Models are tested by eliminating each 241am data value in turn, then using nearby data and the
model being tested to estimate the missing value. The difference between the estimate and the
measured value is called the "kriging error," and can be used to compare different models and check
the statistical assumptions. For example, one assumption is that the kriging errors are normally
distributed, and this was shown to be a valid assumption for both models. Because the model which
ignored windrow effects gave fewer kriging errors on the 241 am activity in excess of 6 pCi/g, it was
chosen for making the initial characterization estimates. The model for the 25 m data in the west
area was also tested to confirm that it would yield acceptable estimates.
In two areas of Janet, both of the 50 m model tests produced more large kriging errors than
anywhere else on the island.
One was a 450 x 250-m rectangle near the center of Janet, and the
other a triangle on the northern edge of the island just west of the north baseline, near the old Test
Grid. The fact that estimates using both models gave poor results in these areas indicated the
activity itself was more variable, so that more measurements would be useful. Therefore both areas
were staked on a 25 m grid and measured with the IMP at the finer spacing from 6-21 January 1978.
The original characterization had resulted in an estimate of 21.25 ha with TRU activity in excess of
40 pCi/g with the additional 25 m data, this estimate dropped to 20.75 ha.
6.4.3 Grid Location Problems
Because benchmark PORKY had not yet been uncovered in the dense vegetation when the surveyors
began staking the 25m grid in the west area, benchmark LEE was used as a reference instead.
Unfortunately, an error was made in the process of setting out the grid from LEE, which was
discovered when the vegetation was cleared from PORKY. The error resulted in the 25 m grid being
shifted 7.32 m (24 feet) west and 4.88m (16 feet) north of the intended location. In order to
minimize further confusion, the area was not restaked at the time, but the 50 m grid with origin at
PORKY was extended far,enough west to assure complete coverage of the island.
The situation remained unchanged until the subsurface excision in December 1978. The excision site
was in the area with the shifted grid, and had been sampled at locations referenced to the shifted
grid. Therefore, the boundaries for the excision were transmitted to the Joint Task Group (JTG) in
terms of the shifted grid. However, JTG was not informed of that fact until later and the first two
lifts were made with the location based on PORKY coordinates. The misunderstanding was
eventually cleared up, the excision completed as intended, and all locations thereafter were
referenced to PORKY, even in the west area. For the Fission Product Data Base sampling, the 50 m
grid was extended to cover all of Janet, so that all FPDB samples were taken at 50 m nodes of the
PORKY grid.
6.4.4 Other Activities
In April 1978, seven additional locations were chosen for soil sampling as part of an investigation of
the variability of TRU in the soil and of an apparent discrepancy between soil and IMP data (see
Tech Notes 22 and 23). To try to estimate the variance of soil TRU activity within an IMP view,
four composites instead of the usual two (see section 4.2.1) were taken at each location. The
samples also provided a check on the ratio of TRU to 24! Am eomputed from the original soil samples.
The chemical analysis results for these samples confirmed that the soil TRU activity within a single
IMP detector field of view is highly variable. Also, the variance of the sample TRU activities
increased in proportion to the average TRU activity in the field of view. However, the ratio of TRU
to 24lAm from these samples was not significantly different from the previously-estimated ratio,
and the ratio variance was independent of TRU activity. The ratio of TRU to 24! Am for Janet was
therefore not changed, and eventually soil sampling reverted to the usual two-composite method.
162