Machine learning in igneous petrology: opportunities, challenges, and insights into magmatic plumbing systems

Recent advances in analytical techniques, experimental studies, and computational modelling have significantly improved our ability to investigate magmatic plumbing systems. At the same time, the increasing availability of high-dimensional petrological datasets, ranging from crystal-scale chemical maps to multimodal geochemical and textural data, poses new challenges for data integration, interpretation, and physical consistency. In this context, machine learning (ML) emerges as a powerful tool to complement classical petrological approaches, offering new ways to explore complex datasets and quantify magma storage conditions and the evolution of plumbing systems.

In this contribution, we discuss how ML can be integrated into volcanology, with a particular focus on igneous petrology. We first outline the main opportunities offered by ML approaches, particularly their potential to automate tasks, enhance modelling strategies, and accelerate knowledge discovery. Then, we address key epistemological and practical challenges, such as ensuring transparency, model interpretability, calibration limits, reproducibility, and ethical considerations. These issues become especially critical in high-risk contexts such as volcanic hazard assessment, risk mitigation, and crisis management, where reliance on ML outcomes can have serious consequences for human lives (Ágreda-López & Petrelli, 2025).

Building on these considerations, we present examples of ML-based applications to reconstruct magma storage depths and plumbing system architectures. We conclude by discussing best practices for integrating ML in volcano science and by outlining future directions for combining physics-based models and data-driven approaches to improve our understanding of magmatic systems and their associated hazards.

References

Ágreda-López, M. & Petrelli, M. (2025). Opportunities, epistemological assessment and potential risks of machine learning applications in volcano science. Artificial Intelligence in Geosciences 6 (2). https://doi.org/10.1016/j.aiig.2025.100153