Application of geographical information systems to the interpretation of exploration geochemical data and modelling of gold prospects, South Devon, England.

Geographical Information Systems (GIS) are rapidly developing computer systems for managing spatial data. This study used GIS techniques to analyse and visualize exploration geochemical data, to extract spatial information from Landsat images, gravity and aeromagnetic data and to model gold potential using logistic regression, weights of evidence, Dempster-Shafer and fuzzy logic methods. Regional exploration geochemical data from various data sets were represented using the catchment method. Experimental regression analyses for evaluating the influence of Fe-Mn scavenging effect and lithological variations show that zinc is most affected by these factors. The importance of anomalous catchments and dilution effect for zinc was also evaluated using mineralization rating and productivity methods. Soil samples collected in this study suggest that there is no base metal mineralization associated with gold at Whympston area, but limited anomalous haloes around the Loddiswell Mine. Lineaments were extracted from Landsat images using the objective lineament extraction and enhancement method. A shaded relief technique was used to enhance the variations of topographical, gravity and aeromagnetic data. Gravity and aeromagnetic data were also processed using second vertical derivative and reduction to the magnetic pole. Results show structural features possibly related to faults on the processed images and highlight highly elevated magnetic anomalies over the Kingsbridge area suggesting an association with possible unexposed hornblende schist related to the Start Complex. Both descriptive and conceptual models for gold occurrences were constructed, which were then correlated with results from the inductive and deductive modelling methods used. Logistic regression and weights of evidence methods have relatively good correlations with known gold occurrences and highlight the importance of felsite in the area. However, very low weights are related to binary patterns over the areas associated with unexposed gold potentials. Dempster-Shafer and fuzzy logic methods can overcome the shortcomings for incomplete data in data driven methods. These methods allow the inputting of weights (degree of belief and membership grade) from well documented mineral deposit models without any prior knowledge of gold mineralization in the area. The deficiency of deductive methods is that there are sometimes large discrepancies between the predicted favourable areas and known gold occurrences.