Volcano activity classification from synergy of EO data and machine learning: an application to Mount Etna volcano (Italy)

Abstract

Abstract This study investigates the integration of Earth Observation (EO) data and Machine Learning (ML) techniques for classifying volcanic activity states at Mount Etna, one of the world’s most active and monitored volcanoes. Using satellite data, including ground deformation, radiance, land surface temperature, sulfur dioxide emissions, and gravity anomalies, five volcanic activity states were identified: Quiet, Preparatory, Unrest, Eruption, and Cooling. Supervised ML algorithms, such as random forest, support vector machines, decision trees, and k-nearest neighbors, were employed to classify these states. Random forest achieved the highest accuracy, demonstrating its robustness for this application. The study addresses challenges like temporal and spatial disparities and class imbalances through data preprocessing, ensuring a reliable dataset for training and validation. A k-fold cross-validation approach was used to evaluate model performance systematically. The results underline the potential of ML techniques combined with EO data for volcanic hazard monitoring, with implications for improving risk assessment and early-warning systems. This methodology, tested on a well-instrumented volcano like Mount Etna, provides a foundation for extending the approach to other less-monitored volcanoes. These findings are one of the first attempts of integrating satellite data with Artificial Intelligence (AI) to enhance the accuracy of volcanic state predictions and mitigate risks associated with eruptions, while emphasizing the need for rigorous validation against well-documented case studies.

Keywords

• Machine learning
• Earth observation data
• Volcanic hazard
• Optical data
• SAR interferometry
• Mount Etna