Creating large scale probabilistic boundaries using Gaussian Processes

Abstract

A crucial step in geological modelling for mining is creating accurate geological boundaries from multiple information sources such as geophysical logs and chemical assays. The accuracy of these boundaries directly affects the accuracy of the resulting geological models. This in turn can affect future work such as mine planning. Manually created boundaries are labour intensive, and it can be difficult to reassess results, fix mistakes, or update the boundaries as new information becomes available. An easier method of including production data as it becomes available is important for capturing finer detail in the deposit. This paper presents a method for automatically estimating surfaces in local regions from exploration and production blast hole data using Gaussian Processes. Models in overlapping regions are then fused, producing a global surface estimate that captures boundary variations at a local scale. The results demonstrate that the inclusion of blast data produces locally more informed surface estimates and that blast data can contribute to the production of boundary estimates evaluated to be geologically sound by an expert in the field.