A novel method for mineral prospectivity mapping integrating spatial-scene similarity and weights-of-evidence

In this study, a novel method that integrates weights-of-evidence and spatial-scene similarity (WESS) was proposed for mineral prospectivity mapping. The weights-of-evidence model (WofE) was used to rank the importance and determine the weight of each ore-controlled factor. A spatial-scene similarity model was employed as the fundamental kernel theory, utilized to extract the spatial relation between ore-controlled factors and evaluation cells in the spatial-scene, and then used to measure the similarity between two scenes. The spatial-scenes of the known deposits were deemed the mineral cases, and all other spatial scenes were deemed new cases. We performed a similarity computation for each new case and all mineral cases one by one, attached the maximum similarity value to the central evaluation cell of the new case and adopted the value as the cell’s metallogenic potential index. A case study for Fe-Cu-Pb-Zn prospectivity mapping was performed in the Qimantage area of the eastern Kunlun metallogenic belt in China. The WofE and WESS models were used to evaluate the metallogenic potential and the receiver operating characteristic curve (ROC), area under curve (AUC), and a study area cumulative percentage curve (SCP) was utilized to perform a precise evaluation. Our experiment consisted of three sub-experiments (deemed A, B and C). In experiment A, all known deposits were used as training samples and verification samples simultaneously; the evaluation precisions of the WofE and WESS models were 75.8 % and 92.6 %, respectively. In experiment B, two thirds of the known deposits were selected as training samples, and the remaining one third was selected for verification; the evaluation precisions of the WofE and WESS models were 77.8 % and 88.9 %, respectively. In experiment C, half of the known deposits were selected for training, and the other half served as the verification sample; the evaluation precisions of the WofE and WESS models were 66.7 % and 81.6 %, respectively. The results showed that the proposed WESS model was more precise than the traditional WofE model.