Water samples were collected at 10 points in parts of the eastern Quadrilátero Ferrífero (QF), located in a mining region in the southeast of Brazil. The aims of this study were to find possible relationships among dissolved organic carbon (DOC), metals and other parameters measured in the region studied and evaluate the Kohonen neural network as a tool to analyse this geochemical multivariate data set. Physico-chemical analyses were performed in situ and in the laboratory, where concentrations of DOC and a suite of metal ions were determined. The Kohonen neural network allowed an easier visualisation and interpretation of the results and helped to define the relationships among them. In this way, a relationship between DOC and Fe and a possible effect of seasonality on the distribution of the samples were indicated. Signs of lithology were detected in the analyses, especially considering the elements Ca, Mg, Mn and Sr.
Quadrilátero Ferrífero - Brazil; dissolved organic carbon; Kohonen neural network; environmental geochemistry
