The partitioning of petrophysical data: a review

Petrophysical analyses are undertaken within a framework of ordered geological systems. The recognition of the natural geological ordering through the grouping and partitioning of petrophysical data is known to be a fundamental requirement of successful petrophysical interpretations. The mathematical basis and examples of the use of a variety of techniques of value in partitioning data are reviewed. Beginning with the simple use of curve discrimination on conventional crossplots, the discussion proceeds by covering increasingly complex but relatively well-known methods including discriminant function analysis, cluster analysis and principal component analysis, and goes on to include multi-dimensional scaling, ‘pigeon-holing’, fuzzy clustering, characteristic analysis, projection pursuit and neural networks. The discussion highlights some of the benefits, and some of the pitfalls, of each technique through the included examples. The discussion encourages the use of exploratory data analysis techniques in the processing and interpretation of petrophysical data and provides a reference framework from which can follow further evaluation of the value of the techniques in circumstances particular to the interested reader