Effect of Estimation Technique on Inventory Estimates
Before turning to our work on estimating Pu concentration contours, it
is important that we mention the considerably different estimates of Pu
inventory that can be obtained depending on whether the "average" Pu
concentration for a stratum is estimated by the arithmetic mean or some
statistic that is less affected by unusually high concentration samples
such as the geometric mean or median.
Gilbert et al. (1977a) illustrate
the problems involved with the Pu data from stratum 6 (near ground zero)
in Area 13.
If the geometric mean is used in place of the arithmetic
mean, the estimate of Pu inventory in surface soil at Area 13 drops from
the 46 curies reported by Gilbert (1977b) to roughly 24 curies. This
difference occurs because the statistical distribution of the observed
Pu concentrations for each stratum is highly skewed (a few concentrations
much higher than the bulk of the data).
correct?
Which estimate is more nearly
The answer to this is not known with assurance.
Some insight may be gained by examining the most recent estimates of Pu
inventory as reported by Delfiner and Gilbert (1978) for strata 3 through
6 of Area 13. These were obtained using kriging techniques making use
of a linear relationship between FIDLER and Pu measurements (each in
logarithmic scale).
The inventory estimate for strata 3 through 6
combined is about 17 curies compared with the 28.6 curies reported by
Gilbert (1977b) for these strata.
Delfiner and Gilbert note, however,
that their estimates are probably biased low due to ignoring changes in
mean concentration levels over distances less than 100 feet when transforming inventory results computed in logarithmic scale back to arithmetic scale. Also, the method used to make this transformation is based
on the assumption that the Pu data are lognormally distributed, which
may not be the case, particularly near ground zero in stratum 6.
These
authors suggest that taking a large number of FIDLER readings on a fine
mesh grid may be a viable option for obtaining more reliable inventory
estimates, at least in relatively high activity strata where the linear
log FIDLER-log Pu relationship appears to be well established.
ESTIMATING SPATIAL DISTRIBUTION OF PLUTONIUM AT SAFETY-SHOT SITES
Initial Ffforts
The first estimates of the spatial distribution of Pu at safety-shot
sites appeared in Gilbert and Eberhardt (1974).
They gave threedimensional plots of FIDLER Am count per minute data collected on 100and 400-foot grids over the Area 13 (Project 57) study site.
Threedimensional plots of actual Pu concentrations over space were also
shown.
Additional information was published by Gilbert et aZ. (1975) in
the form of FIDLER Am activity strata for nine safety-shot sites, and
estimated Pu contours for Area 13 and the GMX site in Area 5.
455