Geological concept revision during appraisal drilling: a replay of the McLeod VMS deposit, Matagami

Appraisal drilling is designed around a geological concept, yet discovery records show that such concepts are often partly wrong and revised late. We propose a procedure that updates the geological concept during drilling and quantifies the cost of delayed revision. The concept is written as a lattice of geological elements, each present or absent with a prior weight, so that revision becomes Bayesian inference over their combinations. A surprise score, the discrepancy between an observation and the prediction of the current model, locates where the data contradict the concept. The approach is tested by replay on the McLeod Zn–Cu volcanogenic massive sulfide (VMS) deposit, Matagami, Quebec. A reference model is built from the complete drilling record, and the deposit model of 2004 is rebuilt from the information available at that date. Three campaigns of similar cost drill the reference model from the 2004 state: the program drilled in 2005, a program planned on the five exclusive hypotheses of 2004 kept fixed, and a program planned on the lattice with sequential updating. Under the fixed hypotheses, drilling converged on the wrong hypothesis and ended at 0.9–1.0 Mt against a 3.0 Mt reference, with the value of delineation at 0–3 MCAD, close to walking away. The lattice ended at 2.5–3.4 Mt and valued delineation at 35–109 MCAD. The principal benefit of revision lies in a better quantification of economic potential and model uncertainty, more than in savings on drilling cost.