Hybrid multivariate typological model for the banded iron formations from the Bonito mine, Northeastern Brazil

Abstract

Typological ore delineation is a basic procedure that provides various bases in ore geology on which crucial developments rely on, in particular, those supporting mineral resource evaluation. To deal with several complex issues related to typological delineation of heterogeneous ore bodies, an unusual hybrid multivariate approach, based on a skillful combination of geochemical/geological knowledge and correspondence analysis (CA), was established. This approach was tested on banded iron formations (BIF) in the Bonito mine, Northeastern Brazil, enhancing the available geochemical knowledge by revealing grade combinations related to geological processes causing such deposits. Petrographic studies helped identify four main BIF-types: amphibolitic, hematitic, martitic, and magnetitic itabirites. An exhaustive geochemical database comprising 1,384 BIF samples assayed for Fe2O3, SiO2, Al2O3, P, and Mn grades was established and classified for the aforementioned BIF-types. The methodology provides an updated typological model based on a unique aspect of CA that allows a simultaneous projection of samples and their characteristics onto axes of the same abstract space; this helps validate mathematical results with geological information linked to sample attributes. The BIF typology model was established by applying K-means clustering to factorial scores derived through CA. Hence, the BIF typological model was updated, and new geochemical BIF-types helped interpret some of the main geological processes related to them. Geochemically, the factorial axes could be related to: 1) Iron rich BIF rocks, which are clearly represented on the positive side of Axis 1, comprising almost 50% of the total variance; 2) Manganese and alumina, associated with high silica content, by attachment to Axis 2, which describes a possible and critical terrigenous input.