Abstract:
The paper describes a principle and algorithm proposed for the cluster analysis of the chemical compositions of rocks from magmatic complexes with the aim of identifying parameters of the spaceless and spatial geochemical structures. The approach is underlain by the following information processing logic. First, geochemical data are processed (regardless of the geological and petrographic characteristics of the samples) by means of hierarchical cluster analysis. The aim of this stage is the natural grouping of samples on the basis of similarities between the values of the chosen geochemical parameters, with neither the number of these groups nor their compositions assumed a priori. This algorithm makes it possible to combine, in the course of n - 1 steps, n samples into a single final group. Then the level of the working variant of the combination is determined considering available textural petrographic and geological data. The results of geochemical data processing by this method are (i) a set of discrete geochemical rock types (clusters) that are convenient for describing a given magmatic body; (ii) the evaluated average compositions and the contents of minor and trace elements in the recognized geochemical rock types (clusters); (iii) systems of correlations between the average concentrations of elements in the geochemical rock types (clusters); and (iv) relations between the spatiotemporal distributions of the geochemical rock types (clusters) in the magmatic body. It is expedient to consider this set of characteristics to be the spaceless [characteristics (i) through (iii)] and spatial [characteristic (iv)] structures of a magmatic complex. Applications of this approach are illustrated by the processing of data on the chemistry of different rock types and their distribution in the reference vertical section of the Kivakka intrusion, northern Karelia, which was examined in detail by E.V. Koptev-Dvornikov et al.