A MODERN LOOK AT THE DECCAN CHEMOSTRATIGRAPHIC SCHEME

The Deccan Traps are a continental Large Igneous Province covering large swaths of west-central India, with onshore erupted lava volumes that may have exceeded one million cubic kilometers. Although the total duration of magmatism is a matter of debate, recent geochronological work has demonstrated that the vast majority of volcanism occurred in a short (<1 Ma) interval straddling the Cretaceous-Paleogene boundary. The thickest and best exposed sections are in the Western Ghats, a mountainous escarpment that runs roughly parallel to the west coast of India. There, more than 3200 m of near-continuous lava stratigraphy is exposed. Seminal work done in the 1970s and ‘80s revealed the presence of several regionally traceable geochemical transitions in these sections. Beane et al. (1986, Bulletin of Volcanology v. 48, pp. 61–83), using cluster analysis on a compilation of over 1200 samples, identified stratigraphically and geochemically distinct lava packages in the Western Ghats. This chemostratigraphic scheme has served as the foundation for a wide range of Deccan studies, from mapping to understanding the province’s architecture, temporal evolution, and feeder system. However, the usefulness of the chemostratigraphy has been put into doubt when expanding it beyond the type sections of the Western Ghats, and the validity of interpreting units as true chronological markers has been questioned. The original statistical analysis focused on elements readily available via X-ray fluorescence: SiO₂, Al₂O₃, TiO₂, CaO, K₂O, P₂O₅, Ni, Ba, Sr, Zr, and Nb. However, issues caused by variable degrees of alteration and, particularly, fractional crystallization, have not been addressed, which has limited the predictive power of the geochemical clusters as currently defined. Here, we discuss the proposed origin of these geochemically distinct lava packages, and implications for geochronology. We propose a modernization of the chemostratigraphic scheme that takes into account a much greater suite of elements now commonly analyzed thanks to advances in analytical capabilities. We present preliminary results of statistical analyses of an updated Deccan sample database, discussing random forests and classification and regression trees as the basis for a more robust chemostratigraphy of Deccan lavas.