The Gaia app is based on artificial intelligence and is able to predict the depth of magma chambers for volcanoes. Knowing where magma reservoirs are located can help predict the occurrence and intensity of an eruption. The analysis tool developed by a group of researchers from the University of Florence, to which an article is devoted in Earth and Planetary Science Letters, is available to the scientific community. “What inspired our research – explains Simone Tommasini, professor of petrology and petrology and coordinator of the team – was the consideration that we needed new research methods to assess and manage volcanic risks, creating synergies between different skills.”
From available data on the chemical composition of clinopyroxenes, which are minerals commonly found in volcanic rocks, Gaia is able to determine the pressure, temperature and thus depth of the magma chambers from which these minerals originated. “The application – Tommasini continues – will allow us to analyze the history of the eruption and see if there is a correlation between the depth of the magma reservoir and the intensity of the eruption.”
“We developed the neural network system in two steps,” adds Lorenzo Cecchi, PhD student in the Department of Physics and Astronomy and the first signer on the article, “using a database found in the literature, first training it on one part of the data and then testing its predictive ability on the remaining part.” Gaia was then applied to study the ‘dissections’ of five active Italian volcanoes, Etna, Stromboli, Vesuvius, Vulcano and Campi Flegrei, and to reconstruct the properties of the magma reservoirs beneath each volcano during its entire eruptive history. The tests carried out “confirm the predictive power of the neural network in past episodes – conclude Duccio Vanelli and Luca Bindi, teachers of physics of matter and mineralogy, respectively – and we hope that the application will become an effective working tool thanks to which it will be possible to reveal the dynamics of volcanic systems, thus contributing to the collection of evidence Useful for assessing explosion risk.
© All rights reserved
“Infuriatingly humble social media buff. Twitter advocate. Writer. Internet nerd.”
