Prospectivity Mapping for Porphyry Cu–Mo Mineralization in the Eastern Tianshan, Xinjiang, Northwestern China.

In order to comprehensively utilize regional-scale geological, geochemical and geophysical datasets for future exploration of undiscovered porphyry Cu–Mo polymetallic deposits (PCMPDs) in the Chinese Eastern Tianshan orogenic belt, three data-driven mineral prospectivity mapping (MPM) methods, namely ordinary weights of evidence (WofE), fuzzy weights of evidence (FWofE) and logistic regression (LR), were employed to integrate these datasets for mapping prospectivity of undiscovered PCMPDs. Firstly, the geological setting and mineralization of PCMPDs in the Eastern Tianshan district are reviewed. Then, spatial datasets based on geological maps, stream sediment geochemical data, and Bouguer gravity and aeromagnetic data are introduced, and on the basis of the prospecting model for PCMPDs, layers of structural, lithological, geophysical and geochemical evidences are constructed using the spatial datasets by means of GIS-based techniques. Finally, these evidential layers were integrated by using the WofE, FWofE and LR methods to obtain posterior probability maps of PCMPDs and the results are critically compared. The main conclusions are that: (1) the porphyry Cu–Mo mineralization in the Eastern Tianshan was occurred in the subduction boundary of the Late Paleozoic Dananhu-Dacaotan arc system of Kanguertag-Huangshan deep fault belt. This geological inference is supported by all the data-driven MPM methods; (2) the conditional independence assumption for both WofE and FWofE can be easily violated in practical applications. This issue seems very difficult to be circumvented due to geological correlations of evidence layers; (3) the uncertainty of the LR modeling approach particularly with respect to models using multiclass response variables mainly arises from over-fitting of the (ln-transformed) linear relationship; and (4) if there is no need for estimation of the number of undiscovered PCMPDs, the prospectivity map biasedly estimated by either WofE or FWofE modeling can be recommended for targeting new exploration areas with more detailed reconnaissance of potential undiscovered PCMPDs. Otherwise, the prospectivity map unbiasedly estimated by LR modeling with binary evidence modeling approach can be priority of use. [ABSTRACT FROM AUTHOR]