143 A Showcase of Student Research in Geoinformatics and Data Science (Posters)

Students are invited to present original research posters relevant to geoinformatics and data science, including geospatial analysis, machine learning, big data, modeling, and visualization. Posters will be judged for divisional awards.

Monday, 20 October 2025
8:00 AM - 5:30 PM
HBGCC, Hall 1

Statistical Discrimination of Volcanic Rock Units using Handheld Spectrometer Devices

Paper number: 143-1
Program: Technical Sessions
Time: 8:00 AM – 8:00 AM

Authors

Abstract

This study reviews the viability of using parial least square-disctiminant analysis (PLS-DA) and linear discriminant analysis (LDA) to discriminate volcanic rocks using reflectance spectra. In geologic remote sensing applications, key absorbane troughs in the visable near-infrared (VNIR) are the ferric (Fe3+) and ferrous (Fe2+) oxides at various bands from 0.4-1.1 µm. In the shortwave infrared (SWIR), many hydroxyl and carbonate compounds are captured from 1.1-2.4 µm. If absorbance troughs are observed in these ranges, they can be attributed to specific minerals that contain these compounds in their chemical formula. Geologic remote sensing, typically based on spectral libraries, has largely been confined to minerals that contain these compounds in their chemical formula (i.e. goethite and hematite for iron, gypsum for hydroxyls). However, this leaves several types of geologic materials essentially 'undetectable' in these spectral ranges.

 

Both PLS-DA and LDA are common machine learning methods used with spectral data. We apply these techniques here to discriminate compositionally diverse volcanic rocks. Six rock unites were tested, some more geochemically similar than others, including: three felsic units, one intermediate, one mafic, and one intrusive granodiorite. After spectra were collected, we used PLS-DA to test whether rock units could be differentiated and LDA to illustrate our results. First, we tested whether all six rock units could be separated and then ran a condensed model where all three felsic units were grouped togehter. The six-unit PLS-DA model was able to differentiate the non-felsic units with an average accuracy of 86.34%, but struggled to separate the three felsic units from each other. Overall accuracy of the PLS-DA model was 78.17%. The condensed model was able to differentiate the non-felsic units with over 90% accuracy, but still struggled to separate the felsic group. These models were an overall success in discriminating rock units based off spectra and provide a new path forward for identifuying rock inits wih non-distinct spectral features. 

Geological Society of America Abstracts with Programs. Vol. 57, No. 6

doi: 10.1130/abs/2025AM-412481

© Copyright 2025 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.