Gilbert et al. (1975) and Gilbert (1976) found that the use of random sampling within strata tended to leave gaps in the pattern of sample location points which contributed to biases observed in estimated Pu contours obtained using computer algorithms. This problem might have been reduced had a systematic (grid) sampling plan been used within each stratum to insure more uniform coverage. The grid spacing could have varied between strata in accordance with the allocation of samples obtained using the ideas discussed above under Guideline 2. An alternative sampling plan for these safety-test sites would make use of both systematic and random sampling. One approach would be to grid off each stratum (different grid sizes being allowed for the various strata) and to choose two or more samples at random within each grid square. The grid size could be chosen to insure that the maximum possible distance between samples is acceptably short, while the number of samples within each grid square could be chosen on the basis of the variability expected between samples. This ‘approach attempts to use the best features of both systematic and random sampling. eribute or measurements are ne oe eee Acceptance sampling by attributes eee en u 1 an area is necessary. of p cleanu r whethe ng decidi for basis o f the proportion of samp roach, the decision is made on the is needed, level. No assumption of normality trations exceeding some present using the measurewhen than r greate be may ed but the number of samples requir ma ke a Ment a PP roach. The measurement ac h us es appro P The design of a transuranic field study must take into account site specific characteristics of the contamination such as deposition patterns, and the quantity, kind, and properties of the transuranics present. However, general design guidelines can be formulated to aid in the design of these studies. Principal planning steps include (1} a clear statement of objectives, (2) appropriate definition of that part of the environment about which information 76-C-06-1830. is desired (the "target population") and that part from which samples are (6) design or selection procedure used to select sampling units for analysis. The design of studies at sites contaminated by local sources of contamination may include surveys using ground or aerial in situ radiation detection devices. These can help define the spatial pattern of contamination and, in some situations, be used in conjunction with wet chemistry analysis of soil samples to estimate total] amounts or average concentrations using double sampling. Stratified random sampling has been shown to be an efficient design for totals Or averages at safety-shot sites. Selecting soil samples from undisturbed locations is an acceptable procedure when estimating total accumulated deposition from worldwide fallout, or from local sources such as stack emissions where the size of Pu particles is in the micron range. However, undisturbed sites are, in general, not appropriate at safety-shot or nuclear detonation sites where patterns of deposition may be complex and/or when interest centers on determining the present deposition pattern. 586 to ACKNOWLEDGMENTS actually taken and to which inferences apply (the “sampled population"), (3) determining the kinds of data and the degree of precision required to meet study objectives, (4) appropriate choice of measurement instrumentation and laboratory procedures, (5) division of the sampled population into sampling and s i c oncen tr ation and offersns or for reading the manuscript We would like to thank Pam Doct the veyeder Biomed!by d funde seat paper This e suggestions for improvements. y ce an fons 0 Croup, USERDA, Nevada Operat Ecology ngton, D.C., under Contract EY Washi A, USERD rch, Resea tal cal ord environmen SUMMARY units, a av erage tion of norma decision and depends on the assump s: aid. the most promise for cleanup studie tial sampling plans seem to offer tran to the design of ve relati ted sugges are nes neral guideli ey oeing area into homogeneous subareas (tret arudies!. (1) divide the study ia, t dies if necessary to obtain data stu pilot t conduc (2) le, possib , Sy eheetives with in each stratum use random : the main sampling effort, and (3) atic samp ling unless study obje some combination of random and system riate. “undigturbed" locations are approp specifically require that only 587