Mr. Tom McCraw
September 22, 1976
Page 3.
The variability in Pu concentrations would probably be greater for high
In-Situ readings than for low readings (as indicated in the diagram).
If Yo is the level of Pu in soil signifying clean-up, this corresponds
to an average In-Situ reading Xj.
But the data in the diagram indicates
individual Pu readings considerably greater than Y, for In-Situ reading Xj}.
Hence, if the In-Situ device is used to meet the clean-up criteria in
terms of Pu concentrations, the level of In~Situ indicating clean-up should
be less than Xj}.
One candidate is the value of the In-Situ measurement
{Xg in the diagram) such that the upper (one-sided) 95% confidence on average
Pu concentration is Yop. An alternative approach would be that Tevel of In-Situ
reading such that some large percent (P) of the Pu concentrations associated
with that In-Situ level are less than Yp with probability l-a. The main
point here is that if clean-up is to be based on In-Situ measurements, the
Tevel of In-Situ measurement indicating clean-up should probably be lower
than indicated by the average linear relationship.
In the remainder of this letter I have addressed the five questions you
handed out at the meeting with Roger Ray, Paul Dunaway, and others in
Joe Deal's office on July 29, 1976. Hopefully, this discussion will help
clarify same of the different kinds of statistical probability statements
that can be made based on sample results. I direct your attention particuTarly to the discussion of "acceptance sampling" for Question 3. This
seems to be a much more satisfactory approach than using average Pu concentrations for deciding whether an island needs to be cleaned up. There are
a good many details that would need to be worked out for actual field application in connection with kriging, but these need to be explored with someone Jike Dr. Delfiner. A table of sample sizes required to meet various
probability criteria is included in the section dealing with Question 3 for
the simplest (nonsequential) sampling design. The number of samples would
probably be less for a sequential design.
Question 1: Over what area or areas should Pu-in-soil measurements be
averaged:
a.
b.
In-Situ measurements?
Soil sampling?
The answer to this question depends’ in part on the variability present from
sample to sample, the spacing of samples, whether any trends are present and
perhaps most importantly on how the health standards (cleanup criteria) are
formulated. If there are no trends and the variability between samples is
relatively small, then the area over which samples are averaged can be large.
However, if strong trends are present (such as near GZ for example), it
would be important to define these fairly precisely. In that case rather
few if any areas might be averaged. Presumably In-Situ measurements would
need less (if any) averaging than plutonium concentrations in soil samples
Since each such In-Situ measurement is itself an average of the Americium
activity in the area scanned by the detector.