Sampling for cleanup would most likely be done in a sequencial manner in order to keep costs to a minimum. The basic idea would be to take a limited number of samples and make a decision whether to clean up or to withhold judgment : until more samples are collected. The general approach as it igs ap lied Li industry is discussed by Burr (1976, Chapter 12). The use of acceptance . sampling using either attributes or average concentrations requires that samples be collected at random within relativ ely homogeneous areas The Procedure outlined above should thus be applied separately within each such area. The assumption of random sampling within areas is important in order to preserve the chosen a and 6 risks. Consequently, the use of alternative sampling plans such as sampling on a systema tic grid would need to be carefull evaluated before being used in a field study. _ Once the initial cleanup operation has been completed, additional samples will be required to assure that cleanup to the level desired has been accomp lished The design of this phase of the sampling program can also be based on acceptance sampling, although the details may change from those described above. Tf a final demonstration (certification) is require d and the attribute approach is adopted, then Table 12 in Burstein (1971) can be used to determine the number of samples required to be 100(1-8)% sure that the true Proportion of samples with concentrations greater than the cleanup level, L, is less than P "This approach fs based on the assumption that the cleanup operation has ben success~ a 50 that an samples will have concent rations less than L. Sampling to ans tha eous areas ene have been met nsure homogen should also be done independently The design of a cleanup sampling plan may involve a combination of in situ measurements and Pu concentrations in soit samples. If so, these two measurement systems must be well calibrated so that their results can be related The double sampling approach investigated by Gilbere and Eberhardt (1976) at NTS safety-shot sites offers an approach to estimate the optimum allocation of effort between the in situ measurements and soil samples. ° SAMPLING DESIGN GUIDELINES Based on our experfence in helping design some of the NAEG soil and vegetation studies at safety-shot sites, we would like to suggest several guidelines relative to the design and statistical analysis of environmental radionuclide studies, These guidelines are directed towards estimating the total amount and spatial distribution of transuranics, but the design principles are als applicable to ecosystem-type studies, ° Guideline 1: Break Study Area Into Homogeneous Subareas On the basis of field surveys, information from similar studies, topography of the land, vegetation patterns, or any other relevant information attempt to divide the total study area into subregio ns (commonly termed "blocks" or strata") such that the variability in radionuc lide concent rations within subregions is less than that expected for the entire study site. If this stratification is successful, the end result should be a more precise estimate of inventory for the whole site. In addition, of course, estimates of inventory will be obtained for each stratum separately, which are useful for relating trends in soil concentrations with concentrations in other ecosystem components, The success achieved in increasing the precision of the inventory estimate for the entire study site depends on how effective the stratification has been in dividing the area into relatively homogeneous blocks or strata (Gilbert et al., 1975). An inappropriate choice of strata will yield little if any benefit in increased precision. It has been our experience that it may not be possible to construct a satisfactory statification plan without conducting field surveys and/or collecting samples prior to the main study. This leads us to the next guideline. Guideline 2: Consider Pilot Studies If relevant information on the variability and spatial pattern of the radionuclide is not available, a pilot study may yield information that will aid in the design of the main study. It will, hopefully, give indications of trends and levels of variability necessary for defining strata. The kind of information obtained in the pilot study should be appropriate for estimating the number of samples needed in each of the strata to achieve the maximum possible increase in precision of the estimate of inventory or to achieve a specified precision needed to meet the objectives of the study. For example, FIDLER survey readings for 7"!Am would not in themselves be adequate for estimating the variability of 24!am concentrations in 10-gram aliquots of surface soil. A FIDLER survey in conjunction with soil samples collected at a number of FIDLER locations would, in general, be a better approach, Then it might be possible to relate the variability and level of FIDLER readings to those for soil sample concentrations in order to approximate the optimum number of samples for each strata. Of course, the total funds available for the study put an upper limit on the total number of samples that can be collected and analyzed, The number of samples allocated to the various strata could also involve costs of collecting samples if this cost varies substantially between Strata. Gilbert et al. (1976a) discuss how the allocation was accomplished for the Area 13 (Project 57) safety-test site on NTS. Guideline 3: Use Random Sampling Within Strata or Random Sampling in Conjunction With Systematic Sampling Careful consideration needs to be given to the method used to determine the actual sampling locations. Statisticians tend on theoretical grounds to favor random over systematic sampLing. The theory of statistics that permits the formulation of probability statements and inferences about the universe from which sample data are taken 1s based on the notion of drawing samples at tandom. However, the greater ease with which systematic samples on a grid, say, can be collected under field conditions, and the need in many cases to insure that all portions of the study site are sampled are reasons for considering systematic sampling. 585