Application of hierarchical clustering, singularity mapping, and Kohonen neural network to identify Ag-Au-Pb-Zn polymetallic mineralization associated geochemical anomaly in Pangxidong district

Abstract

In order to identify Ag-Au-Pb-Zn polymetallic mineralization associated geochemical anomaly in Pangxidong district, hierarchical clustering (HC), singularity mapping (SM) and Kohonen neural network (KNN) methods have been employed to analysis stream sediment geochemical data in the study area. Firstly, the geochemical variables (i.e. 16 elements) were grouped by HC method, and Au, Ag, Pb, Zn, Hg, As and Sb were classified into the same group, which is in accordance with the fact that these elements are either the main ore-forming or accompanying elements that are associated with Ag-Au-Pb-Zn polymetallic mineralization. Then, the concentration values of the Au, Ag, Pb, Zn, Hg, As and Sb were processed by SM method to calculate the singularity indexes (i.e. α-values), which can be used to enhance the geochemical anomaly signature due to characterizing the local enrichment and depletion of these elements. Finally, α-values were classified by KNN method and a total of 9 groups of data were derived. The distributions of the samples in group 8 almost cover all the low α-values areas (i.e. anomaly areas with positive singularity). It can be thus inferred that the samples in group 8 indicate the enrichment of the mineralization associated elements Au, Ag, Pb, Zn, Hg, As and Sb. The main conclusions are: (1) HC method can be used as a data-driven approach to explore the proper association of ore-forming and accompanying elements linked with Ag-Au-Pb-Zn polymetallic mineralization in the study area; (2) the application of SM method has revealed the significant properties of local enrichment and depletion of the mineralization associated elements including Au, Ag, Pb, Zn, Hg, As and Sb; (3) KNN is a useful method to identify Ag-Au-Pb-Zn polymetallic mineralization associated geochemical anomaly by classifying α-values of these elements; and (4) the hybrid approach that combines HC, SM and KNN methods is an applicable tool for delineating target areas to prospect potential undiscovered deposits.