A study of the lake sediment geochemistry of the Melville Peninsula using multivariate methods: Applications for predictive geological mapping

• Lake sediment geochemical data were assessed for predicting underlying bedrock geology.
• The use of logratio transforms on the geochemical data removes the closure problem.
• Spatial statistics describe both the spatial extent and anisotropy of the data.
• Multivariate statistics reveal patterns and features of the underlying geology.
• Study confirms that lake sediment geochemistry can predict underlying geology.

The effectiveness of predictive geological mapping was tested using re-analyzed lake sediment geochemical data from lakes across the Melville Peninsula in Nunavut, Canada. The treatment of lake sediment geochemical data within the compositional framework of logratio analysis and the corresponding use of principal component analysis, analysis of variance, linear discriminant analysis and spatial analysis with ordinary kriging provide an informative and quantitative manner for lithologic mapping. Principal component analysis provides useful information on the multi-element associations based on the contrast of rock types in the area. Supracrustal sedimentary rocks have a multi-element character that is distinctive from rocks derived from granitoid and gneissic rocks. The analysis of variance provides details on which elements are the best discriminators for differentiating between the rock types represented by the lake sediment geochemistry and the spatial analysis provides insight into the direction and spatial continuity of the elements associated with specific map units. Linear discriminant analysis provides a basis for distinguishing between the different map units and provides a method of validating the predictive capability of mapping the underlying map units based on the lake sediment geochemistry. The application of multivariate statistical methods on lake sediment geochemical data provides the basis for establishing an objective approach for discovering and classifying geochemical processes from which existing geological maps can be tested and validated and new geological maps can be made in areas where sufficient geological information is lacking.