SUBSAMPLING we Related to the problems of sampiing in the field are the problems of subsampling in the laboratory. The sample presented to the analyst must be of an acceptable configuration, which implies subsampling of a large mass--often kilograms of soil. Subsampling and sample preparation are interdependent to the extent that they must be discussed as an entity. Cline (1944) discussed subsampling of soil masses and stated that the mixture must be reduced in size and subsampled in such a way that a single analysis will produce an unbiased estimate of the mean of the entire sample, He stated further that the accuracy with which a subsample can be drawn depends on its size and degree of heterogeneity. Harley (1972) discussed sample preparation and subsampling of soi] masses for the determination of plutonium in global fallout. The method employed crushing, grinding, and sieving to obtain a mass acceptable for subsampling. However, as pointed out earlier, global fallout is relatively homegeneous in particle size and distribution and is not subject to the vagaries encountered in the accidental or operational releases. The preparation and subsampling techniques outlined by Harley may not be adequate for samples representing accidental ot operational releases. The radionuclide distribution may be heterogeneous and contain particles of a wide range of sizes. The particles are difficult to reduce to a size which can be uniformly distributed throughout the sample. As with sampling, much work remains to be done on subsampling. It is probable that multiple subsampling and multiple analyses are the only techniques available at the present time to adequately define the radionuclide content in the sample. DISCUSSION The importance of careful sampling techniques appears to have been given an unjustiffably iow priority in the reviewed literature. Much of the literature on sampling related to sampling for fission products with only a few references directed toward transuranics and uranium. Evaluation of individual sampling methods or the comparison of methods for sampling of soils for transuranics could not be accomplished since detailed procedures and results often were not given. In many cases, the only sampling information given was: “samples were obtained" or "samples were taken." All too often, sampling is conducted with insufficient attention to the means by which the objectives are to be met. Proper specification of (1) the intrinsic variability of the medium to be sampled; (2) thé number of replicates necessary to provide defensible results; (3) the choice of procedure needed to minimize cross-contamination; and (4) the cost of sampling, sample preparation, and analysis must be defined by the mission in terms of the objective to be fulfilled. 32 In some cases, data obtained for one mission--more often for an objective with an undefined mission--have been used by other workers in the field to support conclusions related to an entirely different mission. Such a practice may lead to erroneous or unsupportable conclusions. Sampling of soils with a coring tool has been quite popular. <A problem associated with almost all coring devices is that of smearing, which introduces a variable not easily controlled, particularly when sampling large differences in radicnuclide concentration within the soil profile. The coring tool may be used for determination of total radioactivity content from the surface to some predefined total depth. An example of the smearing problem (Table 1) was reported by Fowler e¢ al. (1968). Soil cores were collected at the Palomares, Spain, site from an untilled hillside plot in the pathway of local fallout, about 200 m from the point of impact of a nuclear device. The area had received about 5 cm rainfall per year. Five months had elapsed between the event and sampling. The 3.7 fold increaSe from the 5.0- to 15-cm depth to the 15- to 25-cm depth is explained as contamination of the core, probably by particulate plutonium dragged from the surface into deeper soil fractions by the coring tool. Transuranic concentrations in soils are generally low, but their distribution in space may have a range of several orders of magnitude and cross-contamination is a problem which must be addressed. Accidentally- or operationally-released radionuclides may be distributed heterogeneously, not only on a areal basis, but also as a wide range of particle diameters and types. Those particular types of heterogeniety have been observed at several locations. Sampling of the heterogeneous distributions of particles requires special care that all particle sizes are adequately represented in the samples. An example of a heterogeneous distribution of highly radioactive particles ts shown in Fig. 1. The autoradiograph was prepared from soil obtained from an area near the point of impact of one of the nuclear devices at Palomares, Spain. The wide range of particle diameters and agsoctated radioactivities is suggested. If one assumes a single spherical particle of PuQg 0.6 um-diam, the resultant radioactivity amounts to about 0.2 dis/min/particle; an 18 ym-diam particle will produce about 4700 dis/min/particle. In the case of a sample from an area receiving accidentally- or operationallyreleased radionuclides, the heterogeneous distribution of radioactive particles Fowler et of varied sizes poses serious problems associated with subsampling. al. (1968) discussed subsampling and the variability in observed radioactivity Seven one-gram due to a heterogeneous radioactive particle distribution. aliquots were taken from a single sample; the radioactivity of dry soil corrected There is no for background ranged from zero to 778 dis/min/p (Table 2). technique that will effectively reduce Pu02 particles of the typical sizes ebserved by Fowler to a population of subparticles that can be uniformly Faced with such a dilemma, the researcher distributed throughout the sample mass. In the NAEG program, ball-milling, sieving, turns to replicate subsampling. Analytical data from different and replicate analyses have been employed. laboratories on subsamples centaining radioactive particles often vary by factors of 2-3. 33