Optimal composite sample size selection, applications in geochemistry and remote sensing

Nearly all the data in exploration geochemistry and remote sensing represent composites. However, composites may arise implicitly or be created explicitly. Bearing in mind that a common exploration task is the classification of data as being above or below some predetermined threshold the size of the composite may be critical to the recognition of a relatively rare, subcomposite, anomalous event. Two approaches are developed, based on statistical, and cost-analytical considerations. The statistical model allows for spatial correlation in the data, of importance when sampling is undertaken continuously along a drill core or flight line. Tables are presented for optimal composite sample size selection based on both models. The procedure is illustrated by an example taken from a drilling program. In general, the cost-analytical model leads to smaller composites than the statistical model. When spatial independence may be assumed the cost-optimal composite sizes are almost always smaller than those suggested by the statistical approach.