Mathematics of Mine Sampling VII: The Contradictory Spherical Semivariogram Model

This note calls attention to a contradiction in current eatistical theory and to corresponding difficulties in practice. These are significant concern in evaluating almost all ore deposits. The theoretical contradiction is that the commonly used spherical model for the semivariogram is not consistent with the usual definition of a "regionalized variable" as a mathematical function, the direct study of which "is excluded because its spatial variability is usually extremely erratic, with all kinds of discontinuities. . ." (Journel and Huijbregts, 1978). Specifically, this means a grade function that is too "rough" for interpolation to be useful. The practical difficulty is that the spherical semivariogram interprets the data upon which it is based in a manner that is, more often than not, demonstrably inconsistent with reality. This is attributable to the widely disregarded interaction between the sample pattern and the grade function. Each sample pattern "blurs" the grade function in its own particular way. It is as though the grade structure of the deposit were seen through a lens of poor resolution or reflected in a warped mirror, such as those found in amusement parks. The fitted spherical model embeds this sometimes dangerously blurred and distorted image in the Krige regression. A full analysis of these inconsistencies is to be presented later but, in the meantime, it seems urgent to communicate their effects so that those engaged in estimating ore reserves may be on guard. The spherical semivariogram has a sill and can be replaced by the corresponding covariogram. The equation of the spherical covariogram in normalized form, that is, variance and range both equal to one, is represented as: [ ]