use the kriging approach discussed here, making specific corrections using the factor 107/10" (v) . The magnitude of this factor could, hopefully, be estimated from the finer grid FIDLER readings. In Tablé 3, Part B, we have recomputed Pu inventory estimates for stratum 6 and the total for strata 3 through 6 after deleting from consideration the two 100- x 100-foot cells adjacent to GZ. Recall that these had kriging estimates of 826 and 832 uCi/m? and that they contained the three highest observed Pu concentrations in stratum 6 (1,170, 1,300, and 14,300 pCi/m?). Both kriging and stratified random sampling estimates of inventory were recomputed. The two inventory estimates for stratum 6 now differ by 2.9 curies rather than by the 10.5 curies when these two cells were not ignored (Part A, Table 3). The stratified random estimate for strata 3 through 6 drops by 9 curies, whereas, the kriging estimate drops by only 1.5 curies. Hence, the inventory estimates obtained by strata mean concentrations are much more sensitive to extreme values than those obtained using kriging. This is partly due to a different weighting scheme and also to the use of logarithms in the kriging procedure. The results in Table 3 suggest that inventory estimates obtained using two stage kriging or arithmetic average Pu concentrations for strata may tend to be similar in magnitude for study sites where extreme trends (drift) or variability in Pu concentrations do not occur over the region for which estimates are required. DISCUSSION The kriging approach used here may be particularly applicable for those regions that lie between the immediate GZ area and the relatively lowlevel areas at distance from GZ. The tremendous variability present close to GZ makes questionable, perhaps, the application of any “smoothing" or weighted average procedure in that region, i.e., concentrations may be simply too chaotic to model. At distances far from GZ where relatively low levels of Pu are evident, the kriging approach has not been adequately studied to allow us definite conclusions as to its applicability. However, we have seen in this report that Pu concentra- tions are not as well predicted from the FIDLER when FIDLER readings drop below 5,000 cpm (Figure 5). We have also noted (Table 2) that along Line 2, which is in an area where FIDLER counts are < 5,000 cpm, there appears to be no correlation structure even for adjacent FIDLER readings, i.e., the variogram is a horizontal line with a- large nugget effect. In other words, along Line 2, we have pure "noise." correlation structure, kriging. In the absence of a there appears to be little advantage to using 401