The S3 orogenic gold targeting algorithm: A case study from the Timmins gold camp, Ontario, Canada

Mineral exploration targeting of orogenic gold deposits is challenging, especially in concealed terrains. To meet this challenge, a practical orogenic gold targeting method using structural complexity (SC), self-organizing map (SOM) and supervised deep neural network (SDNN) is proposed. As orogenic gold deposits are structurally controlled, the use of high-resolution aeromagnetic data in order to delineate concealed structures is routinely performed by mineral explorationists. SC is defined as the intersection density or the orientation diversity of linear structures, which can be derived from those magnetic data and/or acquired by geological structural mapping. Common features from both the SC and the magnetic data can be grouped together into one class by the SOM, essentially an unsupervised neural network classification method. A small percentage of the data can also be designated as anomalous classes by the SOM. The SOM results can support gold exploration targeting in areas lacking any known gold deposits or occurrences. However, the predictive targeting of orogenic gold mineralization can also be achieved with the support of a SDNN using the SC results. The SDNN can be trained by the SC data over known gold deposits or occurrences. The top few percentiles of the target probabilities are used for target generation. As the orogenic gold targeting process involves SC, SOM, and SDNN, it may be called the S3 method. This proposed S3 method is tested using public domain high-resolution aeromagnetic data covering the world-class Timmins gold camp in the Abitibi greenstone belt, Superior Province, Ontario, Canada. The results document that the S3 algorithm is an effective and robust AI-supported method to target concealed orogenic gold mineralization.