Joshua Rines, PhD

Decision making based on smooth interpolation (e.g., simple kriging) of geophysical and geochemical data, when used in mineral prospectivity mapping yield high false positive rates, thereby ineffective exploration.  Most interpolators result in maps that have less variance than the data itself.  We present a novel interpolation technique, called the multigrid simulation, which is a stochastic interpolation method designed to improve the interpolation of geophysical flightline data.  The multigrid method advances the performance of standard methods (e.g., simple kriging) by offering enhanced resolution, computational efficiency, and robust uncertainty quantification.  Standard methods yield deterministic maps without properly accounting for the spatial uncertainty, and are often plagued by non-physical artifacts introduced by the interpolation.  These maps thus inject substantial bias into the decision making process when it comes to critical mineral exploration (e.g., planning of drill sites).   In our work, we address both of these confounding biasing effects by employing the multigrid method.  The multigrid method treats non-stationarity associated with large-scale datasets separately as an underlying trend, from which local variation is interpolated iteratively at increasing resolution until acquisition resolution is reached. Implemented in the Julia programming language, which takes advantage of the parallel computing efficiencies therein, we demonstrate the utility of multigrid by using it for interpolation of flightline residual magnetic intensity (RMI) data over the Cape Smith Belt in northern Quebec and demonstrate clear bias reduction with respect to decision making.