kriging over more conventional approaches.
The necessity of careful
planning of field sampling studies in order for kriging (or any statis-
tical technique) to give reasonable estimates of inventory and spatial
distribution of plutonium and other radionuclides is noted.
INTRODUCTION
The Nevada Applied Ecology Group (NAEG) has been conducting environmental
transuranic studies at safety-shot sites on the Nevada Test Site (NTS)
and the Tonopah Test Range (TTR) since 1971. One important objective of
these studies is to estimate the total amount (inventory) and spatial
distribution of 239°240py and 24lam (henceforth denoted by Pu and Am) in
surface soil. Of the 10 safety-shot sites studied by the NAEG, the most
intensive soil sampling program took place in Area 13 at the Project 57
site. The first estimates of Pu inventory in surface soil (0-5 cm) were
given by Gilbert et az. (1975), with corrections to these estimates being
published by Gilbert (1977). These inventory estimates were obtained
using stratified random sampling, i.e., by collecting soil samples at
random locations within Am activity strata (Figure 1) that had been
defined on the basis of FIDLER* surveys taken on 400-foot and 100-foot
grids (Figure 2) about GZ.
The inventory estimate for a given stratum
was obtained by multiplying the average Pu concentration (yCi/m*) for
that stratum by the size (m2) of the stratum.
Information on the spatial distribution of Am and Pu was available from
the FIDLER Am activity strata map for Area 13 (Figure 1) since Pu and Am
are correlated at this study site.
The spatial distribution of Pu was
also studied by estimating isopleth (contour) lines of concentration on
the basis of Pu soil concentrations using a computer program "SURFACE IL"
(see Figures 16, 20, and 21 in Gilbert et aZ., 1975; also see Figures 8
through 14 in Gilbert et al., 1977, for the Area 5 (GMX) site). These
estimated contours were unsatisfactory in several respects.
They were
found, for example, to be biased in the sense that estimated concentrations near GZ appeared to be too low, and those at distance from GZ were
too high. Hence, we began to look for alternative methods. An iterative
estimation approach on both untransformed and log-transformed data was
investigated by Gilbert (1978).
The iterative procedure appeared to
reduce the bias in estimates of plutonium concentration contours mentioned
above when calculations were done in logarithmic scale with transformation back to arithmetic scale as the last step.
This work also pointed
out that highly spurious estimates can result in regions of sparse data
where Pu concentrations change rapidly within short distances.
This
iterative approach grew out of suggestions offered by Professor John
Tukey at the first ERDA Statistical Symposium (Eberhardt and Gilbert,
1976).
*Field Instrument for the Detection of Low Energy Radiation.
363