184 TRANS URANIC ELEMENTS IN THE ENVIRONMENT 



be two dimensional (area sampling), a logical approach is grid sampling. If we then use 

 stratification, we will, in most instances, want to vary the mesh of the grid between 

 strata. This can be done easily if mesh sizes change geometrically; i.e., for areas of the 

 highest concentration, use a grid size of one unit, for the next lowest concentration 

 stratum, use a grid twice as large, for the next four times as large, and so on down to the 

 lowest concentration stratum, which will have the coarsest grid mesh. With such a scheme 

 the problem of matching grid meshes at stratum boundaries is fairly simple. Some of the 

 flexibility of stratified sampling, however, is thus given up, and this probably reduces its 

 efficiency in situations where concentration gradients are known to change rapidly. 

 Hence a variety of field trials will be needed to work out details of this and other 

 problems. 



Many other problems might be considered here, but we conclude by nofing one that 

 may go largely unnoticed, i.e., the practice of sieving soil samples and doing chemical 

 analyses on only certain sieve fractions. Such a practice can introduce important biases, 

 some of which are described by Gilbert et al. (1976a). Examples using plutonium data 

 from safety tests are given in the report by Gilbert and Eberhardt (1976b, pp. 131-153). 



Conclusions 



We have attempted to briefly mention some of the many facets of statistical and sampling 

 methodology which are relevant to studies of the transuranic elements. We believe that 

 too many investigators working with these elements are not sufficiently aware of the very 

 large component of "chance'' error inherent in their data. Much of our experience has 

 been with data on plutonium in soils, mostly that resulting from "safety shots" at the 

 Nevada Test Site, but we have also studied data from a number of other sites (Enewetak, 

 Los Alamos, Rocky Flats, and other locations). In other circumstances variability may be 

 much reduced. However, if it is not, then it is likely that many "significant" findings to 

 date are largely artifacts resulting from inadequate sampling. 



References 



Agterberg, F. P., 1974, Geomathematics, American Elsevier Publishing Co., New York. 



Aitchison, J., and J. A. C. Brown, 1966, The Lognormal Distribution, Cambridge University Press, 



London. 

 Atkinson, A. C, and W. G. Hunter, 1968, The Design of Experiments for Parameter Estimation, 



Techno metrics. 10(2): 271-289. 

 Barnes, M., R. O. Gilbert, and P. Delfiner, 1977, A New Statistical Tool, Kriging, and Some 



Applications to Area 13 Data, in Transuranic s in Desert Ecosystems, M. G. White, P. B. Dunaway, 



and D. L. Wireman (Eds.J, USDOE Report NVO-181, pp. 431-445, Nevada Operations Office, 



NTIS. 

 Box, M. J., 1968, The Occurrence of Replications in Optimal Designs of Experiments to Estimate 



Parameters in Nonlinear Models, J. R. Stat. Sac. Ser B, 30: 290-302. 



, 1970, Some Experiences with a Nonlinear Design Criterion, Technometrics, 12(3): 569-589. 



, 1971, An Experimental Design Criterion for Precise Estimation of a Subset of the Parameters in a 



Nonlinear Model, Biometrika, 5 8: 149-15 3. 



, and'H. L. Lucas, 1959, Design of Experiments in Nonlinear Situations, Biometrika. 46: 77-90. 



Cochran, W. G., 1963, Sampling Techniques, 2nd ed., John Wiley & Sons, Inc., New York. 



, 1973, Experiments for Nonlinear Functions, / Am. Stat. Assoc. 68(344): 771-781. 



Conover, W. J., \91 \, Practical Non-Parametric Statistics. John Wiley & Sons, Inc., New York. 

 Delfiner, P., and J. P. Delhomme, 1975, Optimum Interpolation by Kriging, in Display and Analysis of 



Spatial Data. pp. 96-114, J. D. Davis and M. J. McCullagh (Eds.), John Wiley & Sons, Inc., New 



York. 



