close proximity are not completely independent. The degree of dependence with distance between samples is required information for kriging. The design and results of this study are discussed by Doctor and Gilbert (1978). Our capability for using kriging techniques was greatly enhanced in 1977 by arranging to access the BLUEPACK kriging program available on the Nevada Operations Office computer in Las Vegas via a high-speed computer terminal operated by Boeing Computer Services in Richland, Washington. This will allow us to handle efficiently the rather large amounts of data commonly required for kriging. Thus far, our major use of BLUEPACK has been by Dr. Delfiner in the analysis of the Area 13 (Project 57) data (Delfiner and Gilbert, 1978). ESTIMATING PLUTONIUM INVENTORY AND SPATIAL DISTRIBUTION AT NUCLEAR SITES Design Aspects We have used the experience gained at safety-shot sites to try new sampling designs at nuclear test sites for estimating Pu inventory and spatial distribution. Perhaps a fundamental change in our approach is to design field sampling plans for the primary purpose of estimating spatial distribution as opposed to inventory. If a field sampling plan allows for the efficient estimation of spatial distribution, then inventory estimates can also be estimated from the data. However, the reverse is not necessarily true. That is, a sampling plan for inventory may not be efficient for estimating spatial distribution. The basic field design approach is set out in sampling protocols developed during 1977 by NAEG scientists for nuclear site studies. These are discussed by Essington (1978). A basic design change from safety-shot sites is to no longer insist that sample locations be chosen at random. Instead, field instrument readings (beta + gamma, alpha, and FIDLER) as well as soil and vegetation samples are collected on systematic grids over the study site. Sampling is done in two phases. During Phase l, the grid spacing is rather wide (usually 400 feet). Data at these locations are augmented by instrument readings every 20 feet taken along eight radials at 45° intervals commencing at ground zero and continuing out as far as 2,000 feet. This grid and transect information is used to design Phase 2 sampling where soil, vegetation, and/or instrument measurements suitable to the particular features of the study site in question are obtained. Phase 2 designs are expected to use grids of different mesh sizes, the finer grid spacing being used near ground zero areas or in other regions where concentration levels appear to change rapidly on the basis of information gathered during Phase 1. The grid data should be suitable for estimating both Pu spatial distribution and inventory 457