Measuring the performance of mineral-potential maps

D. P. Harris and others have proposed a new method for comparative analysis of favorability mappings. In their approach, Weights-of-Evidence (WofE) consistently shows poorer results than other more flexible methods. Information loss because of discretization would be a second drawback of WofE. In this paper, we discuss that the random cell selection method proposed by Harris and others necessarily results in higher success ratios for more flexible methods but this does not necessarily indicate that these methods provide better mineral-potential maps. For example, a good point density contouring method that does not use any geoscience background information also would score high in the random cell selection approach. Additionally, we show that discretization usually is advantageous because it prevents occurrences of overly high posterior probabilities. For more detailed comparison, we have conducted a number of experiments on 90 gold deposits in the Gowganda Area of the Canadian Shield comparing WofE with the more flexible weighted logistic regression method. Mineral occurrences should be modeled as discoveries at points instead of randomly sampling them together with their surrounding environments in small cells.