Multivariate spatial analysis of lake sediment geochemical data; Melville Peninsula, Nunavut, Canada

- Two till geochemical surveys have been studied for their ability to confirm and predict the underlying bedrock lithologies.
- Min/max autocorrelation factor (MAF) analysis for spacified lag values is applied to logratio transformed geochemical data.
- MAF components are used to classify lithologies by analysis of variance and linear discriminant analysis to create maps of posterior probabilities.
- The predictive mapping process using MAF is an improvement over classification methods that do no employ the spatial relationships of the data.

A multivariate spatial analysis was conducted on a suite of glacial till geochemical data collected over the Melville Peninsula, Nunavut, Canada. Previous studies demonstrated through the application of multivariate statistical techniques that the composition of the lake sediment geochemistry reflects the underlying geology in northern Canada. In this study, the application of minimum/maximum autocorrelation factor analysis (MAF) to glacial till geochemistry has extended the knowledge and description of the underlying geology through the recognition of spatially correlated factors that represent distinct lithologic features and glacial transport processes.Maps of posterior probabilities for the underlying lithologies were estimated based on the MA factors and compared with those from standard principal component analysis. The use of MAF provides a measured improvement in predictive mapping irrespective of the choice of lag separation.