Statistical Evaluation of Classification Diagrams for Altered Igneous Rocks

Authors: SURENDRA P. VERMA, RODOLFO RODRIGUEZ RIOS, ROSALINDA GONZÁLEZ-RAMÍREZ

Abstract: The International Union of Geological Sciences (IUGS) has proposed recommendations for the classification of relatively fresh volcanic rocks, but with no specific instructions for altered volcanic rocks, other than discouraging the use of the total alkalis versus silica diagram. The Nb/Y-Zr/TiO2 diagram has been in use for the classification of altered rocks now for over 30 years. Recently (during 2007) another diagram (Co-Th) has been proposed to replace this old diagram, particularly for altered arc rocks. Using an extensive database of all kinds of relatively fresh rocks from four tectonic settings (island arc, continental rift, ocean island, and mid-ocean ridge), as well as from three settings excluding island arc, we carried out an objective evaluation of the old Nb/Y-Zr/TiO2 diagram for rock classification. Similarly, for the evaluation of the new Co-Th diagram, an extensive database of similar rocks from island arcs, the Andean active continental margin, continental rifts, ocean islands, and the Mexican Volcanic Belt, was used. Statistical parameters of correct classification or success rate and minimum misclassification defined in this work, respectively, were used to evaluate these diagrams. Our results of the quantification of these parameters showed that none of these diagrams seems to work precisely for the classification of fresh rocks. It is therefore difficult to imagine that they would work well for the classification of altered rocks. Thus, there is an urgent need to apply correct statistical methodology for handling compositional data in proposing new classification diagrams that could provide classification and nomenclature to altered volcanic rocks fully consistent with the IUGS nomenclature for fresh rocks.

Keywords: TAS classification, volcanic rocks, plutonic rocks, chemical classification, correct statistical analysis of compositional data

Full Text: PDF