Using self-organising maps to derive lithological boundaries from geophysically-derived data in the Mt. Isa region, Queensland

Abstract

Susceptibility and density volumes, derived from the inversion of airborne magnetic and ground-based gravity measurements for a region of the Mt Isa inlier, Queensland, Australia, were analysed using a self-organizing map (SOM) approach. Three-dimensional sub-surface voxel distributions of susceptibility and density were derived from the inversion of magnetic and gravity data using the University of British Columbia (UBC) codes. These petrophysical volumes are often difficult to interpret because of their nebulous nature, with subtle differences between adjacent volume elements. As the SOM approach uses vector quantization, it is an ideal tool to identify subtle relationships in such volumes of disparate data. The CSIRO data-mining SOM tool (SiroSOM) was used here as it was designed specifically for the analysis of such spatially-located, diverse exploration data. Our SOM analysis of the petrophysical voxels has identified (1) some structural features that are evident on the previously constructed Geological Survey of Queensland (GSQ) model of the area; (2) anomalous voxels that form coherent patterns, which may be related to mineralisation and hence be exploration targets; and (3) explicit domains that relate to lithological packages. Further work is needed to validate the SOM results; however, our analysis has shown the value of the SOM approach for analysis of such data. By using SOM, we have been able to assess petrophysical volumes to extract information related to structure and lithological packages in addition to identifying geophysical targets with potential for mineralisation.