Quantification of Uncertainty in Geochemical Anomalies in Mineral Exploration

Abstract

Classification methods capable of identifying signals or groups of samples, whose geochemical composition is affected by dispersion from mineralisation, are critical in regional and local scale mineral exploration projects. This study compares various population and spatial fractal classification models with several new models to identify populations associated with VMS-style mineralisation in regional till geochemical data from Sweden and both Cyprus-style VMS deposits and anthropogenic contamination in soil data from Cyprus. The new models include concentration-distance from centroids (C-DC), concentration-concentration (C-C), and simulated-based and category-based fractal models applied to representative and simulated samples (CF-R and CF-S). The precision (stability) of the models and spatial uncertainty were tested using Monte Carlo and sequential Gaussian simulations, as well as the effects of pre-processing of the geochemical data. In the Sweden till data, CF-R, spectrum-area (S-A) and the related simulated (SS-A) approach proved more effective in delineating known VMS mineralisation in some regions than single element patterns for mineralisation-related metals such as Cu. In Cyprus, both established and new fractal approaches were marginally more effective at separating areas of known mineralisation (including the major deposits) against a backdrop of generally elevated levels of VMS-related elements in the pillow basalts and underlying sheeted dyke complex. The C-C and C-DC approaches define a contiguous zone whose multivariate patterns are closely linked to either geogenic dispersion or anthropogenic contamination including historical contamination that cuts across current land use zoning. Population or spatial features in geochemical data delineated by different fractal approaches are dependent on the mathematical basis of specific fractal models. Application of a wide range of fractal methods, along with assessment of uncertainty in sample classification and stability of spatial patterns, provides a firmer basis for quantifying the processes and features that control element distributions in regional geochemical data. It also provides criteria for selection of the most effective combination of data pre-processing and fractal modelling to extract desired features or signals in the data.