Re: Radioactivity in soil

[clement@vaxxine.com (Clement)]

When trying to compare soil radioactivity levels at a potentially
contaminated site to the natural background range there are several
difficult problems to address.  The most difficult to solve generally
revolve around the question "what is background?".

In my opinion, the most important consideration is that the "background"
should be a function of both locality and soil type.  It would be ridiculous
to apply background data collected where the range of natural radionuclide
concentrations are generally quite low to a project in an area where they
are relatively high.  The result would be to classify all the soil as "above
background".

Once a background area has been selected, a reasonable sampling protocol
must be followed to collect sufficient and representative "background"
samples.  Although I'm not familiar with it, I understand from previous
posts that some guidance is given in NUREG-5849 in the form of a statistical
sample adequacy equation.

Then the background data must be analyzed.  Fitting the data to a reasonable
distribution (generally normal or log-normal) is the simplest, and a very
useful, approach.

Finally, sample data must be compared to the background distribution.  At
least two very different approaches can be used here.  The simplest is to
compare each sample result to a measure of the upper end of the background
distribution.  Recently, our office has used the 97.5th percentile, used by
the Ontario Ministry of Environment and Energy, as a measure of the upper
end of the background range.  We've found this measure more useful than
others we've tried in the past (such as mean + 3 std. dev) since it can be
applied to data that may not be normally distributed, and can still be
reliably calculated if a large proportion of the backround results are
reported as less than the lower limit of detection.  The biggest problem
with this approach is that 2.5% of background samples will exceed this measure.

If the samples have been taken to characterize an area, and many samples are
available, another approach is available.  One can compare the distribution
of sample results to the distribution of normal background.  Although this
involves more sophisticated analysis and requries multiple sample
comparison, it avoids the problem of "statistically" failing "clean" material.

One problem sill remains:
>the actual soil concentration for Ra-226 in soil for the 15 mrem/y
>residential scenario according to NUREG-1500 is 0.787 pCi/g

This will often be less than the difference between the background "mean"
and the highest of the background samples collected.

>This means that assuming the background sample represent the true
>variability in background, 1/20 of all uncontaminated areas will
>require cleanup! Ouch.

Not at all.  I'm not familiar with the US regulations, but I don't see why
one should think that any individual sample with a Ra-226 specific activity
greater than 0.787 pCi/g above the mean background must be considered an
absolute indicator of contamination.  It would seem more sensible to
consider a distribution of samples, and test whether or not the distribution
indicates that the samples GENERALLY exceed the limit of 0.787 pCi/g above
the mean background.

Chris Clement
Low-Level Radioactive Waste Management Office
Atomic Energy of Canada Limited
clementc@crl.aecl.ca
clement@vaxxine.com

===================================================
The opinions expressed above are completely my own,
and may not reflect those of my employer.
===================================================

>With regard to detection of 1 pCi/g, based on my field experience, I believe
>that you are correct in you assessment of the delectability of a 1 pCi/g
>change in Ra-226 concentrations using NaI(Tl) detectors.  However, the
>problem with the pending 15 mrem/y requirements is that 1 pCi/g is often in
>the "noise" range for background concentrations of naturally occurring
>radionuclides.  That is, within a given well defined soil type, it is not
>uncommon to see background radionuclide concentrations vary from sample to
>sample by as much a 0.5 to 1+ pCi/g.  This is usually addressed by saying
>"take more samples and characterize the population statistically."
>
>Statistical analyses of a background sample population has a great deal of
>utility, however there are problems with this as well.  Using the
>statistical sample adequacy equation, as given in NUREG-5849, it is possible
>to determine how many samples are required to calculate the mean of a given
>population.  But, even so, there is an inherent variability, both in the
>calculated mean, and in the soil.
>
>Given these sources of significant variability (relative to 1 pCi/g) there
>are many of us in the environmental cleanup realm who are very concerned
>about the proposed 15 mrem standards as they apply to naturally Occurring
>isotopes.
>
>As an example, consider the following site specific data from a uranium
>tailing site:
>
>20 background samples were collected from non-affected areas.
>Sample concentrations ranged from 0.5 to 1.8 pCi/g.
>Sample adequacy analyses determined that 20 sample were statistically
>sufficient to calculate the mean.
>The mean was calculated to be 0.99 pCi/g.
>
>Given a limit of 1 pCi/g above the mean of  0.99 pCi/g we are now trying to
>distinguish a difference in count rate resulting from a difference in
>concentration of only 0.19 pCi/g (1 + 0.99 - 1.8) for at least one
>background location which is statistically representative of 1/20th of the
>sight as a whole.  A 0.19 pCi/g distinction between clean and dirty is
>difficult whether you are using gamma surveys or laboratory analyses.