Se
disturbance in the access lanes. No precise estimate of this is available, but the experiment
described in Tech Note 4,0 indicated that it is on the order of 2 percent.
A possible source of error that was not included is a bias in the estimates of brush cover, which were
subjective. There appeared to be good agreement between the two regular IMP operators, but there
may have been differences in judgment for substitute operators. For example, the two brush
distributions for Belle shown in Figure B-20-11 and B-20-12 are quite different. Figure B-20-11
comes from the initial survey by an experienced operator, and Figure B-20-12 from a later survey by
a substitute operator. As shown by Figure B-20-13, the later brush estimates are consistently
lower. No brush removal occurred between the surveys, and seasonal variations would result in more
cover during the later survey, not less, so the difference is not due to a real change in brush cover.
However, at a maximum, the computed TRU value is only 6 percent higher for the original brush
estimate than for the later estimate. No other information on the presence or extent of this
possible bias is available.
Table B-20-1 shows the range of values for the sources mentioned above for which a standard
deviation can be estimated. There are also other possible errors which cannot be estimated. For
example, during the fall of 1977, the soil sampling procedure was being done incorrectly for some
unknown length of time. Because the TRU/Am ratio remains fairly constant on an island, the
mistake was assumed not to have affected the data adversely, but there is no way to check this
assumption. There were also a number of equipment problems such as changes in detector
efficiency or resolution and analyzer malfunction. Many of these were detected and corrected, but
others may have been overlooked. Similarly, human errors crept in, for instance on sample labels,
sample weights and results transcriptions.
All of these that were found have been corrected, but
Some may have been missed. The data were checked several times to minimize these "man and
machine" errors, but it is unlikely that they were eliminated totally. Overall, however, the
propagated error value represents a reasonably good assessment of the TRU measurement variance,
Since all of the significant contributors to that variance are included.
TABLE B-20-1: RANGES OF STANDARD DEVIATION ESTIMATES
Source
Ranges
Location
of Values
Propagated error on TRU, pCi/g
Janet
0.6
-
51.6
Propagated error on TRU, percent
Janet
27
-
398
Counting error-IMP 241 4m, pCi/g
Janet
0.1
-
4.6
Standard deviation of TRU/Am ratio*
Northern Islands
0.12
-
2.72
Computed brush correction factor
Pearl
1.05
-
1.42
IMP 241 Am-reproducibility study,
pCi/g
Pearl
7.6
-
9.0
Standard deviation of detector
Lojwa
0.07
-
90.58
Standard deviations of kriging
Northern Islands
0.6
-
Counting error-lab gamma data, pCi/g
Janet
0.17
-
1.66
Counting error-lab alpha
Spectroscopy data, pCi/g
Janet
0.19
-
6.39
effective area measurements, cm2
error, pCi/g
*Due to a programming error the standard deviations reported here are overestimated.
B-20-4
16.2