Evaluating volcanic-hosted massive sulphide favourability using GIS-based spatial data integration models, Snow Lake area, Manitoba.
|Title:||Evaluating volcanic-hosted massive sulphide favourability using GIS-based spatial data integration models, Snow Lake area, Manitoba.|
|Authors:||Wright, Daniel Frederick.|
|Abstract:||Maps showing spatial variation of volcanic-hosted massive sulphide (VHMS) potential in the Snow Lake area have been generated by combining digital data from a variety of sources. An exploration model, based on a VHMS deposit model, but expanded to include regional exploration datasets, provides a conceptual framework for the extraction of predictive evidence from the raw data sources. The evidence is divided into five factors: stratigraphic evidence, evidence related to available heat, alteration evidence, geophysical evidence and geochemical evidence. Stratigraphic evidence is derived from the presence of favourable volcanic lithologies, and the proximity to contacts between volcanic flows and volcanic breccia. Heat evidence is derived from proximity to subvolcanic tonalite sills (the inferred primary heat source) and proximity to synvolcanic dykes. The alteration evidence is deduced from the presence and proximity to mapped alteration, mainly silicification, Fe-Mg metasomatism and pyritization. The geochemical evidence is a combination of a lake sediment signature (derived by principal components analysis), and individual maps of Pb, Zn and Cu in till. The geophysical factor is a combination of magnetic, gravity and VLF components. Four approaches have been applied for combining spatial evidence to predict mineral potential. (1) The weights of evidence method (data-driven) is used to characterize the spatial associations between the known deposits, and to calculate weights for each predictive map layer. The output is a favourability map expressed as the posterior probability of a VHMS deposit occurring per unit area. The magnitude of the weights allow the evidence to be ranked according to predictive capability. (2) A second data-driven method, area weighted logistic regression, was applied using the same binary evidence maps as used in the weights of evidence method. The binary maps, representing the independent variable in the model, were combined to form irregularly shaped polygons with unique combinations of evidence. The observations for each polygon were weighted according to the area of the unique-polygon. The dependent variable was the presence of a deposit. This model was not constrained by the assumption of conditional independence. (3) In the fuzzy logic method, evidence is expressed in terms of fuzzy membership functions, subjectively assigned by an expert, for each predictor map. An inference network is constructed to mimic the decisions made by an exploration geologist, using a variety of fuzzy combination rules that reflect varying degrees of logical AND and OR. (4) Dempster-Shafer belief theory is more flexible than fuzzy logic for representing uncertainty in the data, but is somewhat restricted in the expressiveness of the combination rules. The Dempster-Shafer output consists of favourability maps for "support" (conservative), "plausibility" (optimistic), and "ignorance" (uncertainty). Comparing the results from the four different methods shows that they all have a similar spatial distribution of areas of high favourability. The known deposits are in highly favourable zones, as expected, and a number of prospective areas have the right combination of factors, but no known deposits. The new VHMS discovery at Photo Lake made during summer 1994 is in a favourable zone as predicted by these models.|
|Collection||Thèses, 1910 - 2010 // Theses, 1910 - 2010|