Claire Penczak
Evaluation of Weights-of-Evidence Modelling: A Red Lake, Ontario Case Study © 2010
To define gold exploration targets, geologists delineate areas with multiple geological characteristics (evidence) associated with known gold deposits (training data). Based on Bayesian probability theory, weights-of-evidence modelling refines this delineation process by weighting the absence and presence of evidence on binary evidence layers. The purpose of this study was to validate weights-of-evidence as a method for quantifying the spatial association between training data and evidence and to evaluate its predictive ability in areas without training data.
Step-wise addition of evidence layers demonstrated that evidence strongly associated with training data determined the spatial patterns on the final favourability map. Evidence layer thresholds derived from the Bubi gold deposit area in Zimbabwe were applied to the Red Lake gold deposit area in Ontario, Canada in a weights-of-evidence analysis that predicted 65% of known Red Lake deposits.
Results validated weights-of-evidence as a method to quantify the spatial association between training data and evidence, and demonstrated its potential for application in unexplored areas.