Citation

Longo R, Lacanna G, Innocenti L, Ripepe M. IEEE Trans. Pattern Anal. Mach. Intell. 2024; ePub(ePub): ePub.

Copyright

(Copyright © 2024, Institute of Electrical and Electronics Engineers, Publisher IEEE (Institute of Electrical and Electronics Engineers))

DOI

10.1109/TPAMI.2024.3399689

PMID

38728129

Abstract

Explosive volcanic blasts can occur suddenly and without any clear precursors. Many volcanoes have erupted in the last years with no evident change in the eruptive parameters and with dramatic consequences for the population living nearby the volcano and the tourists visiting the active areas. In recent years, a big effort has been made to develop Early Warning systems to issue timely alerts to the population. At Stromboli volcano, the development of sensitive instruments to measure the deformation (tilt) of the ground has revealed that the volcano edifice is inflating tens of minutes before the explosion following a recurrent exponential ramp-like pattern. This scale-invariant of ground deformation has allowed the development of a quasi-deterministic Early Warning system which is operative since 2019. In this article we show how Artificial Intelligence and Machine Learning can be successfully applied to improve the efficiency and the sensitivity of Early Warning systems, provided the availability of a comprehensive experimental data set on past explosive events. The approach presented here for the Stromboli case demonstrates promising results also in forecasting the intensity of explosive events, offering valuable insights and new perspectives into the potential risks associated with volcanic activities.

Language: en