Artificial intelligence AI has helped researchers to peer into the hidden interior of one of Europe’s most active volcanoes. Using advanced seismic networks, scientists have created the most detailed 3D model yet of the Vulcano volcano in northern Sicily. The study marks a milestone in efforts to understand volcanic systems and could shape future monitoring and evacuation strategies for millions living near active volcanoes worldwide.
Artificial intelligence-powered model reveals volcanic structure
The research team of the University of Geneva in collaboration with the National Institute of Geophysics and Volcanology of Italy (INGV) was successful in integrating the artificial intelligence with the nodal seismic data to reproduce the inner structure of Vulcano. The project, according to lead author Douglas Stumpp, a doctoral researcher in the Department of Earth Sciences at the University of Geneva, provides current insight into processes that happen deep within the surface that were not previously visible with such accuracy.Earthquake signals that take place beneath volcano systems have been traditionally studied in volcano seismology.
General structural patterns were described by large-scale studies, but scientists could very seldom study the fine details of the underground volcanic activity. This had shortcomings in the knowledge of how magma and fluids flow towards the surface.The Vulcano project shows that with the help of machine learning algorithms, seismic ambient noise can uncover faults and underground routes with amazing precision. Stumpp noted that the method is a big step towards the understanding of the internal plumbing systems which cause eruptions.
History and renewed activity of Vulcano.
The Vulcano volcano is a volcano of more than 2,000 years that is situated on the Aeolian Islands in the northern side of Sicily. There are several eruptions in the historical records, and ash and steam frequently hit the communities around. The word volcano in English, and other languages, originated because of its modern name.In 2021, Vulcano passed a new stage of turmoil.
The volcano became more explosive and created a number of ash and steam eruptions. The exercise gave scientists a natural chance to gather useful seismic data. Approximately, 200 portable seismic sensors were spread all over the island to capture the ground vibrations within a one month span.These state of the art instruments recorded data at a broad frequency spectrum. This mass of data was digested by the supercomputer of the University of Geneva, Yggdrasil, and thus a complete 3D model of the volcanic system was recreated.
Seismic imaging and magma pathways
The resulting model disclosed the way faults at the bottom of the volcanoes act as pathways to magma and hydrothermal fluids. Such faults create an elaborate plumbing system that directs materials to the surface. Such pathways are fundamental in predicting hazards when volcanoes are about to erupt because the variation of the fluid circulation usually precedes the eruption.
Scientists have pointed out that although the model cannot predict volcanic activity at this time, it offers more detail than ever before on how Vulcano is internally configured. Such a combination of resolution provides a basis to volcanologists of creating dynamic monitoring systems capable of tracing changes in real time.
Artificial intelligence in volcanic monitoring
The fact that AI was integrated was essential to the success of the project. The neural networks were then used to process the massive amount of seismic signals and patterns were brought into view that would be obscured. Through machine learning and deep learning, the team was able to make quicker and more precise reconstruction of underground structures.Stumpp elaborated that the end product is to receive seismic ambient noise data in real-time.
This may enable scientists to determine changing conditions within volcanoes in real time. This ability would allow the authorities to develop flexible evacuation strategies that directly react to identified changes in the ground.Nevertheless, scientists admitted that processing of large-scale datasets in ultra-fast mode is still a technical challenge. Nevertheless, the study by Vulcano shows that the mechanical integration of automated seismic networks and AI-based methods is becoming more and more realistic.
Academic publication
The complete research was entitled Neural Network Nodal Ambient Noise Tomography of a transient plumbing system under unrest, Vulcano, Italy and was published by Nature Communications. The authors attributed their results to both the cooperation of the University of Geneva and the INGV of Italy and the high computing power of the Yggdrasil supercomputer.Although predictive eruption models do not exist yet, the study proves that AI-improved seismology can unveil structures that are concealed under volcanic systems
Scientists are confident that this body of work is going to experience a booming development with real-time monitoring becoming the new edge.Stumpp concluded that machine learning will become an even more important part of volcanology. He observed that hazard planning in the decades to come would change based on a dynamical analysis of the volcanic systems instead of a retrospective analysis.