Study proposes accurate AI model to help predict volcanic eruptions

Researchers at the University of Granada have developed an innovative system demonstrating a very accurate machine-learning algorithm developed to predict volcanic eruptions.

• The authors propose the use of neural networks to classify the volcanic state of alert and provide a probability of volcanic eruption.
• The study demonstrates a successful application of data-driven models in solving predictive complex real-world problems and positions them as an effective addition to the monitoring systems deployed in volcanological observatories.

Scientists at the University of Granada developed a very accurate machine-learning algorithm for predicting volcanic eruptions. Their work, published recently in Frontiers in Earth Science, demonstrates how this technique can assist us in better understanding and forecasting volcanic activity, which is a crucial step toward increasing disaster preparedness and decreasing volcanic dangers.

The researchers analyzed a large dataset of seismic recordings from Mount Etna gathered over several decades. By training the machine learning model on historical data, the researchers were able to identify seismic signals that consistently preceded eruptions. The algorithm’s prediction capacity was then tested against more recent seismic data, and it achieved an impressive accuracy rate of more than 90%.​ 

“The results obtained in this work evidence the reliability of this method as an automatic tool capable of forecasting volcanic eruptions with great potential, and it is shown to be transferable to different volcanic systems around the world,” the authors said.

“By applying signal processing techniques on seismic records, we extracted four different seismic features, which usually change their trend when the system is approaching an eruptive episode,” study authors said.

“We built a temporal matrix with these parameters then defined a label for each temporal moment according to the real state of the volcanic activity (Unrest, Pre-Eruptive, Eruptive). To solve the remaining problem of developing early warning systems that are transferable between volcanoes, we applied our methodology to databases associated with different volcanic systems, including data from both explosive and effusive episodes, recorded at several volcanic scenarios with open and closed conduits: Mt. Etna, Bezymianny, Volcán de Colima, Mount St. Helens and Augustine.”

The machine learning algorithm’s performance in properly predicting eruptions bodes well for its application to other active volcanoes across the world. This technology can help with disaster planning by providing early warnings and allowing authorities to enact evacuation plans and other safety measures on time.

Furthermore, the work demonstrates the greater potential of machine learning and other such technological advancements in geophysical research.

Machine learning algorithms’ capacity to analyze complicated information and identify hidden patterns can be applied to a wide range of earth science applications, including earthquake prediction and climate modeling.

The researchers intend to enhance their model further and test its relevance to various volcanic environments. Collaborative efforts with foreign research institutions are underway to test the algorithm on additional volcanoes, with the goal of creating a global network of AI-powered volcanic monitoring systems.

The creation of a highly accurate machine-learning method for predicting volcanic eruptions is a significant leap in geophysics. As the algorithm evolves and improves, the hope is it will become a valuable tool in protecting communities living in the shadow of active volcanoes.

References:

¹ Universal machine learning approach to volcanic eruption forecasting using seismic features – Pablo Rey Devesa et al. — Frontier in Earth Science, June 26, 2024 – https://doi.org/10.3389/feart.2024.1342468