi/1 (-37 Bq/m3) and lower must be
measurable. As concern over levels of indoor radon has increased, it is likely
that thoron will also become more important as an indoor air pollutant. It has
been estimated that the dose equivalent of thoron progeny is 20% that of radon
progeny, and that there are 1,000 to 5,000 lung cancer deaths per year in the
U.S. due to exposure to thoron progeny (1). Another reason is that the
fundamental mechanisms responsible for thoron concentrations and their variations
have not been investigated thoroughly. Outdoor levels are affected by
meteorological and diurnal cycles (as are radon levels), but difficulty in
measuring thoron due to its short half-life has hindered research in this area
The ability to measure thoron concentrations indoors will address the mitigation
issue of whether soil or building material is the dominant source of thoron
Although thoron levels can be inferred from thoron daughter measurements
measuring thoron directly avoids the problem of uncertainties in the extent of
equilibrium between thoron and its daughters.
The two-filter system described here is one of the few instruments in the
world that can continuously monitor, unattended, low levels of radon and thoron
simultaneously. A few other laboratories, such as the Elliot Lake Mining
Laboratories and the Australian Radiation Lab have operational systems capable
of measuring thoron gas, but designed for higher concentrations. In the research
literature can be found descriptions of a few systems, some based on the
ionization chamber principle, capable of low-level measurement but not
operational today (see, for example, refs. (2), (3)). We report here results
for collection and counting intervals of 2 hours, sampling outside air at a rate
of 65 1pm.
THE TWO-FILTER METHOD
The system operates by the two-filter principle, which has been used to
study airborne radioactivity for over twenty years. The basic theory is reviewed
by Thomas and LeClare (4), so only a brief summary will be given here. The
sampled air (which contains radon, thoron, and their progeny) enters the decay
chamber through the entrance filter which removes the progeny. As the air
travels through the chamber, the radon and thoron decay to their daughter
products. The air that leaves the decay chamber passes through the exit filter
which removes those products that have been produced by decay of radon and thoron
in the chamber. At the end of the sampling period, a detector counts the decays
of these deposited progeny. Because this activity is only due to decays within
the chamber, the radon and thoron concentrations in the entering air can be
calculated from the decay equations. In fact, for a fixed sampling and counting
protocol, the concentration of thoron and radon will be directly proportional
to the decay counts from the respective daughters requiring only an experimental
calibration.
EXPERIMENTAL DESCRIPTION
The air is supplied by a diaphragm pump (made by Gast) at a rate of about
-------
65 1pm. The entrance filter, made by Gelman, is a type A/E glass filter. Just
interior to the entrance filter are 10 closely-spaced layers of window screen,
installed to reduce channeling and to produce a uniform flow down the length of
the decay chamber. The pressure in the chamber is roughly 5 psi above
atmospheric pressure, so any inflow due to leaks is minimized. The exit filter
is a 5.7 cm wide roll of 1.2 /xm acrylic copolymer Versapor filter made by Gelman.
The chamber is approximately 1 meter long, 35 cm in diameter, and has a
volume of 79 liters. A Hewlett-Packard HP-71B microcomputer controls the entire
two-filter system, including the operation of the pump, logging the detector
output, and advancing the exit filter paper.
At the end of the sampling period, the area of the filter paper upon which
the progeny are deposited is advanced to a silicon detector which counts the
alpha decays of the progeny. An Amptek A-225 integrated circuit provides
amplification of the signal. Simultaneously the next sample to be counted is
collected. The microcomputer logs the detected a-decays in one of the following
energy ranges: 4.7 to 6.4 MeV, 6.4 to 8.1 MeV, and 8.1 to 9.7 MeV from three
electronic gates set on the output of the amplifier. The following progeny a-
decays are recorded in these windows: 218Po (6.0 MeV) and 212Bi (6.05 and 6.09
MeV), 21APo (7.7 MeV), and 212Po (8.8 MeV). A multichannel analyzer (Canberra
Series 10) was frequently used as a check on the microcomputer and to monitor
gain drifts, which were typically 0.05% per hour. An alpha spectrum collected
by the MCA for 2 hour sampling and 2 hour counting periods is shown in figure
1. From the number of events in the 214Po peak, the radon concentration in the
entering air is calculated to be 18 Bq/m3 using the experimental calibration
factor. Similarly the thoron concentration in the sampled air is 18 Bq/m3 as
calculated from the number of events in the 212Po peak.
To calibrate the system for detecting radon daughter decays, air was first
pumped through a column of uranium ore and sampled with a Lucas cell (with a
calibration checked by the intercomparison program at the U.S. Environmental
Measurement Laboratory) before entering the chamber. For a 2 hour sampling
and 2 hour counting protocol, the sensitivity is 33.5 (counts/hr)/(Bq/m3). For
experimental calibration of thoron, a 66 Bq reference source of 228Th was used
which had been verified with a manual two-filter system that had been checked
against equipment at Elliot Lake Mining Laboratories. The resulting sensitivity
(for the same time intervals) is 2.2 (counts/hr)/(Bq/m3). The background for
both thoron and radon progeny is 0.5 (counts/hr).
RESULTS
To correlate outdoor radon and thoron levels with meteorological
conditions, a microprocessor-controlled data acquisition system (Meteorite M800
by Weather Measure Corporation) continually logged barometric pressure, air
temperature, wind speed, rainfall, and relative humidity at the sampling point.
The air was sampled 1 meter above the ground and 5 meters outside the laboratory
at the New Mexico Institute of Mining and Technology in Socorro, New Mexico.
Several trends were apparent. Both the radon and thoron concentrations followed
diurnal cycles; the averages for 70 days are shown in figures 2a and 2b. The
radon levels follow an already well-established diurnal pattern (see, for
example, ref. (3)), so we will focus on our thoron results. The typical
uncertainty shown for the thoron levels are greater than that for radon due to
-------
the decreased sensitivity. Because meteorological variables typically follow
diurnal cycles, correlations with radon and thoron concentrations must be
determined from data taken during intervals in which these variables do not
follow their normal diurnal patterns. For the period 11-4-89 to 11-10-89, the
pressure did not follow its usual cycle; the thoron concentrations for this
period are shown as connected open circles in figure 3. These concentrations
do not correlate with the pressure, but anticorrelate with the temperature. This
can be explained by the same mechanism responsible for observed variations in
radon concentration, which is already established (5) (6): warming of ground
level air increases convection and mixing in the lower atmosphere, thereby
reducing the concentration at 1 meter. This relationship can be visualized by
modeling the thoron as a constant minus a term proportional to the temperature-
this fit is shown as filled circles in figure 3. Increased wind also coincided
with decreased concentrations, but in all cases the temperature was also high
For both radon and thoron, rainfall coincided with decreased concentrations
due to a decrease in flux. In figures 4a and 4b are shown the daily averages
of the concentrations, with the time of rainfall denoted by vertical lines with
lengths proportional to the amount of rainfall. On these days, the thoron level
was 37% of that on dry days, and for radon, the level dropped to 75% of its dry-
day average. This trend of a greater decrease in thoron level is expected for
two reasons. First, a larger fraction of the soil volume emanating thoron will
be blocked than the soil volume emanating radon due to thoron's shorter half,
life. Second, because the detected thoron must have been emanated very close
(meters, as opposed to kilometers for radon) to the sampling point, localized
rainfall will have a greater effect on thoron concentrations. The correlation
between thoron and radon levels can be seen from figure 5, in which all 2-hour
levels are plotted; the linear correlation coefficient is 0.40.
The average concentrations for the data in figures 2 and 3 are for radon
22.9 ± 2.2 Bq/m3, and for thoron, 28.1 ± 3.7 Bq/m3. Both values are
representative of outdoor concentrations (refs. (5) (6)) although there is very
little previous thoron data with which to compare. These data confirm an
important point not commonly recognized: in terms of activity the concentration
of thoron near ground levels outdoors is typically comparable with or higher than
radon.
COMPARISON WITH THEORY
For a simple one-dimensional model of the atmosphere (5) with decav
constant A and constant eddy diffusion coefficient K, the concentration at a
height Z above the earth can be written as:
e (1)
where J is the flux density. The steady-state flux density from homogeneous soil
can be written in terms of soil parameters:
J - 7C(-®)y«AD
(2)
-------
where C(-«>) is the concentration at depth, e is the soil porosity, D is the
diffusion coefficient in the soil, and 7 is a measure of the relative
contributions of convective flow from a pressure gradient and diffusion.
Specifically,
where V is the Darcian flow velocity (assumed constant) through the soil. For
typical flow velocities (of the order of 1CT3 cm s"1) at this site, 7 < 2 for
radon and 7 is always very close to 1 for thoron. This explains the lack of
correlations between thoron concentrations and pressure.
Considering the other terms in equation (1) , the exponential varies only
slightly with K for both radon and thoron. The eddy diffusion coefficient can
vary between approximately 10 cm2/s and 10* cm2/s (2) , depending on temperature-
induced convection and wind speed. The dominant cause of variations in the
concentration is K in the denominator. Using values of J, «, D, and C(-<*>) as
measured by Schery et al (4) , Table 1 shows the range of radon and thoron
concentrations predicted by this range of K.
V f V2
7" mW * urar+ XJ
(3)
TABLE 1: COMPARISON OF PREDICTED AND OBSERVED CONCENTRATIONS
Gas
Predicted (Bq/m3)
Observed (Bq/m3)
Radon
11 - 74
3-37
Thoron
15 - 56
2 - 91
Considering the simplicity of the model and the uncertainty in k, the
agreement is fairly good.
DISCUSSION AND CONCLUSION
1. We have successfully operated one of very few systems capable of low
level measurement of outdoor thoron.
2. Data at a Socorro, NM site show wide variation in outdoor thoron
between 2 Bq/m3 and 91 Bq/m3, with a mean of about 28 Bq/m3.
3. Thoron and radon are strongly correlated although the thoron
concentration was more affected by rain and reached a maximum sooner,
suggesting that thoron levels are more affected by local conditions.
4.
The system should be useful for a number of research projects such
as investigating the source of indoor thoron, determining typical
-------
values and variation of outdoor and indoor thoron, studying factors
affecting variation of thoron, and studying thoron in special
locations (for example, maritime and highly urbanized areas).
Acknowledgements: The assistance of Doug Gonzales and Ed Johnson in design arid
construction of the system is appreciated.
The work described in this paper was not funded by the U.S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the views
of the Agency and no official endorsement should be inferred.
REFERENCES
1. Schery, S. Radon isotopes and their progeny in the indoor environment.
In: P.N. Cheremisinoff (ed.), Encyclopedia of Environmental Control
Technology. Vol. 2, Gulf Publishing, 1989, p. 897.
2. Israel, H. , Horbert, M. , and de La Riva, C. The thoron content of the
atmosphere and its relation to the exchange conditions. Final tech. rept
European Research Office, U.S. Army contract DA-91-591-EUC-3761, Aachen
1967.
3. Crozier, W.D. Direct measurements of radon-220 (thoron) exhalation from
the ground, J. Geophys. Res. 74: 4199-4205, 1969.
4. Thomas, J.W. and LeClare, P.C. A study of the two-filter method for radon-
222. Health Phvsics 18: 113,1970.
5. Ionizing radiation: sources and biological effects. (United Nations
Scientific Committee on the Effects of Atomic Radiation) Publication E
82.IX. 8.06300P, United Nations, N.Y., 1982
6. Schery, S,, Gaeddert, D., and Wilkening, M. Factors affecting exhalation
of radon from a gravelly sandy loam. J. Geophvs. Res. 89: 7299, 1984.
7. Schery, S. Studies of thoron and thoron progeny: implications for transport
of airborne radioactivity from soil to indoor air. Tq: Indoor Radon APCA
International Specialty Conference. Philadelphia, PA, 1986, p. 25.
-------
100
MCA ALPHA SPECTRUM OF OUTDOOR AIR SAMPLE
Hi
z
z
<
z
o
Ui
209
zso
300 350
CHANNEL
Figure 1. Typical alpha spectrum collected by multichannel analyser. Count rate
In ThC' (212Po) peak corresponds to 18 Bq/m3 of thoron. Count rate In
RaC' (214Po) peak corresponds to 18 Bq/m3 of radon.
-------
HOUR
Figure 2a. Average diurnal cycle of radon concentrations based on data from 70
days, shown with a typical value of the uncertainty.
Figure 2b. Average diurnal cycle of thoron concentrations based on data from
70 days, shown with a typical value of the uncertainty.
-------
60
CO
I 40
a
cn
z
0
g 20
1
h*
102 104 106 108
DAY
Figure 3. Measured thoron (connected open circles) for 11-4-89 to 11-10-89, and
prediction (filled circles) based on phenomenological linear tempera-
ture dependence.
-------
DAY
Figure U&. Daily averages of radon concentrations 1 meter above Earth's surface
from 7-23-89 to 10-13-89. Vertical lines denote rainfall, with the
maximum equal to 0.9 cm.
Fieure 4b. Daily averages of thoron concentrations 1 meter above Earth's surface
from 7-23-89 to 10-13-89. Vertical lines denote rainfall, with the
maximum equal to 0.9 cm.
-------
RADON (Bq/m3)
Figure 5.
Plot of 2-hour thoron levels versus radon levels from 7-23-89 to
10-13-89
-------
B-III-5
ACCURACY AND PRECISION OF PASSIVE LONG-TERM RADON DETECTORS
AS A FUNCTION OF CONCENTRATION AND EXPOSURE TIME
Robert J. Lyon, Richard D. Hopper and Barry S. Parks,
U.S. Environmental Protection Agency, Office of Radiation
Programs, Las Vegas, Nevada 89119
Mark Dickson and Michael Boyd, U.S. Environmental
Protection Agency, Office of Radiation Programs,
Washington, D.C. 20460
ABSTRACT
Four types of long-term passive radon detectors were exposed
under three different exposure conditions to evaluate the effect
of concentration and exposure time on accuracy and precision.
The three sets of exposure conditions were designed to evaluate
the effect of high concentrations (200 pCi/L) for short exposure
periods (seven days) and low concentrations (0.4 to 60.0 pCi/L)
for longer exposure periods (three and six months). The detectors
were commercially available devices, with at least thirty of each
type used for each set of exposure conditions.
Radon gas monitors (RGMs) were used as the standard
measurement device for radon concentrations. At least five
devices of each type were used for background monitoring. The
detectors were sent to the manufacturers for processing in a
routine manner.
The results of these exposures are presented with an
analysis of precision and accuracy for the three sets of exposure
conditions.
This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's peer and administrative review
policies and approved for presentation and publication.
-------
INTRODUCTION
Three types of alpha track detectors (ATDs) and one type of
electret ion chamber (EIC) were exposed at selected
concentrations and time periods to obtain estimates of precision
and accuracy under different exposure conditions. Thirty to
forty devices of each type were subjected to three different
radon exposures.
Information regarding the accuracy and precision of long-
term radon measurement devices is incomplete. This study will
enable the U.S. Environmental Protection Agency (EPA) to better
understand the benefits and limitations of these devices. This
study may also indicate whether a brief, high radon concentration
will produce the same results as an extended exposure to a low
radon concentration.
MATERIALS AND METHODS
Two different types of radon detection methods were used in
this study. The more common alpha-track method was used in three
of the detector groups (designated as devices A, B, and C); while
an electret was the detector media for Device D. Device A and c
used CR-39 plastic, and device B used a cellulose nitrate film
(LR-115).
The alpha-track detector (ATD) for measuring radon is based
on alpha particles producing tracks on the detector media
(Alter81, Lovett69). Alpha particles emitted by the radon decay
products in the air impinge on the plastic or film, producing
submicroscopic damage to the detector, commonly known as tracks.
The detector media is placed in an etching bath of sodium
hydroxide to enlarge the tracks. The tracks are accentuated when
backlighted and viewed under a microscope or similar device. The
number of tracks in a counted area are proportional to the
concentration of radon to which the detectors have been exposed.
The electret ion chamber utilizes a charged detector
(electret) to collect ions produced by radon and radon decay
products (Kotrap88). The reduction in the charge of the electret
over a known period of time is proportional to the integrated
radon concentration. This voltage is measured on the electret
before and after sampling with a portable surface potential
voltmeter.
One group of detectors was exposed to 217 ± 7.2 pCi/L
simultaneously for seven days for a total exposure of 1519
pCi/L-days. The radon chamber was controlled at 21°C and 50%
humidity. Five to 10 detectors were unexposed to determine
-------
background radon concentrations. Both the exposed and unexposed
ATDs were sent to the appropriate manufacturers for normal
processing. The EICs were analyzed by EPA staff at the Office of
Radiation Programs (ORP) Las Vegas Facility using the
manufacturer's methods and calibration factors to simulate
routine analysis.
A second group of devices was exposed to naturally occurring
radon concentrations for a three month period. The radon
concentration (measured hourly by RGMs) varied from 0.4-60.0
pCi/L. The average concentration was 7.6 pCi/L for 98 days, for a
total of 745 pCi/L-days. Both the exposed and unexposed ATDs
were sent to the manufacturers as for the first group. The EIC
results were obtained as described above. After the three month
exposure, the same detectors were returned to the test facility
since electret analysis is non-destructive.
The third group of detectors was exposed to naturally
occuring radon concentrations for a 6-month period. The radon
concentration averaged 6.2 pCi/L for 182 days for a total of 1130
pCi/L-days. Both exposed and unexposed ATDs were shipped to the
manufacturers for processing. The EIC results were obtained
after a 6-month exposure.
-------
RESULTS AND DISCUSSION
Table 1 compares the known exposure levels and experimental
standard error to the average measured values of each device.
TABLE 1. Known and Measured Radon Exposure (pCi/L-days)
Exposure Known Measured Exposures ( ± la)
Period (RGM) Device A Device B Device C Device d
(days) (X ± a) (X ± a) (X ± a) (X ± a) (x ± CT)
7
1520
±
50
3040
±
2900
1244
+
369
1485
+
87
1495
±
96
98
746
±
37
975
±
347
600
+
179
625
±
36
763
±
68
180
1135
+
45
937
+
258
1232
±
244
909
+
126
1100
+
87
For convenience in determining accuracy and comparing
different manufacturers at different exposure levels, the
measured values were normalized by dividing by the known exposure
in Table 2 and 3. Table 4 lists the accuracy and precision for
devices A, B, C and D.
The results of the radon measurements are given in the
appendix in Tables 5 through 8. The known values of the
exposures were determined by calibrated radon gas monitors
(RGMs).
-------
TABLE 2. Normalized Radon Exposures of Devices A and B
Device A
Device B
7 days
98 days
180 days
7 days
98 days
180 days
1.21
1.41
0.66
1.41
0.75
1.10
0.93
0.72
1.09
1.27
0.89
1.12
0.96
0.49
0.94
0.75
0.58
1.40
0.42
1.75
0.72
1.02
0.83
1.18
1.11
0.39
1.00
1.09
0.85
0.88
1.18
0.50
0.99
1.20
1.07
1.46
7.09
1.60
0.90
1.20
1.06
0.76
1.08
1.42
1.00
0.87
0.94
1.20
6.17
1.77
1.03
0.82
1.04
0.81
4.78
1.89
0.68
0.90
0.77
1.48
1.16
0.55
0.08
0.81
1.04
0.61
5.84
1.52
0.90
1.04
0.63
1.31
1.07
1.66
0.73
0.85
0.69
0.75
1.13
1.86
0.84
0.95
1.12
1.10
1.38
0.89
0.20
0.72
1.12
0.84
1.06
1.80
0.89
0.72
1.12
1.35
1.16
1.47
1.06
0.42
0.74
1.08
1.26
1.89
1.05
0.81
1.31
0.89
5.29
1.67
1.10
0.51
0.84
1.04
1.00
0.94
0.79
0.39
0.81
1.27
1. 16
1.21
0.98
0.40
0.43
1.32
1.11
1.09
0.74
0.70
1.12
1.21
1.17
1.82
0.73
0.72
0.52
0.87
1.21
1.33
1.01
0.62
0.80
0.96
1.13
1.63
0.80
0.86
0.30
1.19
0.98
0.97
0.78
0.69
0.38
1.02
1.44
0.82
0.71
0.71
1.05
0.91
0.88
0.72
0.92
0.81
1.42
0.72
0.74
0.95
0.92
0.64
0.51
0.30
0.95
0.93
0.96
1.11
0.59
0.61
0.95
0.44
0.56
1.14
1.05
0.87
1.33
0.93
0.54
0.98
1.16
0.87
1.17
0.77
1.03
1.40
0.92
0.73
0.86
0.73
0.61
1.36
0.78
1.16
-------
TABLE
3. Normalized Radon
Exposures of
Devices C
and D
Device C
Device D
7 days
98 days
180 days
7 days
98 days
180 days
1. 00
0.83
0.78
0.95
1.04
0.95
0.96
0.80
0.88
1.08
1.03
1.00
0.98
0.89
0.90
0.98
1.01
0.96
1.00
0.83
0.78
0.95
1.04
0.95
0.96
0.80
0.88
1.08
1.03
1.00
0.98
0.89
0.90
0.98
1.01
0.96
0.96
0.89
0.79
0.99
1.04
1.00
0.97
0.88
0.72
0.95
1.02
0.96
1. 03
0.83
0.67
0.97
1.02
0.98
1.01
0.80
0.83
0.99
1.01
0.98
1.05
0.81
0.89
0.94
1.01
0.96
1.03
0.94
0.84
0.99
0.93
0.87
1.00
0.83
0.81
1.00
0.98
0.94
1.05
0.80
0.61
0.95
1.00
0.95
0.97
0.82
0.60
0.93
1.04
0.99
0.88
0.92
0.89
0.95
0.95
0.90
0.96
0.83
0.88
0.97
1.02
0.97
1.01
0.84
0.61
0.94
1.09
1.03
1.03
0.86
0.80
0.96
1.11
1.01
0.94
0.77
0.89
1.03
0.98
0.92
0.98
0.84
0.79
1.00
1.00
0.96
1.00
0.90
0.87
0.95
0.94
0.90
1.01
0.88
0.85
0.99
0.99
0.97
0.93
0.87
0.86
0.98
0.97
0.93
1.05
0.89
0.93
0.94
1.01
0.97
1.04
0.82
0.89
1.15
1.46
1.33
0.89
0.85
0.55
1.22
0.93
0.87
0.97
0.90
0.66
0.94
0.97
0.92
0.91
0.84
0.84
0.94
1.02
0.97
0.89
0.87
0.89
0.94
1.04
0.95
0.98
0.89
0.84
0.94
1.05
1.01
0.88
0.90
0.66
0.98
1.03
0.98
1.03
0.78
0.85
1.00
1.02
0.98
0.85
0.88
0.86
0.96
0.81
0.91
0.87
0.85
0.74
0.91
0.80
0.68
1.01
0.77
0.80
1.00
0.77
0.84
1.09
0.81
0.89
1.03
0.79
0.83
1.02
0.75
0.86
0.95
0.74
0.74
-------
TABLE 4. Accuracy and Precision for Devices A, B, C and D
Device/
Exposure
Period
(davs) Min.
Device
A
7
.42
98
.39
180
.08
Device
B
7
.39
98
.30
180
.61
Device
C
7
.85
98
.74
180
.48
Device
D
7
.93
98
.93
180
.87
% of Values
Different
Std. from True
Dev. % Value by
MeXt Mean (?) 25% or more
7.09
2.00
1.91
96
35
1.89
1.31
.44
33
79
1.10
.83
.23
27
33
1.41
.82
.24
30
65
1.31
.80
.24
30
46
1.48
1.09
.21
20
36
1.09
.98
.06
6
0
.94
.84
.05
6
7
.93
.79
.11
14
30
1.22
.98
.08
8
0
1.46
1.02
.09
9
3
1.33
.97
.08
8
3
-------
CONCLUSIONS
Device A shows an improvement in precision as the exposure
time increases from the 7-day to the 180-day exposure. The
measured values of the 3-month exposure cannot be distinguished
from the measured values of the 180-day exposure. Device A is
biased high at low exposure times and biased low for longer
exposures.
Device B shows no statistically significant change in
precision as exposure time increases from 7 days to 3 months, but
there is an improvement between 3 months and 6 months. The
measured values of the 7-day exposure cannot be distinguished
from the measured values of the 180-day exposure. Device B is
biased low for short exposures and biased high for longer
exposures.
Device C shows better precision than either A or B, but the
precision decreases and the bias increases with increasing
exposure time. Thirty percent of the measured values are more
than 25% below the RGM values for 6-month exposures.
Device D, the EIC, showed superior precision and accuracy
and retained the precision and accuracy as the exposure duration
increased. In this study, device D was the most precise and
accurate radon measuring device for long-term exposure
conditions.
Research is continuing to determine the reason why, in these
experiments, the ATDs manufactured with CR-39 appear to have an
increasingly negative bias as the period of deployment increases.
If there is some phenomenon that causes tracks to be lost over
time, then it will be important to characterize this process and
correct for it. The next phase of this project will cover a 12
month exposure period.
-------
Appendix
TABLE 5.
Measured Values of
Radon by Device
A
Radon fDCi/L d)
Backaround (DCi/L
7 days
98 days
180 days 7
days 98 days
180 days
1838
1051
755
97 47
164
1410
540
1233
97 87
52
1456
363
1066
47 38
46
639
1305
815
161 25
39
1683
291
1140
77 30
15
1791
372
1123
89
10
10771
1192
1016
30
1637
1062
1136
36
9385
1321
1167
46
7255
1413
776
106
1761
408
90
8875
1138
1026
1622
1235
828
1714
1386
956
2100
666
230
1606
1345
1009
1761
1094
1201
1915
1413
1188
8042
1246
1251
1525
624
900
1761
900
1114
1683
816
845
1776
1359
833
1838
990
1151
1714
1213
909
1490
726
885
1073
931
678
1000
1062
813
724
-------
TABLE 6. Measured Values of Radon by Device B
Radon
(pCi/L d)
Background (pCi/L d)
days 98
days
180 days
7 days 98 days 180 days
2146
557
1254
246
1931
662
1277
157
1140
429
1590
0
1543
616
1344
0.0 ± 65 0.0 ± 65 638
1663
637
997
0
1819
795
1658
29
1823
790
862
15
1322
700
1366
252
1247
778
918
378
1361
571
1680
216
1234
774
694
1579
470
1490
1285
515
851
1445
835
1243
1095
835
956
1100
839
1529
634
550
1220
1225
980
1014
770
627
1176
591
604
1441
613
319
1499
1057
835
1382
1092
385
985
941
599
1088
1312
223
1352
1053
280
1161
1090
531
1191
1098
685
918
1120
706
1044
773
224
1080
1419
720
1260
899
454
1080
664
420
1296
1590
647
1512
1413
403
1116
1770
650
1332
1165
770
1584
1399
547
972
1104
454
1548
582
1314
-------
TABLE 7. Measured Values of
Radon (pCi/L d)
days
98 days
180 days
1521
622
884
1452
595
1004
1485
663
1018
1458
666
896
1479
658
816
1560
616
759
1533
597
941
1603
607
1011
1564
704
956
1521
617
919
1600
595
690
1479
614
682
1334
683
1006
1458
619
1002
1539
624
687
1560
638
907
1428
575
1014
1497
624
891
1524
668
991
1539
656
966
1419
648
979
1594
665
1051
1574
614
1009
1360
631
622
1468
670
749
1381
629
954
1357
651
1009
1490
661
959
1336
671
745
1574
582
964
1290
660
973
1456
602
1029
1321
631
838
1387
599
770
1532
572
941
1517
576
542
1659
606
1006
1568
587
940
1550
559
976
1444
554
844
Radon by Device C
Background (pCi/L d) 7
7 days 98 days 180 days
18
18
44
18
12
30
18
15
33
0
9
12
15
6
38
21
3
47
18
6
27
9
9
53
18
12
35
27
6
53
-------
TABLE 8.
Measured
Radon (pCi/L d)
days
98 days
180 days
1448
774
1077
1642
768
1136
1485
751
1088
1504
779
1134
1440
760
1092
1476
759
1107
1506
750
1116
1429
754
1090
1503
692
986
1520
734
1065
1439
743
1074
1415
775
1121
1441
710
1021
1469
761
1097
1423
811
1165
1453
825
1152
1565
733
1039
1516
743
1085
1441
702
1023
1504
740
1097
1486
725
1057
1434
754
1096
1744
1092
1509
1858
691
990
1430
723
1048
1427
761
1103
1424
779
1083
1427
782
1141
1483
767
1108
1515
760
1117
Background (pCi/L d)
7 days 98 days 180 days
0 ± 6
0 ± 6
0 ± 6
-------
Alter81
Lovett69
Kotrap88
REFERENCES
Alter, H.W. and R.L. Fleischer, "Passive Integrating
Radon Monitor for Environmental Monitoring", Health
Physics 40, 693 (1981)
Lovett, D.B., "Track Etch Detectors for Alpha Exposure
Estimates", Health Physics 16, 623 (1969)
Kotrappa, P., J.C. Dempsey, J.R. Hickey, and L.R.
Stieff, "An Electret Passive Environmental Rn-222
Monitor Based on Ionization Measurement", Health
Physics 54, 47-56 (1988)
-------
B-III-6
The EPA Diffusion Barrier Charcoal
Adsorber for Radon Measurements in
Indoor Air
David J. Gray and Sam T. Windham
Eastern Environmental Radiation Facility
U.S. Environmental Protection Agency
Montgomery, AL. 36109
ABSTRACT
An improved charcoal adsorber radon monitor capable of integrating
radon concentrations from two through seven days has been developed
by EPA. The EPA open-faced charcoal canister was modified by
inserting a diffusion membrane on top of the activated carbon bed.
This resulted in a decrease in the adsorption and desorption rates
of the canister, as well as a decrease in moisture gain of the
carbon. Excellent results have been obtained with this canister
in environments where the radon concentration has varied by more
than 10 to 1 and at humidity ranges from 20 to 85 percent. Radon
measurements made with diffusion barrier canisters are less
affected by temperature extremes and air movement than open-faced
canisters.
This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's peer and administrative review
policies and approved for presentation and publication.
-------
INTRODUCTION
The open-faced passive charcoal adsorber used by the
Environmental Protection Agency (EPA) and others in the radon
industry responds quite well for a 2-day exposure period in most
indoor situations, however, researchers have documented limitations
of these devices under extreme conditions. Examples of these
conditions include exposure in areas with relative humidity >60%
(1), when radon concentrations vary by more than a factor of io
(2), and measurements made with air movement in the vicinity of the
adsorber (3) (4) .
At the EPA's Eastern Environmental Radiation Facility (EERP)
we have designed a diffusion barrier charcoal adsorber which
responds more uniformly in extreme conditions and extends the
integration measurement period. The EPA open-faced canister was
modified by inserting between the retaining screen and the carbon
bed a 1.25 mil polyethylene membrane containing twenty 0.092.1t
diameter holes. This results in a decrease in the amount of water
vapor and radon adsorbed by the carbon, a reduction in the rate of
adsorption/desorption between the carbon and the environment, and
improved integration capability. In this paper we will describe
these improvements in radon measurement by the diffusion barrier
adsorber.
RESULTS AND DISCUSSION
Radon Adsorption
Sets of 5 diffusion barrier canisters were exposed to the sane
radon concentration for periods of l to 10 days at low, medium, and
high relative humidities in the radon chamber at the EERF. The
chamber was operated in an "active" configuration thus presenting
air flows of 10-30 LFM, not unlike conditions found in homes we
-------
have measured. Three hours post exposure, the canisters were
counted for 10 minutes using EERF SOP's (5). The collection
efficiency of each canister, in terms of adsorbed radon on the
carbon, was calculated from the following equation.
Adsorbed radon (pCi/g per pCi/L) = N
E X 70 X Rn X Df
where:
N is the net count rate, (CPM),
E is the counting efficiency of the detector, (CPM pCi"1),
Rn is the radon concentration during the exposure, (pCiL"1) ,
Df is the decay factor for radon from the midpoint of exposure
until the time of counting,
and
70 is the weight of carbon in the canister, (g).
The results for the diffusion barrier canister are shown in
Figure l. The collection efficiency has a positive slope for the
first five days over the range of relative humidities and then
jaegins to plateau, especially at high humidity for the remaining
exposure period. This is in contrast to the collection efficiency
of the open-faced canister at high relative humidities (>70%) (6)
where the collection efficiency begins to decrease after an
exposure period of 2 days. This difference is attributed to the
characteristic of the diffusion barrier to reduce the adsorption
of water vapor by the carbon. The break point of the carbon is the
point where the amount of adsorbed water vapor is sufficient enough
to prevent further adsorption and desorption of radon in the carbon
jyed (7). The diffusion barrier canister has not reached the break
point of the carbon for up to 7 days at high relative humidities.
-------
Adsorbed Radon on Diffusion Barrier Canister
as a Function of Humidity and Exposure Duration
Exposure Time (Days)
Figure 1. Diffusion Barrier Canister Radon Adsorption Curves
-------
Radon Desorption
The effective half-life of radon in the charcoal canister is
the combination of radioactive decay of the radon activity and
desorption of radon from the carbon bed. To determine the
effective half-life of radon in the diffusion barrier canister, a
set of 5 canisters was exposed in the EERF chamber to 25 pci/L for
7 days at a relative humidity of 50%. The canisters were sealed
and counted to determine the radon activity (Q0) . They were then
reopened and left to desorb in a low radon environment (
-------
1.2
Desorption of Radon from Diffusion Barrier Canisters
as a Function of Humidity and Time
1.1 -
o H—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—
0 20 40 60 80 100 120 140 160 180 200
Time (hours)
Figure 2. Diffusion Barrier Canister Desorption Curves
-------
Response Of The Diffusion Barrier Canister During Variable Radon
Concentration
The ability of diffusion barrier canisters to integrate time-
weighted average radon concentrations was evaluated through a
series of exposures. Sets of 5 canisters were exposed for periods
of 2 and 7 days during which radon concentration varied from 6 to
1 to an extreme case of 100 to 1. The results of these exposures
are shown in Figures 3 through 5.
In Figure 3, diffusion barrier and open-faced canisters were
exposed at 70'F and 85% relative humidity for a period of 2 days.
During the exposure period radon concentration varied from a low
of 6 pCi/L to a high of 43 pCi/L. At the end of exposure the
canisters were analyzed and the mean performance ratio (the mean
of the reported radon concentration of the set of 5 canisters
divided by the true average radon concentration of the continuous
monitors) was calculated. The mean performance ratio for the open-
faced and diffusion barrier canisters was 0.99 and 1.0
respectively, for the 2 day exposure period.
The result of diffusion barrier canisters exposed for 7 days
at 70#F and 85% relative humidity under extreme radon variability
is shown in Figure 4. The radon concentration varied from a low
of 0.5 pCi/L to a high of 53 pCi/1 during the period. The 5
canister mean performance ratio of 0.85 reflects the ability of the
diffusion barrier canister to provide acceptable results under
conditions of extreme radon variability.
The result of diffusion barrier canisters exposed for 7 days
at 70#F and 50% relative humidity is shown in Figure 5. The
variation in radon concentration was less extreme, varying from a
low of 5 pCi/L to a high of 50 pCi/L. The mean performance ratio
was 1.06 for the exposure period.
-------
Diffusion Barrier Integration Test - High Humidity - 2 Days
Canisters vs. Continuous Radon Monitors
0 20 40
Time (Hours)
CRM A CRM B Average a Scintillation Cells
Figure 3. Diffusion Barrier and Open—Faced Canister Passive Chamber Test
-------
60
50 -
40 -
30 -
20
10 -
Diffusion Barrier Integration Test - High Humidity - 7 Days
Canisters vs. Continuous Radon Monitors
CRM A
Time (Hours)
CRM B Average
180
A Scintillation Cells
Figure 4. Diffusion Barrier Canister Passive Chamber Test
-------
Diffusion Barrier Integration Test — Medium Humidity - 7 Days
Canisters vs. Continuous Radon Monitors
60
50 -
40 -
30 -
Canister Mec
(27.9 pCi/lit
I"")
20 -
10 -
ean
CRM M
(26.3 pCi/lit(
r)
—T~
20
~r~
40
-1 1 1 1 1 r~
60 80 100
Time (Hours)
CRM B Average
0
CRM A
120
140
160
a Scintillation Cells
Figure 5. Diffusion Barrier Canister Passive Chamber Test
-------
Field tests were also conducted in several residential
structures throughout the country. Figure 6 shows the result of
5 diffusion barrier canisters exposed for 160 hours in the basement
of a house in Michigan. The mean concentration of 7.3 pCi/L for
the canisters compares quite well to the 7.2 pCi/L average measured
by the continuous monitors.
CONCLUSIONS
Testing of the diffusion barrier canister at the EERF and in
the field has shown that it can measure average radon concentration
for an extended exposure period compared to the open-faced
canister. The diffusion barrier decreases the amount of water
vapor adsorbed by the carbon and reduces the adsorption/desorption
rate of the carbon creating a more uniform sampling rate of the
canister. This improves the retention and integration ability of
the canister under extreme environmental conditions.
-------
Field Test of Diffusion Barrier Canisters vs. Continuous Radon Monitors
Time (Hours)
Figure 6. Diffusion Barrier Canister Field Test
-------
REFERENCES
1. George, A.C. Passive Integrated Measurement Of Indoor Radon
Using Activated Carbon. Health Physics. 46: 867; 1984.
2. Cohen, B.L. Nason, R. A Diffusion Barrier Charcoal
Adsorption Collector for Measuring RN Concentrations In
Indoor Air. Health Physics. 50: 457; 1986.
3. Gray, D.J.; Windham, S.T. The Overresponse of Open-Faced
Charcoal Adsorbers Used For Measurements Of Indoor Radon
Concentrations. Paper Presented at 1988 EPA Symposium On
Radon and Radon Reduction Technology, Denver, CO.
October 11-15, 1988.
4. Pearson, M.D.; Spangler, R.R. The Effect Of Moving Air On
The Accuracy Of Radon Measurements Made With Charcoal
Canisters. Paper Presented at 1988 EPA Symposium On Radon
And Radon Reduction Technology. Denver, CO. October 11-15,
1988.
5. Gray, D.J.; Windham, S. T. EERF Standard Operating
Procedures For Radon-222 Measurement Using Charcoal
Canisters. EPA 520/5-87-005; 1987.
6. Ronca-Battista, M.; Gray, D. The Influence Of Changing
Exposure Conditions On Measurements Of Radon Concentrations
With The Charcoal Adsorption Technique. Radiation
Protection Dosimetry. 24: 1/4; 1988.
7. Scarpitta, S. Factors Affecting The Rapid Estimation Of
Indoor Radon Using Passive Activated Charcoal Canisters.
Doctoral Dissertation, Environmental Health Sciences, New
York University; 1988.
-------
Session B-IV:
Short-/Long-term Radon Measurements—POSTERS
-------
B-IV-1
SURVEY OF RADON 222 IN MONROE COUNTY, PENNSYLVANIA
P. N. HOULE AND DOUGLAS N. BLICK
DEPARTMENT OF PHYSICS
EAST STROUDSBURG UNIVERSITY
EAST STROUDSBURG. PA
ABSTRACT
A high density of home radon concentration measurements over
one Pennsylvania County was made. Agreement between these results
and larger area but lower density measurements is found to be good.
It is concluded that a measurement density of approximately .15
measurements per square mile will yield results in good agreement
with a density of 1.39 measurements per square mile. It is also
found that Pennsylvania, in part or taken as a whole, has three
times as high a percentage of homes with radon concentrations
greater than 4.0 Pci/L than present E. P. A. estimates indicate.
-------
Survey of Radon 222 in Monroe County, Pennsylvania
P.N. Houle and Douglas N. Blick
Department of Physics, East Stroudsburg University, East
Stroudsburg, Pennsylvania
Introduction
Numerous surveys of radon concentrations over
1,2,4
relatively large geographic areas have been reported. This
report is of radon concentrations over a relatively small
area. (Monroe County covers an area of approximately 601.4
square miles, and is in the northeast corner of
Pennsylvania.) Eight hundred thirty-three (833) measurements
were made yielding a measurement density of 1.39
measurements per square mile. The data reported herein
covers all 23 Monroe County zip codes.
East Stroudsburg University has set up a radon testing
program primarily for homes within its county—Monroe
County—with the purpose of determining average radon
concentrations. All persons interested in the program could
write to the university to request test kits. These kits
consisted of the now traditional charcoal canister with
diffusion barrier, and were distributed for a price, at
first, of $10.00, then one year later, $12.00.
The test results are represented by zip code area, as
shown in Figures #1 and #2.
1
-------
2
Materials and Methods
The E.S.U. Radon test kit consists of two metal
canisters, each, 2 inches in diameter and 1.5 inches tall.
Inside each canister are 15 grams of activated carbon-type
PCB 12x30 made from coconut charcoal. The carbon is held in
the canister by a metal screen, with mesh openings of
0.5mm. This metal screen, which is cut to fit inside the
canister, is held securely against the charcoal by a bead of
silicon caulk.
The lid of this canister has a 1/2-inch diameter hole
through its center. A 1-inch square nylon mesh with openings
of 30 microns and a thickness of 70 microns is glued to the
bottom of the lid covering the hole. A 3-gram, drop-in
style* sorb-it bag is taped to the bottom of the lid
covering the nylon mesh and hole.
To make the canister lid secure and air tight, black
electrical tape is wrapped around the perimeter of the
canister cover along its seam. Duct tape is then placed over
the hole on the lid, thus sealing the canister.
Both containers are weighed together to the nearest
tenth of a gram before and after exposure to determine the
moisture captured within the container. Instructions sent
with the kits indicate the kits should be placed in the home
-------
3
side by side. After exposure for seven days, the detectors
are analyzed with a two-inch diameter sodium iodide crystal
and associated electronics, including a multi-channel
analyzer which was used to count Gamma rays in the energy
range of 220 to 384 kev. The canisters are counted for
thirty minutes in a low background environment of
approximately 600 counts. The conversion from these net
counts to radon concentrations is determined via calibration
factors determined by exposing canisters to known radon 222
concentrations in a radon chamber at East Stroudsburg
University. Calibration factors were calculated for two
different relative humidities of 10% and 80% and numerous
radon 222 concentrations. Linear interpolation between these
values is used to determine radon concentrations for
intermediate ranges.
A daily quality assurance check on the counting system
was performed with canisters which had been exposed at known
concentrations. Further calibration results were made by an
external lab.
-------
4
Distribution of Radon Test Kits
Results reported here have been based on the kits which
were returned to the University from January, 1986 to April,
1989. This is an actual case study, not a random sample
study.
Included in the kits is a one-page list of instructions
for the proper use of the canisters. These instructions are
in accordance with EPA protocols.
A brief questionnaire to collect data on home
characteristics is also included with the radon test
kit. The earlier ones that were handed out from January,
1986 to January, 1987 consisted of a questionnaire that
asked for the zip code, county, the location of the
detectors within the house, the number of stories, the
relative humidity, and the type of heating system used in
the home. Questions were asked, such as: Does the house have
a crawl space, concrete slab, or basement? (This is unlike
the new questionnaire presently in use, which is much more
elaborate, having twenty-nine questions.)
-------
5
Discussion
Figure #1/ shows a map of radon concentrations with
shading from the lightest to darkest, indicating radon
concentrations from low to high concentrations
respectively. This figure also shows that there are six zip
codes in Monroe County whose mean radon concentrations are
lower than 4 pCi/L, as shown in Table 1. The highest mean
concentration is in zip code 18341 with a mean concentration
of 49.2 pCi/L.
Figure #2 shows Monroe County again by zip codes, the
white areas representing those regions where no measurements
exceed 10 pCi/L. The area represented by the hash marks
shows regions where no measurements exceed 41 pCi/L, and the
black represents areas where no measurements exceed 420
pCi/L.
-------
6
Table 1 depicts the percentage of homes whose radon
concentrations are below the value indicated.
Figure 3 compares radon concentrations in homes with
well water to radon concentrations in homes with municipal
water supplies.
In Monroe County there is approximately an equal number
of homes utilizing municipal water supplies as there are
using well water. This data is heavily weighed to those
homes having municipal water supplies. A further
categorization of the data is by floor level —basement#
first floor, and second floor—also shown in Figure 3.
On all three floors (basement, first, and second),
there was no significant difference between radon
concentrations in homes which use well water and those which
use municipal water. In fact, homes which used municipal
water had a higher mean radon concentration on all floors
compared to the radon concentrations found in homes which
used well water.
-------
7
Conclusions
It should be noted in Figure #1 that the darker colored
lines are in the southern part of Monroe County, which is
close to the Reading Prong, yielding one possible reason for
the southern end of Monroe County to have such a high radon
concentration. It could also be speculated that when the
glaciers moved southward towards this area during the last
ice age, they advanced not too much farther than the
southern end of Monroe County. When these glaciers receded
Zy uncovered the top layer of soil leaving uranium
deposits close to the surface.
It is of some interest to note that 42.4% of the homes
tested in Monroe County have radon concentrations greater
than 4.0 pCi/L. This is in modest agreement with the most
recent Pennsylvania D.B.R. results, whereas early radon
concentration estimates by EPA indicated 15% of all U.S
homes would have radon concentrations greater than 4.0
pCi/L.
In his book, Radon: A Homeowner's Guide to Detection
and control, Cohen indicates that of 6606 homes he tested in
Pennsylvania, 3055 homes (46%) were above 4 pCi/L. Michael
Lafavore points out in his boo*, Radon:. The Invisible
Threat, that of 20,527 homes tested in Pennsylvania, 11,770
were above 4 pCi/L or 57.3% of homes tested.
-------
8
If half of all homes in Monroe County were to be
mitigated at an average cost of $1,500.00/home (this
estimate allows for sub-slab ventilation), the total cost
would become 67.5 million dollars. Too, one might note that
original EPA estimates indicated only 15% of all homes in
the U.S. might need to be mitigated, to be lowered to below
4.0 pCi/L.
Another parameter of interest is the number of
measurements per square mile. Pennsylvania covers an area
of approximately 45,333 square miles. The 6606 measurements
obtained by Cohen, yield a density of .15 measurements per
square mile. Lafavore's measurement's yield a density of .45
measurements per square mile. This indicates that even
though these measurement densities are less than that of
this report the results are all in very close agreement.
The results in Figure 3 are somewhat surprising in that
one would expect that homes with well water would have
higher radon concentration than those with municipal water
supplies (even though the data in not significantly
different). Two possible explanations are proposed:
1. There are many small municipalities with their
own water supply coming from small reservoirs that are
replenished by a stream.
-------
9
2. A number of municipalities obtain their water
from deep wells.
There are 90,000 full-time homes in Monroe County, and
at the present time E.S.U. has tested almost 1% of these
homes. The data depicted in Figures 1, 2, and 3, of course,
report on all the data. However, if one chooses at random
10% of this data to analyze, then 20%, etc., one finds that
results identical to within 20% of those reported herein are
produced when only 20% of the data are analyzed. This
suggests that one needs only .2% of all homes to be measured
to determine average radon concentration values to within
20% of a study analyzing five times more data. A further
complication is indicated in these results by the
possibility that a given homeowner may have evaluated the
radon concentration in a given home more than once (perhaps
before and after mitigation). Because of the confidentiality
of the homeowner's name and address, and because many of the
addresses are rural routes, we can only estimate this
effect. We estimate less than 1% of the data is so
affected.
One last important note. Figures 1 and 2 are somewhat
misleading in the depicting actual boundaries for given
average radon concentrations. The boundaries are by zip code
and the average radon concentrations are also reported by
-------
10
zip code. Average radon concentrations taken over the same
county but divided geographically in other ways could yield
results different in some respects from those reported
here.
The work described in this paper was not funded by the U.S. Environmental
protection Agency and therefore the contents do not necessarily reflect
the views of the Agency and no official endorsement should be inferred.
-------
11
References
1 Cohen/ Bernard. Radon: A Homeowner's Guide to Detection and
Control. Mount Vernon, New York: Consumers' Union, 1987
2
Lafavore, Michael. Radon: The Invisible Threat. Emmaus,
Pennsylvania: Rodale Press, 1987.
3
Swedjemark, Gun Astri; Hakan Wahren; Makitalo, Astrid; Tell,
William. Experience From Indoor Radon-Daughter
Limitation Schemes in Sweden. June, 1987.
4
Watson, J. E., Jr. ; Adams, W.C.; Xie, Y. Survey Of Rn 222
In North Carolina Homes. Health Physics. 55:71-75;
1988.
-------
12
Table 1. Radon Levels in Monroe County by Percentage
Radon Level (L) % of Homes with Radon
(pCi/L) Less than (L)
4 57.6%
10 79.9%
15 86 1%
20 88.9%
25 91 5%
50 96.2%
100 98 8%
420 100.0%
-------
1 7
1 7
2.5
29
30
34
4 1
4 4
45
5.0
5.1
5.1
53
56
63
67
72
75
8 1
85
87
105
108
137
15 1
173
174
21 7
24 0
263
339
49 2
>
3.
ZIP
18325
18445
18610
18335
18347 (
18334 I
18352
18466
18372
18326
18320
18333
18342
18350
18323
18322
18301
18356
18370
18355
18354
18360
18344
18332
18327
18331
18349
18058
18353
18321
18330
18341
MAP OF RADON CONCENTRATIONS IN MONROE COUNTY
LIGHTER SHADED AREAS INDICATE LOW CONCENTRATIONS
GEOGRAPHIC DIVISIONS ARE BY ZIP CODE.
P.N. Houle & D.N. Blick
-------
~ NO MEASUREMENT > 10pCi/L
53 NO MEASUREMENT) 41pCi/L
¦ NO MEASUREMENT >420pCi/L
FIG. 2 MAXIMUM Rn222 MEASUREMENT IN MONROE CO.
P N. Houle & D.N. Blick
-------
ZQ<
¦ MUNICIPAL WATER
~ WELL WATER
-i 15
o
o
z:
o
< 10
Or
h-
Z
LlJ
O
O
o
<
UJ
5f
NUMBER OF
HOUSES
409 !00
BASEMENT
2^0 60
FI RST
17 7
SECOND
FIG. 3 MEAN Rn222 CONCENTRATIONS FOR HOUSES ON
MUNICIPAL WATER AND WELL WATER
P.N. Houle & O.N. Blick
-------
B-IV-2
PRIVATE SECTOR SURVEY
by: John Hoornbeek
Office of Radiation Program
U. S. Environmental Protection Agency
Washington, D. C.
Josefina Lago
The Washington Consulting Group
Washington, D. C.
ABSTRACT
The Environmental Protection Agency (EPA) conducted a nationwide survey
of private sector radon mitigators during the summer of 1989. The survey
characterizes the radon mitigation industry, its accomplishments, and its
needs. About 1700 mitigators from all 50 states were questioned on their
radon mitigation activities. Information gained through the effort includes
the geographic distribution of the industry, the number of buildings
mitigated, industry growth rates, and training needs. This paper reviews the
findings of the survey in detail. It also provides suggestions on how
governmental policies and practices can be changed to improve the industry's
ability to reduce the health risks of radon.
-------
B-IV-3
SAMPLING STRATEGIES OF RADON SURVEYS: THE ITALIAN EXPERIENCE
by: G.Campos-Venuti and G.Farchi
Istituto Superiore di Sanita, V.le R.Elena 299
00161 Rome, Italy
S.Piermattei
ENEA/DISP, Via V.Brancatl 48, 00144 Rome, Italy
ABSTRACT
The criteria according to which these surveys are carried out are of
particular importance in obtaining reliable results. The present paper
is concerned with the methodologies employed in national surveys of the
second generation to evaluate the mean exposure of the population to
natural radiation indoors.
The paper describes the probability sampling procedure (a two-stage
stratified technique) adopted in Italy to carry out the national survey.
This type of sampling (with 5000 dwellings) will allow to obtain re-
presentative results both at Regional level and National level. This is
particularly relevant since the survey represents a cooperative effort
between national and local (administrative and scientific) organizations.
INTRODUCTION
An extensive research work (1, 2, 3) has been conducted during the
last twenty years by a great number of laboratories. The aim of this work
consisted in the evaluation of the amount of the exposure to natural
radiation in the domestic environment mainly with regard to radon and its
decay products and in the investigation of the various parameters affecting
their presence indoors (i.e., geology of soil, microclimate, ventilation,
building materials, water supply, heating systems, etc).
Several countries undertook surveys In dwellings (see, for instance 4
and 5), called first generation surveys, to determine the ranges of indoor
radon concentration and to Identify high risk areas. Theoretical and
experimental studies were also carried out to improve the knowledge of the
physical behaviour of radon decay products (aerosol size distribution,
-------
diffusivity, attachment rate, etc.) in the room air, to measure and model
aerosol deposition to better understand the role of the unattached fraction
of aerosols in dose calculation (6,7).
Towards the middle of the eighties the knowledge of radon entry routes
was improved and was evidentiated the major role played by the soil as
source of indoor radon compared to building materials, not only with
regards to the Ra-226 content but also to other important soil parameters
such as grain size, permeability, porosity, moisture content, etc. (8). At
this time, studies were also carried out which demonstrated the existence
of an inverse relationship between the attached fraction of radon daughters
existing in the domestic environment and the degree of equilibrium of radon
gas and its progeny with respect to the aerosol concentration. This sug-
gested a direct relationship between the dose and the concentration of
radon gas, which can be more easily measured by integrating passive detect-
ors (9). During these years reviews of epidemiological studies on under-
ground miners, together with the improvement of the existing dosimetric
models, allowed the validation and refining of the risk estimates as-
sociated to exposure to radon indoors (10,11).
ICRP in its Publication 39 (12) suggested intervention levels at which
remedial actions might be undertaken in buildings where radon concentration
exceeds such values and, at the same time, a distinction was made between
old and new dwellings. Some countries (USA, U.K., Nordic countries) have
adopted recommendations to limit the exposure to radon daughters indoors by
choosing Action Levels for existing and upper bounds for future houses. For
these reasons new series of surveys, called second generation surveys, were
undertaken with two main objectives:
1) the evaluation of the average health risk to the population
deriving from exposure to indoor radon;
2) the knowledge of the number of dwellings where proposed levels are
being exceeded and the assessment of economical and social costs of re-
medial action nationwide.
These objectives can be attained only provided that the measurements
are carried out by selecting a sample of dwellings statistically repre-
sentative of the national situation.
This paper describes the various sampling strategies employed to
choose a representative sample of dwellings and illustrate the one adopted
in Italy for the National survey to evaluate the exposure of Italian
population to natural radiation indoors.
SAMPLING STRATEGIES
The basic concept to be kept In mind when a sample survey Is under-
taken is that reliable information will be obtained provided that: i) the
sample is representative, i.e., reflects correctly the units under study
and it is possible to draw an inference from the sample to the population;
ii) the sample is large enough to yield significant results.
-------
Different methods are available to select a representative sample
(13). The most straightforward one is called List sampling, or Random List
sampling. Every sample has the same chance to be selected and the applica-
tion of the method is subjected to the availability of a reliable and
up-to-date list.
The Stratified sampling method can be used when similar items have to
be studied and comparisons are to be made among them. In this context,
stratum means a group of study units either rather similar or placed In
similar situations. The strata may differ from one another, but within each
stratum the study units can be considered similar. In stratified sampling
the items under study are divided into a given number of strata and a
random selection of study units is drawn from each stratum. This obviously
implies the existence of a list of study units. The stratified sampling
technique offers two main advantages: i) for the same sample size the
information obtained are more reliable if the items to be surveyed can be
divided into strata and ii) comparisons among them can be made.
When the list of study units is not available and its preparation
would be too expensive, the cluster sampling method can be applied. In this
case the sampling units consist of a cluster of smaller items, close
together, obtained, for instance, by dividing the survey areas into smaller
ones, where it is possible to make a list of the study units. The size of
the cluster, as a general rule, should be small.
In some instances it would be more convenient to use a multi-stage
gainpling procedure with separate sampling at each stage.
The sample size should be large enough to give reliable results. The
sample would be formed by all the units under study. This approach
being generally Impracticable, the size of the sample should represent a
compromise between the need of reliable results and the availability of
economic resources and manpower. This decision is based on two considera-
tions: 1) on the basis of the resources the number of study units to be
processed is established; 2) the reliability offered by the chosen sample
size is evaluated by using either previous measurements or results from
pilot studies.
SAMPLE SELECTION METHOD USED IN THE SECOND GENERATION
ITALIAN NATIONAL SURVEY
From the above discussion, it appears that the choice of the sampling
method is strictly related to the existing information and to their relia-
bility. When the decision was taken to carry out the national survey on
natural radiation indoors, two distinct categories of Information were
available on the sources and the places of exposure (dwellings).
The first one Includes the geology of soils; the measurements of
absorbed dose rate in air outdoors covering most of the country (measure-
ments carried out in the early seventies); the radioactivity content in
typical building materials and In mineral waters from particular springs;
the radon content in the deep soil of some areas; the radon concentration
-------
and gamma ray absorbed dose rate in dwellings (14, 15). The characteristics
and the use of these information will be discussed below.
The second category includes data on the dwellings obtained through
the general population Census carried out in 1981 (16). These data concern
the following:
i) type of house: isolated family dwelling or family dwelling part of
a development;
ii) characteristics of building materials: concrete or other;
iii) number of apartments in the same building;
iv) period of construction: before 1919, 1919-45, 1946-60, 1961-71,
1972-75, 1976-80, after 1980;
v) source of drinkable water: aqueduct, well, other;
vi) heating system: central or other, type of fuel.
The final Census data are published and available to the general
public. The Census data include also the number of inhabitants and families
in each Unit (Comuni) and District (Region!). These figures are updated
yearly and stored in a file to which our organizations have access.
The choice of a list sampling method was excluded a priori since a
list of all the dwellings is not available. The use of telephone director-
ies was also excluded since the telephone density files - where telephone
density is defined as the ratio between the number of telephones and the
number of dwellings in each District - prepared by the National Telephon
Company showed large variations from place to place. This would have
introduced a strong bias in the sample. Identifiable groups such as health
physicists, civil servants, volunteers, cannot be used for a representative
survey.
The only option left was a probability stratified sampling cluster
technique. The most important point was the selection of the strata. Given
the paramount role attributed to the soil in determining indoor radon
concentration, the results of the geological surveys carried out during the
sixties were analysed. A cooperative effort involving the Department of
Earth Science of the University of Rome, ENEA/DISP and the Istituto
Superiore di Sanita (17) resulted in the partition of the country into 6
geological groupings (Figure 1) (recent sediments, argillaceous flysch,
limestones, Igneous metamorphic rocks, volcanic basic rocks, recent
volcanic rocks) as strata for sample selection purposes.
Using information derived from the Census a further partition could be
envisioned considering the construction period of dwellings which cor-
responds to different construction styles either for the use of building
materials or for architectural design (the seven periods of construction in
the Census correspond to important steps in the evolution of building
typologies). The partition into geographical areas corresponding to signi-
ficant differences in climate affecting the indoor radon concentration
could be used.
The types of stratification examined, although valid from the
scientific point of view, showed some drawbacks. The administrative borders
-------
were not respected and the access to the Census information proved rather
difficult. In fact, during the Census data processing the addresses of the
dwellings were not recorded and the link between the data on the house and
its address is practically impossible. Moreover, analysis of manpower and
economical resources evidentiated that no more than 5000 dwellings located
in 200 Administrative Units could be included in the survey. Being this the
case, the use of a too large number of strata would have diminished the
significance of the information collected.
Furthermore different needs arose during the discussions of the
sampling strategy. First, the radon concentration indoors represents mainly
a health problem and, in this sense, the two leading Institutes considered
important the involvement of local health Authorities and gave them the
opportunity of contributing to the survey. Consequently, the sampling
Figure 1. Partition of the Italian territory into 6 geological areas.
-------
strategy should provide a sample of dwellings statistically representative
of the situations existing in the Administrative District. Secondly, the
knowledge of radon distribution in towns where the building criteria show
large differences (urban and less densely populated areas) is of particular
relevance.
According to the first one of the above considerations, it should be
remembered that in Italy in 1978 the National Health Service was
established by law and that activities involving health matters are de-
legated to the Administrative Districts by the State. The control of
environmental radioactivity is one of these activities. In 1987 the
Ministry of Health issued a Directive to the Administrative Districts
concerning such control and laboratories are being set up for this purpose.
In this context the creation of research units dedicated to radon measure-
ments was considered of paramount importance, as a valid contribution to
the cultural growth at local level. Local Authorities will also support a
program of detector distribution, by applying their own personnel. Not
having relied on the Postal Service, this support acquires a particular
significance and importance.
The final sampling design resulted in a two-stage stratified sample.
The study units are the families (family means dwelling in this context);
the clusters are the Administrative Units stratified according to the
population size (two strata: greater or less than 100.000 inhabitants) and
to the Administrative Districts (21). This type of stratification accounts
for the building characteristics (predominance of multistory buildings in
big towns with respect to less populated areas) style of livings, etc. and
for the geographical locations, i.e., climate influence (Administrative
Districts correspond to different latitude); it also allows to estimate the
average radon concentration and distribution in the District.
In each stratum the sampling proportion is the same: 1/4000 (Italian
dwellings amount to about 20 million for 56 million of inhabitants) and is
established a priori. All the Administrative Units having more than 100.000
inhabitants (50) are surveyed, while the remaining ones 150 are sampled
randomly among the Administrative Units with less than 100.000 inhabitants
(8042).
In each stratum the number of dwellings being sampled is given by:
where:
d., ¦ 1 for the Administrative Units being in the sample
^ = 0 for the others
i - 1...21 ; k - 1,2 ; j - 1...150
q . = sampling proportion of dwellings in Administrative Units
of stratum ik ¦
FINAL SAMPLING DESIGN
-------
= P^-J f*kJ
Hj d±kj fikj
f... = number of dwellings in Administrative Unit j of stratum
J ik
f' , . = number of sampled dwellings in the Administrative Unit
j of stratum ik
P = final sampling proportion
The clusters are sampled at central level using the updated Census
files, the dwellings are sampled at local level where the family registers
are available. On the basis of the measurements carried out in previous
years in Italian dwellings, the sample size chosen (5000 dwellings
distributed in 200 Administrative Units) will allow to obtain an indoor
radon concentration distribution where the threshold of the 95th percentile
-------
will be evaluated with a standard error less than 10%. For an average
Administrative District with 250 dwellings the same threshold will be
affected by a standard error around 25%. The distribution of the dwellings
in the sample is shown in Figure 2.
The survey is supplemented with a questionnaire carefully studied. It
contains a set of information which should allow to correlate indoor radon
concentration with other relevant parameters. The questionnaire will be
filled by personnel in charge of the distribution of the dosemeters,
trained for the purpose.
CONCLUSIONS
The efforts made to carry out the sampling represent an example of a
cooperative process among central Institutes and local Authorities, which
was deemed important for two main reasons: a national survey implies
financial and manpower burden for a country and has to be carefully planned
in order to get the best information, considering also the social and
economical implication which could derive from it; the involvement of the
local health Authorities contributes to the creation of a correct expertise
in environmental problems.
REFERENCES
1. Nazaroff, W.W. and Nero, A.V. (eds.), Radon and its decay products in
indoor air, New York, 1988.
2. United Nations Scientific Committee on the Effects of Atomic
Radiation. Ionizing radiation: sources and biological effects, New
York, 1982.
3. United Nations Scientific Committee on the Effects of Atomic
Radiation. Ionizing radiation: sources and biological effects, New
York, 1988.
4. Radiation Protection Dosimetry, 7:1, 1984.
5. The Science of Total Environment, 45:1, 1985.
6. Bruno, R.C., Verifying a model of radon decay product behaviour
indoors. Health Physics, 45:471, 1983.
7. Porstendorfer, J., Behaviour of radon decay products in indoor air.
Radiation Protection Dosimetry, 7: 107, 1984.
8. Nazaroff, W.W., Moed, B.A. and Sextro, R.G., Soil as a source of
indoor radon: generation, migration and entry. In: Nazaroff, W.W. and
Nero, A.V. (eds.), Radon and its decay products in indoor air, New
York, 1988, p. 57.
9. James, A.C., Lung dosimetry. In: Nazaroff, W.W. and Nero, A.V. (eds.),
Radon and its decay products in indoor air, New York, 1988, p. 259.
10. ICRP. Lung cancer risk from indoor exposure to radon daughters.
Publication 50, Oxford, 1987.
11. Committee on the Biological Effects of Iozining Radiation. Health
risk of radon and other internally deposited alpha-emitters.
Washington D.C., 1988.
-------
12. ICRP. Principles for limiting exposure of the public to Natural
sources of radiation. Publication 39, Oxford, 1984.
13. Cochran, A., Random Sampling, New York, 1981.
14. Campos-Venuti, G., Grisanti, A., Grisanti, G., Risica, S., Simula, S.
and Borio, R. An indoor radon study to test the methodology for a
national survey. Radiation Protection Dosimetry. 24: 379, 1988.
15. Sciocchetti, G., Baldassini, P.G., Battella, C., Bovi, M. and
Cotelessa, G. Esposizione alle radiazioni naturali in Italia. Energia
Nucleare. 1: 61, 1989.
16. ISTAT (ed.), 12° Censimento della popolazione: dati sulle caratteristi-
che strutturali della popolazione e delle abitazioni. Roma, 1988.
17. Benassai, S., Campos-Venuti, G., Farchi, G., Mancioppi, S., Mariotti,
S., Piermattei, S., Risica, S. and Tommasino, L. Italian survey to
evaluate the average effective dose equivalent due to radon indoors.
In: Proceedings of the Seventh International Congress of the IRPA,
Sydney, 1988, p.224.
The work described in this paper was not funded by the U. S. Environmental
Protection Agency and therefore the contents do not necessarily reflect the
views of the Agency and no official endorsement should be inferred.
-------
B-IV-4
FIRST-PHASE STUDY DESTGN FOR THE U. S. NAVY RADON ASSESSMENT
AND MITIGATION PROGRAM fNAVRAMPI
R. B. Gammage, D. L. Wilson, C. S. Dudney, and T. G. Matthews
Oak Ridge National Laboratory
Health and Safety Research Division
P. O. Box 2008
Oak Ridge, Tennessee 37831-6383
ABSTRACT
In 1988, the Navy initiated a multi-year program for the assessment and mitigation of
radon inside buildings at its worldwide distribution of bases. During the first two years of the
program, a survey is being made of indoor radon levels in residences occupied by Navy personnel
and their dependents. In addition, a small random sample of other structures is being monitored
for elevated radon. Passive alpha-track detectors, numbering about 25,000, are being used as
monitoring devices. A substantial fraction of the monitors (20%) are being used for quality
assurance. Data management programs have been developed to record the chain of custody of the
monitors and handle the associated questionnaire data. Program objectives and implementation
emphasize quality assurance, records maintenance and monitor placement and retrieval.
-------
INTRODUCTION
During the last decade there has been a gradual realization that indoor concentrations of
radon pose a significant national and international problem for radiation exposure to the lung.
For example, Nero et al. (1) estimated that about one million U.S. homes have radon levels
exceeding 8 pCi L"1. The U.S. Environmental Protection Agency (EPA) is currently recommending
(2) that all residences be screened for radon. The recent Indoor Radon Abatement Act (3)
declares the national goal is that "air within buildings in the United States should be as free of
radon as the ambient air outside of buildings."
The EPA recommends that residences be screened for radon and that 4 pCi L"1 be the level
above which remedial action should be considered (4). This action is predicated more by pragmatic
experience in successfully mitigating houses than it is by health risk considerations. The first-phase
screening portion of the U.S. Navy Radon Assessment and Mitigation Program (NAVRAMP)
project aims to identify with reasonable certainty those bases at which one or more buildings have
a time-integrated indoor radon concentration of 4 pCi L'1 or more.
The Department of the Navy (DON) has 100,000 structures under its jurisdiction, ranging
from detached, single-family houses to large hospitals. To investigate radon levels in every structure
in an acceptable manner would be a very time consuming and expensive, if not impossible, task.
In lieu of investigating every structure, DON established the NAVRAMP project which includes
an initial phase to establish priorities at U.S. naval facilities for additional radon assessment and
possible mitigation. This report will describe the measurement technique and sampling strategy
used for the prioritization phase. In addition, it will discuss preliminary plans for analyzing and
reporting data.
MEASUREMENT TECHNIQUES
ALPHA TRACK DETECTOR (ATD)
The naval bases being examined include remote and inaccessible sites. Mail service is often
non-routine. The charcoal canister (CC) method for radon screening is quite unsuitable for use
in such situations; it is necessary to effect prompt return of exposed charcoal canisters back to a
laboratory for counting in order to obtain reliable results. The plastic, alpha-track recording
detector, in contrast, provides a nearly non-fading record of any radon exposure. Delay in returning
an exposed ATD to the laboratory for development has little effect on the result. The ATD was
thus chosen for the NAVRAMP project, even though the ATD is about twice as expensive as the
CC.
VENDOR SELECTION
The selection of a commercial vendor to supply the ATDs and provide readout service
was made using competitive procurement procedures. Specifications describing NAVRAMP
requirements were prepared and distributed to all known U.S. vendors having proficiency
certification. These were vendors currently listed as having passed the most recent round of EPA's
-------
Radon Measurement Proficiency Program. The vendor providing the lowest bid was inspected by
our Program's Quality Assurance (QA) specialist to verify that the vendor's internal procedures
were acceptable.
SAMPLING STRATEGY FOR FAMILY HOUSING
EXPOSURE TIME
The EPA recommends screening at a time of the year when the building can be closed up
as much as possible; the intent is to ensure that measurement is made at the time of highest and
most stable radon concentration (4). It is also recommended that the ATD be exposed for between
one and three months so that accurate measurement can be made at concentrations of a few pCi
L"1. For an integrated exposure of 40 pCi L"1 days, the coefficient variation for a commercial ATD
is typically ±25%.
It was necessary to determine which period of the year was the one most appropriate for
making a measurement at a particular naval base. Climatic conditions vary in extreme from cool
or cold for the whole year (e.g., Adak, Alaska) to tropical (Guam or Subic Bay, Philippines). At
other locations the heating or cooling season may be very short (e.g., Key West, Florida). To
spread the work load evenly and mesh with the U.S. Navy timetable for completing the screening
phase, the following course of action was adopted: (1) exposures would be made for one year at
Bases which are in cool-climate zones and for which the buildings remain closed for all or most of
the year; (2) in warm or tropical zones, exposures would be made for at least three months when
the weather is the hottest and buildings are closed and air conditioned. The first screening results
for some Bases with tropical climates will be forthcoming in late 1989.
SELECTION OF SAMPLING SITES
Worldwide, the Navy's inventory shows approximately 95,000 family housing units and
55,000 non-housing structures. The selection of sampling sites at individual bases was a non-trivial
task. The approach used for family housing was to obtain a copy of the Navy's data base, update
it, and obtain copies of maps of family housing on each Base. We then prepared lists for all Bases
describing which structures required detectors. Ninety percent of the detectors are devoted to
family housing. The greatest concerns for exposure are in family units where occupancy times can
be high and children are often involved.
DATA BASE UPDATE
For the prioritization phase of NAVRAMP, the fundamental group of statistical units for
sampling was the family housing area. A housing area is a group of houses, town houses, or
apartments, of similar structure, built around the same time, and located close together. Using
housing-area data bases, the DON Facility Category Code data base, and maps of the facilities,
selections were made based upon the following factors: (1) size of the housing area, (2) structural
types, and (3) location of the area relative to other housing areas. The available sources of
information were cross-correlated and anomalies were resolved by contacting Base housing officers
(or their staff). The information was used to update a data base of housing area information. In
-------
all cases, the ATDs at a given base were distributed in as wide a geographical area as possible.
After the selections were made, buildings to receive a detector were identified and marked on the
maps. Lists were prepared of buildings in each housing area to receive detectors. If fewer than
33 living units were located in a housing area, every unit was selected to receive an ATD.
Otherwise, enough buildings were selected so that detectors could be placed in 33 units. Discretion
to reassign a detector to an alternate site was given to Base personnel who actually placed the
detectors, in the event that the initial assignment was in error or unforeseen difficulties arose with
a given building.
SAMPLING STRATEGY FOR NON-HOUSING STRUCTURES
Table 1. Allocation of detectors
During the planning stages of the
program we discovered that the central
Navy data base for non-housing structures
was not as current or complete as the
program required. To avoid delays in the
implementation of the program, we
performed a limited evaluation of radon in
(1) hospitals, (2) brigs, (3) child care
centers, (4) schools, and (5) bachelors'
quarters. The main purposes of these
evaluations were to familiarize Base
personnel with the program, obtain
additional information on structural and
mechanical systems, and obtain an accurate
inventory of non-housing structures.
Detectors were allocated, either on a
per-Base or a per-building basis, as shown
in Table 1. A more detailed evaluation of radon in non-housing structures will be performed in
later phases of the NAVRAMP program.
DATA QUALITY ASSURANCE
In any large survey, quality control is
essential. EPA has prepared a report
describing suggested protocols for radon
surveys (5). Table 2 summarizes the
recommendations from that report. For this
study, the EPA recommendations were
followed as closely as was practical. For each
housing area, two detectors or 5% of the
detectors (whichever was greater) were
allocated for field blanks. Three detectors (or
10%) for each housing area were designated
for replicate (i.e., colocated) measurements.
From every manufacturing lot (~ 660 detectors), 33 detectors were set aside for laboratory spiked
to non-housing structures.
Number of detectors
per building per facility
Hospital 3
Brig 2
Child Care Center 2
School 2
Bachelor Quarters 4
Table 2. EPA recommendations for
data quality assurance.
Field Blanks
5%
Colocated Detectors
10%
Positive Controls
a few%
Negative Controls
a few%
(Lab Blanks)
-------
samples and blanks. Eight detectors (of each 33) were held as laboratory blanks. Two of these
were kept in the original envelope, the rest were opened and kept with the unopened detectors
in a radon-free, environmentally controlled environment. For two months, fourteen detectors (12
opened, 2 closed) were exposed at 6 pCi L'1 of radon in an environmentally controlled chamber.
During the exposure, two detectors were kept in the original envelope. For two months, five
detectors (three opened, two closed) were exposed to 100 pCi L"1 in an environmental chamber.
The other six detectors from each lot were held in reserve in case of mishap or for confirmatory
purposes.
In addition to the quality control samples handled by laboratory personnel, there were
quality control samples handled by field personnel. At about 10% of the indicated sampling sites,
field personnel were instructed to deploy duplicate, i.e., colocated detectors. Travel-type control
detectors were shipped to each facility and will be returned without ever having been opened.
Exposure-type control detectors were shipped to each facility and will be stored on site until just
prior to the return shipment. At that time, the detectors will be sequentially removed from their
foil pouches, handled as if they were to be deployed, repackaged, and returned.
Verification of chain of custody of detectors and related data are also very important. A
shipment tracking system was developed to record when detectors were shipped, when they were
received, when a custody report from the point of contact was received, and when placement data
sheets were received. A data form shipped with each detector was to be filled out when the
detector was placed at a designated sampling site. These forms record information about
monitoring location and time. The placement data forms will be returned and entered into a
computer data base. One month before desired retrieval times, notices will be sent to facilities,
notifying the points of contact when to begin retrieval. Shortly before the desired time of retrieval,
a retrieval data form will be generated for each detector and forwarded to the responsible facility.
J)ata from placement and retrieval will be compared to reduce the incidence of misrecorded
information on detector placement.
Instruction packets were prepared to guide Navy personnel in the proper placement of the
detectors in: (1) family housing units, (2) bachelors' quarters (officer and enlisted), (3) schools, (4)
child care centers, (5) hospitals, and (6) brigs. These packets will be revised and reissued for
subsequent phases of NAVRAMP.
IMPLEMENTATION
POINTS OF CONTACT
Implementation of the above sampling design relied on extensive cooperation from Navy
personnel, both civilian and military, at each facility. Early in the programs, each Base provided
us with points of contact for family housing and for operational structures.
-------
DISTRIBUTION OF DETECTORS
Each shipment of detectors for placement in family housing included the requisite number
of detectors, a chain of custody letter/receipt, maps indicating the locations of structures selected
for detector placement, instruction packets, and a placement data form for each detector. Each
shipment of detectors for placement in non-housing structures was divided into subshipments
according to the kind of structure(s); this subdivision was necessary because there were different
points of contact for hospitals, brigs, bachelor quarters, schools, and child care centers. Each
subshipment included a chain of custody letter/receipt, the appropriate instruction sheet and a
detector placement data sheet. In addition, there was a list, derived from our inventory data base,
of structures of the appropriate type at that facility and there was a request for building floor plans
and additional.
HOTLINE INQUIRIES
To facilitate implementation of this plan by field personnel, a hotline phone number was
established to answer questions from the field. The phone was manned during business hours,
Eastern Standard Time. At other times, an answering machine recorded messages received, and
responses were generated as soon as possible.
DATA MANAGEMENT
To date, over 22,000 detectors have been shipped to naval facilities worldwide. For each
of these detectors, information is being compiled on the characteristics of the structure where the
detector was placed. Temporal information is compiled on when the detector was shipped to the
facility, received, placed, retrieved, shipped back to us, received by us, and shipped to the vendor.
The radon exposure (units of pCi L"1 d) recorded by the detector is also included. All paper
records, including maps, chain-of-custody letters, and detector placement data sheets, are being
stored in locked filing cabinets kept in well air-conditioned space at Oak Ridge National
Laboratory. To facilitate data entry into computer data bases and assignment of all received
information to the correct monitoring site and time, bar coded strips are being used extensively.
Bar coded strips are applied to radon detectors and to detector placement sheets. Bar code readers
determine detector identity information from the sheets to reduce the occurrence of transcriptional
errors. Data entry programs have been developed which compare newly entered data with data
already on hand to ensure that only valid identification numbers for radon detectors and Navy
buildings are entered.
DATA ANALYSIS AND REPORTING
An integral part of the development of a work plan for NAVRAMP was consideration of
how the data would be analyzed.
-------
MINIMUM SAMPLE SIZE
From statistical analysis of assumed populations of houses, it was concluded that if 14% of
all units in a population have radon concentrations above 4 pCi L"\ then if at least 30 units are
sampled and none of the results are found to be above 4 pCi L"\ the probability of there being
any house with elevated radon is less than 0.01 (i.e., 1%). Allowing a 10% outage for inadvertent
losses, it was decided to sample 33 units in each housing area with 33 or more units.
DISTRIBUTION OF RADON CONCENTRATIONS IN HOUSING AREAS
After the data are collected, it is possible to assess the distribution of the sample and infer
limits on the distribution within the entire population. From statistical analysis, we concluded
that, if the population is lognormal, it is possible to estimate the worst-case distribution for a given
level of confidence. Using this approach for each housing area, we expect to answer the questions:
1. Does the population follow a lognormal distribution?
2. What is the fraction of the population above any given action level?
3. If we assume that the statistics calculated from the data underestimate the true
geometric mean and true geometric standard deviation, what is the fraction of the
population above the action level?
Based upon the analysis of radon data from housing units, naval facilities will be ranked for priority
scheduling for a more detailed screening survey. No final ranking scheme has been approved, but
those facilities found to have a substantial probability of one or more units with radon in excess
of 20 pQ L"1 will probably receive the highest rank.
ANALYSIS OF DATA FROM LARGE STRUCTURES
The understanding of radon behavior in buildings with complex heating, ventilating, and air
conditioning (HVAC) systems is very limited. EPA's research in schools has shown that in some
buildings, the induced pressure in different rooms in the same building due to the HVAC system
can range from +10 Pa to -20 Pa. It was also shown that even small changes in pressure can
markedly affect radon ingress in ground floor rooms. Therefore, we recommend that data from
each large building be treated on an individual basis.
This is the first of several reports that are planned. After detectors placed in cooling
season facilities are returned to the vendor and results are available to us, the data from those
facilities will be summarized as described in this section and a report submitted. Electronic data
bases will be prepared and provided to die Navy. The data bases will include information on
monitoring sites and times, structure characteristics, and radon concentrations.
-------
ACKNOWLEDGEMENTS
This research sponsored by the U.S. Department of Energy and U.S. Navy under contract
DE-AC05-840R21400 with Martin Marietta Energy Systems, Inc.
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not necessarily reflect the views of the Agency and no
official endorsement should be inferred.
"The submitted manuscript has been
authored by i contractor of the U.S.
Government under contract No. 06-
AC05-840W21400. Accordingly, the U.S.
Governmont retains a nonexclusive,
royatty-fres license to publish or reproduce
the published form of tha contribution, or
allow others to do so. for U.S. Government
purposes."
REFERENCES
1. Nero, A V.; Schwehr, M. B.; Nazaroff, W. W.; Revzan, K. L. Distribution of airborne
radon-222 concentrations in U.S. homes. Science 234:992-997; 1986.
2. EPA Office of Radiation Programs, "Interim Indoor Radon and Radon Decay Product
Measurement Protocols," Report #EPA 520/1-86-04 (Available from NTIS, Springfield, VA),
April, 1986.
3. Amendment of the Toxic Substances Control Act, Section 1. Title III...Indoor Radon
Abatement, Section 301, National Goal, U. S. Congress, Washington, D.C., January 25,1988.
4. Ronca-Battista, M., Magno, P., and Nyberg, P. C. 1988. "Standard Measurement Techniques
and Strategies for Indoor mRn Measurements." Health Physics 55:67-69.
5. Literature provided to Federal Agencies by E.P.A Office of Radiation Programs, April
1989.
-------
B-IV-5
RADON IN WATER AERATION SYSTEM
OPERATIONAL PERFORMANCE
by: Bruce L. Lamarre
North East Environmental Products, Inc.
West Lebanon, NH 03784
ABSTRACT
North East Environmental Products, Inc. is a manufacturer of residential scale aeration
systems for removal of radon and volatile organic chemicals from private water supplies. This
paper is a review of the operational history of residential scale point of entry (POE) radon
aeration systems. Emphasis is placed on the difficulties and solutions encountered in actual
installations caused by both mechanical difficulties and water quality parameters. A summary of
radon reduction efficiency is presented for wells with radon concentrations from 21,000 to
2,600,000 pCi/L. A discussion of customer concerns and attitudes is presented along with other
areas for further technical improvement. Training techniques for dealers and installers are
also discussed.
An update of the current status of the radon in water industry includes current sales
volumes as compared to the potential market and an update on the radon in water MCL standard
setting process from an industry perspective.
BACKGROUND
RADON health effects
Radon is slighdy soluble in water and can enter a home along with the water from a
drilled bedrock well. Because radon is a volatile gas it is quite easily removed from the water
when it is used for typical household activities such as bathing and washing dishes. Some
estimates based on assumptions about water use patterns and house construction details predict
that each 10,000 pCi/L concentration of radon in the water supply will translate into an indoor air
concentration of 1 pCi/L. EPA studies have apparently confirmed that this approximation is
reasonably accurate for an "average home". If a house has a relatively low air exchange rate this
concentration will be higher.
-------
According to health experts, it is the alpha particles from the radioactive decay of radon
and its daughters that cause the most severe health threat. Alpha particles arc relatively large and
will only travel a short distance before striking other matter and give up their high energy. Alpha
particles will only travel a few centimeters in air and only, at most, a few millimeters in the
human body. Therefore, when a person inhales air containing a high concentration of radon, the
most likely organ to be effected by the alpha particles is the lung. In addition, two of the decay
products of radon are also alpha emitters. Polonium-218 and polonium-214 are daughters of
radon. When radon decays, these elements are formed one after another. Since these elements
are solids they attach themselves to dust particles in the air and are carried into the lungs with
each breath of air where they decay and give off harmful alpha particles.
Ingestion of water containing radon has been thought to be much less of a problem than
inhalation. Because alpha particles cannot even pass through a piece of paper it appears to be
very unlikely that when radon, or one of its daughters, decays while located in the stomach or
intestine, the alpha particle will travel to and strike the lining of the digestive track. It appears to
be more likely that the alpha particle will be absorbed by the fluids in the digestive track and
dissipate its energy harmlessly. Of course at extremely high radon concentrations the risk of
ingestion damage may become significant.
Since radon is soluble in blood it can be transported throughout the body and potentially
affect other organs of the body. Studies are being conducted at the present time by several
researchers to establish whether or not radon and its daughters do, in fact, present a health hazard
by the ingestion route. Two preliminary papers have in fact shown a very strong correlation
between radon in water and various types of cancer. The results are both surprising and
alarming.
It has been estimated that of the approximately 120,000 lung cancer deaths each year
between 5,000 and 20,000 of those cancers were caused by radon gas exposure. This is the
second leading cause of lung cancer. The leading cause is, of course, cigarette smoking. Of the
lung cancers caused by radon, the EPA Office of Drinking Water has stated that between 1,000
and 1,800 of these cased can be attributed to radon entering homes through their water supplies.
This range of cancer deaths is greater than the cancer risk from all the other water supply
contaminants combined. Radon contamination of water supplies is certainly a large and serious
problem.
AVAILABLE TREATMENT ALTERNATIVES
Water Treatment - Aeration
Water supplies can be treated through the use of aeration or carbon adsorption
techniques. Aeration techniques simply allow the radon to volatilize from the water and exhaust
it outdoors where it can disperse harmlessly. Adsorption methods collect the radon on activated
carbon and allow it to decay in place.
-------
Aeration devices are relatively simple in that through various methods they contact the
water with enough air to volatilize the radon. The treated water then can be pumped into the
home water system.
There are basically four types of aeration processes that can be used for residential water
treatment: spray aeration, packed columns, diffused aeration and a new process called
horizontally extended shallow aeration.
In spray aeration, shown in Figure 1, untreated water from the well is sprayed into a tank
through a fine mist spray nozzle. The spray nozzle generates a large amount of water surface
area from which radon volatilizes. Usually a small air blower is used to pass a small amount of
air through the equipment to carry the radon out of the tank and to vent it outside of the home.
Typically a simple spray nozzle will remove approximately 50% of the radon from the untreated
water. In order to achieve higher removal efficiencies the water must be resprayed and retreated
several times. Essentially any treatment efficiency desired can be achieved with this system.
The disadvantage of this process is that the water must be repumped several times, (4 to 5 times
usually) and that in order to have a ready supply of treated water the holding tank must be quite
large(about 100 gallons). Equipment using this process is available under the trade name No-
Rad, (patented).
Air lo Vtnl
Pump
Figure 1. Spray Aeration
-------
Packed columns have been extensively used for removing volatile organic chemicals
from contaminated groundwater supplies. These systems can be scaled down for use in a
residential setting for removal of very volatile radon gas. Figure 2 shows a residential scale
system in which the well water is sprayed into the top of a small air stripping column (available
under the trade name Clearadon, patented). The column is filled with about five feet of a
common inert dumped packing material. As the water falls down through the packing a large
amount of surface area is generated from which the radon can volatilize. A small air blower
forces air up through the packing which carries the radon gas out of the column to an outdoor
vent. The efficiency of these systems has been shown to be approximately 90 to 95%. The
principle limiting factor in packed column aeration of radon is the height available for the air
stripping column. Maximum practical packing depth in most residential settings is six feet which
produces a removal efficiency of about 95%. For relatively low levels of radon contamination
(i.e. up to 20,000 pCi/L) this is entirely adequate. Above this level the packed column system
becomes impractical: that is, at a radon MCL of 1000 pCi/L.
Figure 2. Air Stripper
A third aeration method is diffused aeration. Figure 3 shows this type of process. One
supplier manufactures this product (patented) under the trade name of The Stripper. The
contaminated well water is pumped into the first of two or more aeration tanks. Air is forced
into the bottom of these holding tanks through fine bubble diffusers located near the bottom of
each tank by a relatively high pressure (10-15 inches of water column) air blower. As the air
bubbles rise up through the water the radon volatilizes into the air bubbles. In this case the mass
transfer area for the volatilization of radon from water is generated by the small air bubbles as
they rise through the water. Since each tank is essentially completely mixed, very high removal
efficiencies cannot be achieved in a single tank. Usually from two to six tanks are required to
achieve better than 99% efficiency. The efficiency of an aeration tank can be improved by
increasing the residence time of the water in the tank (make the tank bigger), by increasing the
number of bubbles (increase the air flow rate and the size of the diffuser) or by baffling the tank
so that it performs like a plug flow reactor rather than like a completely mixed reactor.
-------
Figure 3. Diffused Aeration
The disadvantages of this system are that a relatively high pressure air blower is required
(10 to 15 inches of water column) and that the air holes in the diffuser foul up easily because
they are very small, about 0.025 inches.
» ExhkuM
Tran*f»r Us*
Pump
Figure 4. Horizontally Extended Shallow Aeration
Figure 4 shows the forth type of aeration device, the Clearadon 310 radon removal
system (patent pending), which uses a horizontally extended shallow aeration tray design to
contact the water and air. Water from the drilled well is piped to the aeration unit where it is
sprayed into the aeration tray. The water flows across the tray between the baffles as air is
blown up through holes in the tray. The air forms a froth of water on the tray creating a very
large area for volatilization of the radon from the water. The air evaporates more than 99.9% of
the radon, which is vented outside the home. The cleaned water then collects in the bottom of
the radon removal unit and is pumped into the water pressure tank. This system, as with the
previous systems, is completely automatic and requires very little maintenance. The shallow
-------
aeration system has three principle advantages over the other three systems. First, the aeration
tray is smaller than two feet in diameter and only 10 inches high. The overall height of the
system can therefore be as little as three feet. The complete system can therefore be smaller, and
shorter than the other designs. Second, the air pressure required for operation is only 4 to 5
inches of water column so a much less expensive type of air blower can be used. And third, the
air holes in the aeration tray are much larger than those used in the diffused aeration design (3/16
of an inch) and therefore fouling problems are virtually eliminated. The main disadvantage of
this system is that it uses approximately 100 cfm of air whereas the previous systems use
between 10 and 50 cfm. In some homes an outside air source may be necessary to prevent
basement depressurization.
In general, aeration devices have several advantages and beneficial side effects. Aeration
devices do not accumulate radioactive elements and therefore do not present a radiation exposure
problem and they do not, and will not, need to be licensed as low level radiation sources.
Aeration devices can also remove hydrogen sulfide, carbon dioxide, and methane from the water
and can assist in the removal of iron and manganese. Aeration devices can also be used to
remove volatile organic chemicals such as gasoline components from water supplies.
Water Treatment - Activated Carbon
Adsorption devices remove radon by adsorbing the radon onto the internal surface of a
specially prepared "activated carbon". Once adsorbed onto the carbon the radon continues to
decay and give off radiation. However, the equipment is usually not located in the immediate
living area of the home. After about three weeks, the amount of radon being adsorbed on the
carbon equals the decay rate of the radon already adsorbed and the system reaches a steady state.
The radiation given off by the unit therefore levels off at some point dependent on the radon
level in the ground water.
The advantage of this type of system is that it has very few moving parts and,
theoretically, should have quite a long useful life. The disadvantages are the radioactive buildup
on the carbon, which may or may not be a real problem depending on the specific situation, the
possibility of fouling of the carbon bed, and difficulty in predicting removal efficiencies due to
the effects of unknown factors. Contaminants in the water such as iron, manganese, and calcium
will be filtered out by the carbon and will eventually plug it. The systems can be cleaned of
solids by back washing but this is not completely effective. Activated carbon should not be used
for radon treatment of water supplies having over 5,000 pCi/L. The EPA has listed aeration as
the best available technology for water system treatment of radon. The liability issue related to
the accumulation of radioactive material on the activated carbon alone should be sufficient
reason for informed people not to use activated carbon for residential treatment of radon.
-------
OPERATIONAL DIFFICULTIES
There have been three types of operational problems with the shallow tray aeration
systems:
• Water leaks
• Noise
• Air in water
As with all types of new mechanical equipment, there have been problems with some of
the mechanical components of our earlier systems. The principal problem has been water leaks
in the piping system.
WATER LEAKS
Rigid PVC water piping is theoretically, and in the shop, great stuff. It's easy to work
with, it has plenty of strength, and it won't corrode. However, threaded PVC piping only has
enough mechanical strength if it is not over tightened. It's amazing how easy it is to over
tighten a PVC fitting. In about a half dozen cases we have had installations in which a threaded
fitting has split shortly after installation. This, of course, caused varying amounts of water
damage in some of the homes which we were responsible for correcting. It did not take many of
these instances for us to change our design and replace all of the PVC fittings with copper.
Another problem with the PVC is that when it is heated up it looses much of its strength.
One of our units was installed at an EPA funded test site in New Hampshire. An unusual
combination of settings in the experimental piping set up caused the repressurization pump to run
continuously for several days against a closed valve. The water in the pump got hot and the
suction piping softened up and swelled. The swelling caused a small leak at a union near the
pump.
NOISE
Most residential wells with high concentrations of radon are drilled bedrock wells. A
drilled well almost always has a submersible pump mounted in the well to supply water to the
house. Submersible wells are virtually silent so home owners are not used to hearing any noise
associated with their water supply. Therefore any new noise caused by water treatment
equipment may be bothersome to some home owners.
The shallow tray aeration devices have a noise quality which is vary similar to a normal
hot air furnace. The noise is not unusually loud but it is unfamiliar in association with the water
system. The earlier version of the product had a metal frame that tended to amplify the noise
-------
made by both the repressurization pump and the blower. In order to eliminate this amplification
the product was redesigned to completely do away with the metal frame. With out the frame, the
pump and blower still generate noise, but at a much lower level.
We also discovered, as I am sure many radon mitigation contractors have also
discovered, that rigidly attaching piping to floor joists can also greatly amplify noise generated in
pumps and blowers. At a test site installation in Connecticut a very significant noise problem in
the living room was eliminated simply by replacing the rigid water pipe clamps with vibration
isolation pipe hangers. We later replaced the original Clearadon II with a Clearadon 310 and the
noise from the unit was virtually unnoticeable.
AIR IN WATER
In approximately one installation in twenty, we have received complaints from the
homeowners that they have air spurting out of their faucet when they turn it on for the first time
in several hours. This had been a very difficult problem to correct because of the infrequency
and inconsistency of the appearance of the problem.
The cause and solution of the problem turned out to be very simple. Flow restrictors are
installed on both the inlet and outlet of the aeration unit. Under some water use conditions and
with some well conditions, the amount of water allowed into the aeration unit was slightly less
than that being pumped out. This marginal condition allowed a small amount of air to be drawn
into the water. The problem was solved simply by installing a slightly larger flow restrictor on
the water inlet.
-------
RADON REDUCTION EFFICIENCY SUMMARY
A well located in Dunbarton, N.H. was found to have a radon level of 2.6 million pCi/L.
At the time, October of 1988, this was the highest concentration ever found in a private water
supply well. Since that time, another well located in Lyons, Colorado has been found to contain
2.9 million pCi/L.
A Clearadon n aeration system was installed at the Dunbarton home. An independent Lab
collected and analyzed a set of samples in early February. The results are as follows:
Radon Influent (pCi/L) Radon Effluent (pCi/L)
2,565,505
2,643,196
2,626,649
2,468,333
2,661,495
2,788,235
2,773
1,786
1,909
1,754
1,699
1,686
2,625,589 +/- 106.201 1,934 +/-419
Removal efficiency - 99.93%
The water quality at this home was excellent with the exception of the radon.
-------
Table 1 lists representative performance data for some of the over 100 units we have installed in
the field.
TABLE 1. RADON REMOVAL EFFICIENCY
Type of
Unit
Inlet Radon
Cone., pCi/L
Outlet Radon
Cone., pCi/L
% removal
Water
Flow
Rate, gpm
Clearadon 310
17,000
130
99.24
8.5
500,000
1500
99.70
8.5
251,400
540
99.79
7
172,450
380
99.78
7
125,200
300
99.76
7
220,140
210
99.90
5
141,870
200
99.86
5
135,900
220
99.84
5
206,900
40
99.98
3
165,950
75
99.96
3
156,800
40
99.98
3
Clearadon II
50,000
>100
99.8
6
32,000
110
99.7
6
21,000
>100
99.5
6
90,000
2100
97.67
6
57,000
510
99.1
6
21,000
>100
99.5
6
2,600,000
1500
99.94
5
-------
RADON REMOVAL VS. TREATMENT RATE
% Radon
Removal
100 -1
99.8 -
99.6
99.4
99.2 -
99.0
Water Treatment Rat*
(QPM)
Figure 5
Figure 5 shows the relationship between treatment efficiency and water flow rate through
the equipment. As one would expect, the percentage reduction of radon decreases with
increasing flow rate when all other factors are held constant
-------
CUSTOMER CONCERNS
Other than the mechanical problems discussed earlier, the principal concern of our
customers has been with the acceptability of any radon treatment device to future buyers of the
home. Since radon and radon treatment equipment are unfamiliar topics to most people it is
understandable that this would be a common concern. We have found that in a small percentage
of cases (less than 10%) this does discourage a potential customer from purchasing the
equipment. Of course it is quite likely that many people with this concern never contact us in the
first place. I believe that our experience under estimates the actual proportion of people that
have this concern.
TRAINING
Dealer training is one of the most important aspects of distributing treatment equipment
such as this. To a person familiar with water treatment and water treatment equipment,
installation and maintenance of a Clearadon unit is very straight forward and simple. To those
unfamiliar with these areas there is a great deal to learn. We have therefore set up training
courses for all of our dealers. Prior to installing any units we now require that the dealer's
service people attend a one day training course at our factory. We are, of course, on call to our
dealers to answer any questions about unusual situations as they come up. The equipment is
normally sold only to our authorized and trained dealers. The equipment is not designed to be
installed or serviced by the final consumer.
CURRENT STATUS OF THE RADON IN WATER INDUSTRY
In comparison to the size of the potential market for radon in water treatment equipment,
the actual market is virtually nonexistant. There are approximately 18 million drilled private
wells in the US. At a radon in water standard of 200 pCi/L approximately one half of these wells
would require treatment. That is 9 million wells. In the past year between 150 and 300 aeration
systems were installed. This low number of installations is due to the lack of a Maximum
Contaminant Level standard for radon.
Representatives of the EPA have repeatedly stated that radon in water presents a larger
risk of death that all other water contaminants combined. The risk posed is greater than the
risks posed by lead, arsenic, hazardous waste site contamination and gasoline contamination of
wells combined. Without a radon standard the vast majority of the population will not know
about radon contamination of water and will not treat their water supplies. By EPA's own
estimates, between 1000 and 1800 people die each year due to exposure to radon originating in
their water supply.
-------
The current market for radon water treatment equipment is currently limited to the
exceptionally health conscious, wealthy population and to some real estate transactions. The
market is evenly split between these two market segments.
The standard will be established for public drinking water supplies only since the EPA
and other federal agencies do not have clear mandates to regulate environmental matters in
private homes. However, it is obvious that if a standard is set for public water supplies it is
prudent for individual homeowners to follow the same standards for their own protection. It is
also possible that state Public Health Departments may, once a federal standard is set, establish
their own water supply standards similar to those for bacterial contamination of private wells.
CURRENT STATUS OF THE RADON IN WATER STANDARD
Under the 1986 amendments to the Safe Drinking Water ACT (SDWA), the EPA is
required to publish Maximum Contaminant Level Goals (MCLGs) and promulgate National
Primary Drinking Water Regulations (NPDWRs) with respect to 83 specific drinking water
contaminants according to a specified timetable. This original list of 83 contaminants included
radon. Standards for all 83 contaminants were specified to be issued by June 19,1989. When
MCLGs are published, EPA must also propose Maximum Contaminant Levels (MCLs) that must
be set as close to the MCLGs as is "feasible". The term "feasible" means with the use of the best
technology , treatment techniques and other means, which the administrator of EPA, after
examination for efficiency under field conditions and not solely under laboratory conditions, are
available (taking costs into consideration).
On September 30, 1986, EPA published an advanced notice of proposed rulemaking. In
that notice it was stated that the health hazard posed by radon contamination of water supplies is
probably larger than that posed by all other water contaminants.
A Fact Sheet issued by the Criteria and Standards Division of the EPA Office of Drinking
Water in September of 1988 stated that MCLGs and MCLs for radon would be proposed in
September of 1989 and projected final regulations in November of 1990. An EPA Office of
Drinking Water publication issued in May of 1989 stated that EPA is considering setting MCLs
for radon in the range of 200 to 2,000 pCi/L of water. This publication continued to delay the
ultimate regulation of radon stating that the "publication of the Notice of Proposed Rulemaking
in the Federal Register is presently scheduled for June 1990." We have been informed that the
Office of Drinking Water has all of the technical information they need in order to propose a
radon standard. The only issues holding up the standard are policy and legal questions. As of
this writing, the standard has been delayed again, to September of 1990.
EPA estimates that between 1000 and 1800 people die of lung cancer in the US each year
from radon contamination of well water. Between June of 1989 and September of 1990,
between 1200 and 2200, people have died because of the agency's inaction. Unnecessary deaths
will continue until the EPA meets its Congressionally mandated responsibilities and proposes a
radon standard. Only then will people be made aware of the danger and begin to take corrective
measures.
-------
The work described in this paper was not funded by the U.S. Environmental Protection
Agency and therefore the contents do not necessarily reflect the views of the Agency
and no official endorsement should be inferred.
REFERENCES
Logtin, J.P., Occurrence of Radon, Radium and Uranium in Groundwater, Jur. AWWA, July
1989
Cothern, C.R., Development of Regulations for Radionuclides in Drinking Water, Keynote
Address, NWWA Conference, Radon, Radium and Other Radioactivity in Ground Water:
Hydrogeologic Impact and Application to Indoor Airborne Contamination, April 7-9,
1987
Deland, M.R., 1986, Indoor radon: A New England Perspective, Environmental Education
Report & Newsletter, v. 15, no. 2, pp 8-10
U.S. Environmental Protection Agency, Office of Research and Development, Radon Reduction
Techniques for Detached Houses: Technical Guidance, June 1986, EPA/625/5-86/019
U.S. Environmental Protection Agency, Office of Air and Radiation, A Citizen's Guide to
Radon: What it is and What To Do About It, August 1986, OPA-86-004
U.S. Environmental Protection Agency, Office of Radiation Programs, Reducing Radon in
Structures, 2nd Edition
-------
National Residential Radon Survey
B-IV-6
by: Frank Marcinowski
U.S. Environmental Protection Agency
Office of Radiation Programs
Washington, DC 20460
Robert M. Lucas
Research Triangle Institute
Research Triangle Park, NC 27709
Abstract
The Superfund Amendments and Reauthorization Act (SARA) requires the EPA
Administrator to conduct a national assessment of radon levels where people normally
live and work, including educational institutions.
The National Residential Radon Survey (NRRS) is the first comprehensive effort
to estimate the frequency distribution of average annual radon concentrations
nationwide. Also, the survey will provide data to correlate radon concentrations with
construction characteristics.
A stratified three stage area probability sample was used to randomly select
approximately 12,000 homes. A questionnaire will provide information on living
patterns, house construction, and heating, ventilation, and air conditioning (HVAC)
characteristics. Two to four alpha-track detectors were placed in each home. It is
expected that approximately 5,000 residents will return detectors with readable radon
concentrations.
With this data, EPA will be able to accurately estimate the magnitude of the
radon problem and evaluate the need for subsequent radon programs.
This paper has been reviewed in accordance with the U.S. Environmental
Protection Agency's peer review and administrative review policies and approved for
presentation and publication.
-------
Introduction
The National Residential Radon Survey (NRRS) is mandated under Section
118(k) of the Superfund Amendments and Reauthorization Act of 1986 (SARA, P.L.
99-499) which requires the Administrator of the Environmental Protection Agency
(EPA) to conduct a national assessment of radon "found in structures where people
normally live and work, including educational institutions". The NRRS is addressing the
component of the national distribution of radon concentrations in occupied housing
units.
The results of the NRRS will be used to accurately estimate the magnitude of
the national radon problem, assess the effectiveness of various radon program strategies
(i.e., government/industry capability to address the radon problem, magnitude of State
and Regional assistance, health risks) and evaluate the need for subsequent radon
programs.
Intended Uses of the Data
The primary uses of the survey data will be to estimate the distribution of
annual average radon concentrations in occupied housing units and the correlations
among specific housing structures and radon concentrations.
In addition to estimating the national distribution of radon concentrations, EPA
also plans to study subsets of the surveyed homes to learn how radon levels vary among
smaller segments of the population. A representative sample of homes selected from
each of the 10 EPA Regions will allow estimation of the distribution of radon
concentrations in each region and also highlight differences in radon levels across the
country.
Comparison of radon concentrations will also be made among other subgroups
of the study population. The radon distribution of single family homes will be
compared with multi-unit homes, rental units will be compared to owner occupied units,
and radon levels will be compared among floors within the same structure to
demonstrate how radon concentrations in basements relate to other levels of a home.
A secondary benefit of the NRRS includes collecting data related to health risks.
The survey will give information on the concentrations of radon to which people are
actually exposed to in their homes, as well as collecting data pertaining to the living
habits of the participants. The survey will investigate homes with smokers, homes
without smokers, homes with children, and various other aspects. Although residential
-------
exposure is only a fraction (60-75%) of a person's total exposure, this information
should help EPA estimate the fraction of lung cancers that can be averted through
reduction of radon concentrations in homes.
Sample Design
A well designed scientific survey must have a prespecified well-defined inference
or target population and specified informational requirements on which to base the
sample design. The target population of the NRRS is:
a) All Census-defined housing units that are continuously occupied for 12 months
that are not on military installations and
b) the permanent residence of these housing units.
The specified informational requirements are given in terms of the required
relative precision of estimates of the upper tails of the distribution of annual-average
radon concentrations in the U.S. The objectives used to develop the sample design are
given below.
The first priority of the NRRS is to furnish a scientifically competent estimate of
the national frequency distribution of annual average radon concentrations in occupied
residences. Specifically, the estimate of the percent of homes with radon concentrations
greater than 10 pCi/1 should have a relative standard error no greater than 0.5, if the
estimate of the percent of homes with radon concentrations greater than 10 pCi/1 is
approximately 0.5%.
The second priority of the NRRS is to furnish estimates of the frequency
distribution of annual average radon concentrations in occupied residences for each of
the 10 EPA Regions. The survey should also provide information from which
correlations can be made between radon concentrations, house construction, and HVAC
characteristics. Specifically, the estimate of the percent of homes with radon
concentration greater than 4 pCi/1 should have a relative standard error of no more
than 0.5 for an EPA Region, if the estimate of the percent of homes with radon
concentrations greater than 4 pCi/1 is approximately 7%.
The NRRS is divided into two phases. Phase I of the survey involved identifying
the households to be sampled, interviewing the residents of each household, and placing
the detectors. Phase II includes several periods of panel maintenance, collection of the
detectors, analysis of the data, and production of a final report.
-------
Phase I
The survey population contains only residential households whether detached or
multi-unit, single or multi-family, owner occupied or rental. A three stage area
probability sampling strategy was used to select the housing units for the survey. The
model assumed that the percentage of homes above 4 and 10 pCi/1 varied among each
EPA region and also across major geographic regions within a single EPA region.
Each of the EPA defined regions was divided into substrata depending on the predicted
percent of homes with a specific radon potential (Table 1). Differences between the
assumed distribution in Table 1 and the true distribution will not induce bias in the
survey estimates but only reduce the precision of the estimates. A probability sample
of primary sampling units (PSUs) was selected from each of the 22 defined strata with
a minimum of 2 PSUs per stratum.
Table 1. Assumed Radon Potential Criteria Used to allocate the sample to
Substrata.
Predicted Percent of Predicted Percent of Substratum
Homes > 4 pCi/L Homes > 10 pCi/L Ranking
13
1.0
High
7
0.2
Medium
1
0.1
Low
The secondary sampling units (SSUs) within each PSU consisted of Census
defined blocks or enumerations districts. Each of the SSUs contained at least 25 but
no more than 150 housing units (HUs). Eight SSUs, with a probability proportional to
the number of occupied HUs in the SSU, were selected within each PSU. Each SSU
was visited by experienced field representatives who counted and listed the HUs within
each of the 1,000 SSUs. Of these 1,000 SSUs, seven were identified as being in
military installations and these contained no eligible HUs.
-------
A random systematic sample of a specified number of listed HUs was selected
from each SSU. A total of 11,441 HUs was selected to be visited by the field
interviewers to determine participation in the survey, administer the NRRS
questionnaire, and place the radon detectors. A total of 2,515 HU's were not eligible
for the survey because they were vacant or not year-round occupied HUs (To meet the
eligibility requirements of the NRRS, the residents had to have no firm plans to move
within 12 months or must live within the residence for 9 months out of the year). Of
the 8.926 eligible HUs, 80 percent (7,134) of the occupants fully cooperated by
completing the questionnaire and having detectors placed in their homes. The targeted
number of HUs expected at completion of the one year detector deployment period is
approximately 5,000.
One of the primary data collection instruments of the NRRS is the household
questionnaire which was administered by trained field interviewers. This questionnaire
provided detailed information on the construction type and structure of the sampling
residence, the amount of time spent on each level of the house, the type of HVAC
system, smoking history of residents, occupancy habits, and detector placement
information.
Following administration of the survey questionnaire, year long alpha-track
detectors (ATDs) were placed in participating residences. Detectors were placed at
each level of the home used as living areas up to three levels (excluding basement or
crawl spaces). Single level homes received one detector in the living area and one
in the bedroom area. Multilevel homes received a maximum of three detectors.
Additionally, the basement received a detector if one wall or more was completely
below ground. Any crawl space received a detector if it could be entered from
inside the home, or could only be entered from outside the home and an adult could
stand in it. Each residence received 2 to 4 detectors, with an average of 2.5 detectors
per residence being placed for the survey.
Phase II
One of the principle tasks of Phase II is to conduct panel maintenance of the
participating households. This will be accomplished through postcards and follow-up
telephone calls for three periods during the yearlong detector deployment period. The
purposes of these contacts are to verify the participation of the residents, confirm that
they will continue to inhabit the same residence, convey a positive sense of participating
in a valuable research venture, and determine if any problems exist with the detection
devices.
-------
The first round of panel maintenance has been completed with 0.7% of the
residents no longer wanting to participate, 3.7% of the residents were moving and no
longer were able to participate, 2.3% of the participants reported lost or damaged
detectors, and 3.9% were not contacted after repeated attempts. The damaged, lost,
and exposed detectors were replaced. Of the remaining eligible HUs, 95 percent
(6,548) are still participating.
The detectors will be retrieved through the mail during the summer of 1990.
The ATDs will be analyzed and the results will be sent to the participants in the
survey. Various statistical analysis will be performed on the data to meet our primary
and secondary objectives.
NRRS Schedule
Developed Survey Design
Pretest Data Collection
Administer Questionnaires
and Place Detectors
Panel Maintenance
Retrieve Detectors
Complete analysis of data
and Announce Results Spring 1991
Possible Data Analysis Strategies
The data from the NRRS will be primarily analyzed using descriptive statistics.
The radon concentrations will be summarized by arithmetic and geometric means and
standard deviations. These statistics will be determined for the entire nation, single unit
homes (attached and detached), and multi-unit homes. The distribution of these
concentrations will be examined to determine if it is lognormal.
Summer 1988
February 1989
June-August 1989
September-October 1989
December 1989 - January 1990
March-April 1990
June-September 1990
-------
The data will also be analyzed on a regional basis. This analysis will include all
the residences sampled or could be limited to single unit homes. The sample size of
the multi-unit homes is sufficient to provide a national estimation but is too small to
provide regional estimations. The analyses will probably be based on the living level
radon concentrations but could also be presented for basements and second floors.
The analyses will also describe the radon characteristics in single and multi-unit
structures. For single unit structures, the radon characteristics of different types of
crawl spaces (covered, uncovered), basements (storage, workroom, living, sleeping), first
floors (full basement, partial basement, crawl space, slab on grade), and second floors
will be examined. The multi-unit structures will be analyzed according to the number
of units at ground level, below ground, and above ground. An important aspect of this
investigation will be studying the relationship of radon concentrations to substructure
characteristics.
The effect of energy conservation methods, designed to reduce air infiltration
and heat loss, on radon concentrations will be examined. The housing population can
be broken down according to slab on grade, basement, and partial basement/ crawl
space, and the type of energy conservation method utilized.
Other analyses can be made between the radon concentrations and the type of
HVAC system, age of the structure, degree of ground contact, and portion of the year
closed to outside air. The data could also be analyzed according to the 11 regional
geologic provinces, as outlined by the U.S. Geologic Survey, and comparisons made
between the geologic characteristics and the radon concentrations within a given region.
u>s. aaowcw WWMN6 1990 748-010/25002
------- |