A preliminary application of weights of evidence in gold exploration in Xiong-er Mountain Region, He-Nan Province

Weights of evidence modeling for combining indicator patterns in mineral resource evaluation is based on an application of Bayes' rule. Two weights are defined for each indicator pattern and Bayes' rule is applied repeatedly to combine indicator patterns. If all patterns are conditionally independent with respect to deposits, the logit of the posterior probability can be calculated as the sum of the logit of the prior probability plus the weights of the overlay patterns. The information to be integrated for gold exploration in Xiong-er Mountain Region comes from a geological map, an interpreted map of a Thematic Mapper (TM) image, and the locations of known gold deposits. Favorable stratigraphic units, structural control factors, and alteration factors are considered. The work was conducted on an S600 I²S image-processing system. FORTRAN programs were developed for creating indicator patterns, statistical calculations, and pattern integration. Six indicator patterns were selected to predict mineral potential. They are conditionally independent according to pairwiseG ² tests, and an overall chi-square test. The potential area predicted using the 32 known deposits generally coincides with the prospect areas determined by geological fieldwork.