Multi-variate statistical analysis of downhole logs from the Meandu coalfield

Abstract

Geophysical data often comprises several variables measured over a large number of samples. Several statistical techniques, e.g. cluster analysis and principal components analysis, have been developed for the study of such data sets. As an example, these techniques are applied to the natural gamma, density and neutron?neutron logs of three boreholes in the Meandu coalfield. The cluster analysis demonstrated that there were seven significant groups in the data. These corresponded reasonably to the lithological groups: coals, 'dirty' coals, carbonaceous siltstones, siltstones, fine sandstones, granular sandstones and conglomerates. Principal components analysis then established criteria for distinguishing these groups from the geophysical data.