Centre for Data-driven Geoscience, School of Earth and Oceans
My work focuses on applying machine learning methods to the geosciences, to provide geoscientists with objective results and provide user-interpretable feedback on those results. Some projects include:
• Developing image analysis methods for automatically detecting planar geological features that appear as sinusoidal lines in borehole televiewer images (though they are rarely exactly sinusoidal due to geological processes).
• Training classifiers to identify subsurface strata from: geophysics logs such as gamma logs since specific strata in the Pilbara region have recognisable gamma “signatures”, though these become stretched or corrupted by noise due to geological processes; modelling geochemistry distributions for strata; and developing a data fusion system incorporating these and other models for identifying the strata intersected by a drillhole for better 3D geological understanding.
• Developing a ML system for objectively logging a geological sample’s composition in terms of mineralogy and handling properties, through analysis of a sample’s laboratory measured reflectance spectrum, geochemistry, photo. The resultant logging is consistent with geologists’ field logging, through analysis of previously logged samples.
