Orange-Volcanoes: Enhancing Data-Driven Petro-Volcanological Analysis with Applications in Petrological Volcano Monitoring

We introduce Orange-Volcanoes, an add-on for the open-source Orange Data Mining platform, designed to enhance data-driven workflows in petrology, geochemistry, and volcanology. Orange-Volcanoes extends the core features of Orange by incorporating tools for Compositional Data Analysis (CoDA), geochemical data preprocessing, and thermobarometric estimations.

These integrated tools enable users to perform machine learning, statistical evaluations, and predictive modeling on large petro-volcanological datasets while providing intuitive, interactive visualizations. The visual programming framework of the platform fosters collaborative research and ensures accessibility for a wide audience (e.g., scientists, educators, and students) without requiring programming expertise.

The combination of advanced machine learning and explainable artificial intelligence techniques, such as feature importance and Shapley additive explanations, supports deeper insights into geochemical variability and improves the interpretation of magmatic processes.

We explore the potential of Orange-Volcanoes through various case studies, showcasing applications such as clustering geochemical data and conducting petrological analyses. As the volume of volcanological and geochemical data continues to grow, this tool facilitates the integration of machine learning and data mining into standard scientific practices. The ability to apply diverse statistical and machine learning tools to geochemical data, while interactively visualizing step-by-step results, makes Orange and Orange-Volcanoes valuable assets for managing large multivariate datasets and supporting petrological volcano monitoring. Orange-Volcanoes represents a significant step forward in promoting reproducible, transparent, and collaborative research methodologies.