define six Am activity strata (Figure 1) within which the locations of the soil samples were randomized. The idea was that a greater density of samples was needed in areas of high concentrations than in low-level areas since variability increases with mean concentration. The objective was to minimize the variance of the estimated total amount of Pu for all strata combined. In this paper, we make a more complete use of FIDLER data on the following grounds: 1. As shown by Church et al. (1975), there is a good overall correlation, in log scale, between wet chemistry Pu analyses and Am FIDLER measurements at 1 foot height. 2. The cost of a FIDLER reading is roughly 50 times less than that of a Pu analysis on a soil sample. The way in which we propose to use the FIDLER data in conjunction with the Pu analysis is borrowed from meteorologists (cf. Cressman, 1959; Chauvet et al., 1976). In order to reconstruct a given field, say the constant presssure surface height (geopotential), meteorologists first construct an initial "guess field," obtained by feeding a numerical weather forecast model with yesterday's data. Then this guess field is updated to make it consistent with today's observations. Here we will use the FIDLER readings to construct our initial guess of Pu and then use the actual Pu data to make local corrections. Clearly the proviso is that the FIDLER data carry enough information about Pu concentrations to devise a sensible initial guess. Our first task is thus to study the FIDLER-Pu relationship. STUDY OF THE FIDLER-PLUTONIUM RELATIONSHIP Since the data range over several orders of magnitude, transform them into the logarithmic scale. stabilizing the variance. it is natural to This also has the effect of In this paper, all logarithms are in base 10. Gilbert et al. (1975) studied the correlation between untransformed Pu and FIDLER data at Area 13 and other sites where the Pu to Am ratio computed on the basis of soil samples was nearly constant. Pu-FIDLER correlations ranged from near zero in low-activity strata to about 0.95 near some GZs. Gilbert and Eberhardt (1976) concluded that except for low-activity strata, a double sampling approach (in the sense of Cochran, 1963, Chapter 12) using the Pu-FIDLER correlation (untransformed data) was feasible on grounds of reduced cost and increased precision. The correlation between Pu and FIDLER in logarithmic scale for Area 13 was found by Church et al. (1975) to be 0.92 for data from all activity 371