Unsupervised clustering of continental-scale geophysical and geochemical data using Self-Organising Maps

Self-Organising Maps is a data-driven approach for exploring and analysing disparate, high-dimensional data. In this experiment Self-Organising Maps is used to cluster remotely sensed geophysical and geochemical data covering the Australian continent into geologically meaningful groups. Our analysis of SOM derived clusters indicates the Australian continent can be represented by five generalised geochronological domains. These geochronological domains contain a number of lithologies symbolising bedrock and regolith units with distinct characteristics.