the subsample results are based on samples from a 2- by 2-m area, whereas the site results are based on samples from the 12- by 14-m site and may include the additional variability associated with coming from a larger This indicates additional significance for the area, as noted above. differences noted in the subsample and composite sample results. These results indicate the desirability of compositing samples (depending on program objectives) to decrease their variability and increase their representativeness for a given area (the parameter "s" in Figure 3 and subsequent figures stands for the sample standard deviation). Variability of Sampling Techniques Figure 4 gives bar graphs of the plutonium-239 activity for the 0- to l-cm and 0O- to 5-cm results for all four sites. The plutonium-239 concentrations are given for both the activity per gram and per unit area. The graphic scales are the same for both techniques at each site, but vary between sites. A horizontal bar with a vertical error bar indicates the mean and standard deviation for each sample group. In most cases, the values are uniformly distributed about the mean. A unique case is Site I for the l-cm samples where a single high value strongly influences the mean. Figure 5 is a bar graph of the ratios of the coefficients of variation (C) for the activity per unit area results (nCi/m2) from the 0- to 1-cm and O- to 5-cm sampling techniques. The bar graph shows that C for the l-cm technique ranged from 1.3 to more than twice the C for the 0- to 5-cm technique for three of the four sites. The average C's for the 0- to l-cm and O- to 5-cm techniques were 0.42 and 0.30, respectively (ratio of 1.4). This indicates somewhat more uncertainty than the estimate of plus or minus 25 percent by Krey and Hardy (1970, p. 30) for samples from 0 to 20 cm. Statistical testing for differences between the C's for the 0- to l-cm and 0- to 5-cm results indicated a lack of general significant differences (Appendix 3). But, because of the small number of results (5 samples in each group), the test results have limited meaning. The summaries of results in Figure 4 and especially Figure 5 indicate that the O- to l-cm sampling technique is somewhat more variable than the 0- to 5-cm technique. Although statistical testing of the data did not indicate this trend was statistically significant (Appendix 3), the limited amount of data results in the test giving only weak confirmation that significant differences are not present between the two techniques. Figure 6 is a bar graph presentation of the mass of sample (that passed 10 mesh) per unit area (g/cm?) for both techniques at all four sites. The variations include the heterogenous nature of the sampled soil; i.e., rocks and gravel, in addition to the uncertainties from sampling. The coefficients of variation (C) for the mass per unit area results are generally less than those for the activity per gram or per unit area (Figure 5). This may be interpreted as indicating that the variability 658