Application of computer-orientated statistical and mathematical techniques to the interpretation of geochemical prospecting data, with particular reference to the Pirejman area, south-east Turkey.

This thesis considers two main topics in the field of applied geochemistry, namely: a) The application of Direct-reading optical spectrometers to the analysis of geochemical samples, and b) The use of computer aided numerical analysis in the interpretation of univariate and multi-variate geochemical data. For the purposes of this exercise, the Pirejman area of SE Turkey was chosen as a study area. This district contained known lead-zinc mineralization, and was also potentially chromite bearing since a large serpentinite body cropped out in the western portion of the area. As a result of this study, a number of chromite and magnetite pods were found. The study of spectrometric analysis draws attention to the importance of the composition of the buffer mixture, ignition temperature and grain size of samples prior to analysis. It is demonstrated that the special background channel used in the ARL 29000B Direct-reading Spectrometer is unsuitable as a base upon which to adopt a correction procedure to eliminate matrix effects. The spectrometer is found to suffer from serious matrix effects, and it is shown that an effective correction procedure must therefore rely on the relationships that exist between variations in the major element contents and the induced interference in trace element readings. A computerized correction and conversion procedure is developed. The limitations, requirements and means of data storage and selective retrieval of geochemical data are investigated. A review of computer-orientated data processing techniques suitable for application to geochemical exploration data is presented. A new approach to the problem of the selection of optimum trend surface is proposed, involving a critical consideration of the components which make the deviations from a given trend surface. The use of Discriminant analysis as a final classification technique for results obtained by both Cluster and Principal Component analysis is discussed. In order to obviate the possibility of misclassifying an unknown sample by Discriminant analysis, it is proposed that a quantity (in standard deviation units), called the "standardised discriminant score distance", is calculated with respect to each of the groups between which the discriminant score of the unknown individual lies. The geology and mineralization of the Pirejman area is described. Primary dispersion patterns adjacent to known mineralization are investigated. Various data processing techniques are applied to regional rock and drainage sediment data obtained from the Pirejman area. Using a number of control features, the resolving capability of these computer-orientated techniques is evaluated. The power of Discriminant analysis as a means of testing a geochemical model, and predicting other areas of interest is demonstrated.