High resolution geochemical mapping in the Mociços mine (Ossa-Morena Zone, Portugal). Contributes from machine learning methods

Abstract

The Mociços mine was exploited for copper in the early twentieth century. A recent soil geochemistry campaign with portable X-ray fluorescence equipment permited to map the surroundings of this ancient mine with high resolution. The analysis of the results using machine learning methods, namely, principal component analysis, hierarchical and k-mean clustering, and the mapping of the observations, allows a better understanding of the geochemical behavior of the elements. The principal component analysis and the k-means method have comparable results and allow to define the zone of mineralization and the outcropping of a dyke of acid rocks. The hierarchical agglomeration method allows to group the mineralized zones with the mine waste sites. Using the spatial mapping of the clusters it was possible to identify the regions marked by the geochemical behaviour of copper and zinc as well as to find relationships between the mineralized vein and outcropping acid rocks in the region.