AUTHORED BY VERACIO’S CORINNA PENA JOHNSON, ‘AN UNSUPERVISED APPROACH TO SPECTRAL CLASSIFICATION’ WAS PRESENTED AT THE SEG 2026 CONFERENCE IN SALT LAKE CITY.

Accurate lithological classification is critical to understanding geological variability, but traditional approaches can be time-intensive and subject to interpretation. Corinna’s work explored how geochemical similarity can be used to identify lithological groupings without relying on spatial proximity or manual classification.

The study introduced an unsupervised classification workflow using XRF geochemistry acquired from core samples. By identifying elements that vary most between lithologies and comparing their spectral signatures, the approach groups samples based on geochemical similarity rather than their physical location along the core.

The results identified a spectral similarity threshold of 98.5% as a strong indicator that samples belong to the same lithological population, demonstrating the potential for more consistent classification while preserving important geochemical anomalies.

The outcome is an approach that allows geoscientists to examine natural geochemical groupings before assigning lithologies, supporting more accurate deposit characterisation and a better understanding of mineral patterns within the core.

DOWNLOAD THE ABSTRACT HERE