The creation of 3D models of ore zones and geotechnical properties is typically performed by the interpolation of borehole data. Kriging and other interpolation methods have serious sampling issues, including bias from borehole orientation. Interpolation and data projection would be better if guided by high resolution geophysical imaging. We propose, and show by example, to use seismic reflection data to robustly interpolate and project data from logged boreholes acquired over the Tropicana Gold Mine. Firstly, a small number of sonically logged boreholes are used to provide missing sonic velocity data via data prediction. In our example, we use Specific Gravity, Magnetic Susceptibility and core-scanned XRF data with machine learning algorithms to create a 'smooth' 3D model of sonic velocity from the 300+ boreholes that have elemental and petrophysical data, but no sonic logging. Then the inversion of 3D post-stack seismic data builds a robust 3D model of acoustic impedance. Along borehole trajectories, relationships between acoustic impedance and geochemical or petrophysical data are built and/or refined by machine learning algorithms. The relationships are then used to create better 3D models of geochemical and petrophysical properties using localised borehole information in conjunction with the high-resolution 3D acoustic impedance model. Thus, we use geophysical data between and below boreholes to interpolate and extrapolate borehole measurements via the seismic reflection data set.

Open-Access Online Publication: March 03, 2023