_ Selection of Sample Locations Two methods of selection were utilized to choose sample locations. The primary method was by statistically random selection and was used on all islands where the surface contamination was considered uniform (or with a slow-changing gradient which could be stratified into approximately uniform areas). When a "hot spot" or anomaly was encountered, or a burial site was vucpected, and at nearly. all 5GZ areas, an objective selection method was used to choose the sample locations. The purpose, in this case, was to determine to a rough degree the extent of the contaminated mass and give some value of its soil concentration. than very crude area limits. No attempt was made to define anything more Determination of actual boundaries of such contamination and specific estimates of volume were considered cleanup projects. Random Selection. The sample locations for the random program were -tmdeed chosen by a statistically random process. Each island was divided by a grid into relatively small areas (actually represented on available maps, drawings and photographs). An attempt was made to make the grid spacing as small as 50 feet or to get several thousand squares on a large island. Each of these grid squares or points was numbered. Areas which would be impossible to sample, such as concrete pads, coral reef, runways, paved roads, etc., were excluded during the gridding process, the appropriate grid square not being numbered. The sample locations were chosen for each island according to the design criteria using random number tables. 2 An excess of 10% was chosen to allow *tandbook of Mathematical Functions, USDC, NBS Applied Mathematics, Series 55, USGPO, Washington, 1965, pp. 991-995 (Table 26.11).