75 Mineralogy, Geochemistry, Petrology, and Volcanology Student Session (Posters)

GSA’s MGPV student representatives invite the division’s student population to present their research (oral or poster). We encourage abstracts from any subdiscipline within our respective fields with hopes of stimulating a multi-disciplinary, early career researcher (ECR) community within MGPV.

Sunday, 19 October 2025
8:00 AM - 5:30 PM
HBGCC, Hall 1

Machine Learning Algorithms Successfully Predict Mantle Reservoirs of Basalts Using Trace Elements as Proxies for Isotopic Data

Paper number: 75-34
Program: Technical Sessions
Time: 8:00 AM – 8:00 AM

Authors

  • Daniel OHare Presenting Author
    Loma Linda University, Dept. of Earth and Biological Sciences, Loma Linda, CA
  • Ana Martinez-Ardila
    Loma Linda University, Earth and Biological Sciences, Loma Linda, CA
  • Benjamin Clausen
    Loma Linda Univ, Geoscience Research Institute, Loma Linda, CA
  • Germán Alférez
    Southern Adventist University, School of Computing, Collegedale, TN

Abstract

Identifying mantle reservoirs is fundamental to understanding mantle evolution and heterogeneity as well as the origin of Mid-Ocean Ridge Basalts (MORBs) and Ocean Island Basalts (OIBs).  The most effective tracers of mantle reservoirs are isotopic ratios, which are minimally fractionated during melting and resistant to alteration.  However, isotopic analyses are costly. Bivariate plots of trace elements (TE) and their ratios have also been used as mantle reservoir signatures, but are more susceptible to fractionation and alteration, limiting their diagnostic power compared to isotopes. This study aims to classify mantle reservoirs using machine learning (ML) with TE data alone, reducing the need for costly isotopic data and overcoming the limitations of bivariate TE plots. We compiled global geochemical data from MORBs and OIBs representing four mantle reservoirs: Depleted Mantle (DM or DMM), Enriched Mantle 1 (EM1), Enriched Mantle 2 (EM2), and HIMU (high μ or high time-integrated ²³⁸U/²⁰⁴Pb). Using the H2O package in Python for automated ML (AutoML), we trained supervised algorithms to classify basalts into these reservoir types based on TE input variables. The best H2O model, a Stacked Ensemble, achieved classification accuracy of 72% for all test locations and 84% for typical locations, demonstrating that TEs can effectively serve as isotopic proxies and distinguish between different mantle reservoirs. We repeated this process using an expanded HIMU class in the training data, in order to improve classification performance on the minority class.  We developed two open-source classification tools, both Stacked Ensemble models: one classifier for all four mantle reservoirs and one optimized for the minority HIMU class. To our knowledge, this is the first ML classification tool that identifies mantle reservoirs using only TE data, offering a robust and cost-effective alternative to isotopic methods.

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

doi: 10.1130/abs/2025AM-414378

© 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.