Modelling 3D Grade Distributions on the Tarkwa Paleoplacer Gold Deposit, Ghana, Africa

Abstract

In the Precambrian Tarkwaian Group of Ghana, gold is preferentially located as paleoplacers within quartz-pebble conglomerates. Gold distributions are intimately associated with sedimentologic and stratigraphic features of the host rocks. In this situation, traditional geostatistical methods have not provided accurate predictions of ore grades and reserves, due to difficulties in properly incorporating geologic information in the geostatistical estimation. Application of Transition Probability/Markov geostatistical techniques allowed us to combine geologic concepts and domain knowledge with indicator and Gaussian-based estimation techniques. Vertical variability relationships within stratigraphic sequences, as measured by borehole data, were used to predict lateral distributions of lithologic facies. The result was a set of 3-D spatial relationships that reflect an integration of geologic concepts and readily observable geologic attributes. This approach provides an alternative to more traditional geostatistical ore deposit modelling. It provides a statistically sound, lithofaces-based prediction of gold grades and uncertainty of the predictions, constrained by geology and the 3D geological framework.