The soil sample results are given in Tables B-8-4 and B-8-5 and plotted in Figures B-8-5.a, B-8-5.b and B-8-6. Several conclusions are noted: A. The activity is highly variable from point to point and as a function of depth. The surface 241 Am activity varied from 2.25 to 14.14 pCi/g. B. Six out of twelve sample locations showed the surface concentrations to be greater than subsurface. The other six showed subsurface activity to be greater. C. The average surface activity (0 to 1.5 em) was 6.98 pCi/g; the average for 0 to 2.5 cm was 7.99 pCi/g; the average for 0 to 3 cm was 9.55 pCi/g, and the average for the IMP reading was 5.44 pCi/g. Additional analysis of the data presented in Table B-8-4 leads to several interesting observations. In terms of accuracy of measurement at different stages of soil sample analysis, one might expect an unbalimilled sample to be least accurate, a ballmilled sample more accurate and counting after chemical separation and isolation to be most accurate of the three stages. In this context, the unbalimilled and ballmilled samples would show high variability around the results by chemistry. Figure B-8-7 shows this to be the case, with 7 of the 12 samples having the results by chemistry at some point between the other two. The magnitude of the differences shown for the Ag sample is unexpected, especially with the ballmilled value so far from the chemistry number. This is further illustrated in Figure B-8-8 where the M1 plot of ballmilled samples shows a definite high side bias due to the one large value from the Ag sample. Deleting the A3 sample produces the plot labelled M2 which reaches stability rather quickly and also indicates the true value of the Ag sample is probably between 15 and 20 rather than 36.6 as reported. Figure B-8-9 is included to show that, in general, with the degree of variability present in these data, six samples are not enough to develop a stabilized mean. VI. Conclusions and Recommendations There appears to be variability in 241 Am activity at any point of measurement (before mixing). Variability has been observed within a given soil sample, as well as within a given area. This means that if soil sample data are to be compared to the IMP data (for a given measurement), a multitude of samples are required. Data in Figure B-8-6 illustrate this problem. Because of the high variability of activity from point to point, this experiment cannot be used to "verify" soil sample to IMP ratios. The IMP "samples" 16 to 20 million grams of surface soil. During this experiment only a few thousand grams were sampled by the soil sample technique. The average surface soil samples read about 40% higher than the IMP readings. However, the average soil sample concentrations (0 to 3 em and 0 to 2.5 em) of 8.33 pCi/g contained a standard deviation of + 3.64. It should be pointed out that the soil samples determine activity in dry soil containing particle sizes less than about 0.5 em in diameter averaged over about the top 2.5 to 3 em. The IMP samples the soil-rock-humus-water matrix in situ to a depth that is variable according to vertical and horizontal distribution of the activity. The IMP conversion factor assumes uniform distribution. Calculations have shown that if the distribution is exponentially decreasing with depth, a soil sampling depth of 0 to 3 em should provide a good comparison with IMP readings (Figure B-8-10). Any other sampling depth would be more dependent on the vertical distribution. It is evident that at half the locations in the experimental area, the activity increases with depth. The area was mostly clear of brush. The soil was coarse sand. It seems reasonable, then, that over a period of 20 years, much of the surface activity has moved down to below the surface. B-8-6