Clusters of lava fountain events identified on strainmeter data at Etna volcano

Abstract

Lava fountains at Etna volcano are spectacular eruptive events characterized by powerful gas jets that expel lava fragments to several hundred meters and volcanic ash to several kilometers above the crater. Ash fall-out and dispersal cause critical hazards to both the vehicular traffic and the aviation, inducing the temporary closure of the southern Italy airports.In 2020-2022, Etna experienced more than 60 lava fountains. The dynamics of such explosive events is usually a gradual process, starting with a strombolian activity that progressively evolves in an intense and continuous explosive activity with a sustained eruptive column. The duration, the degree of explosiveness, the portion of effusive flows, etc., are usually variable implying a different degree of involved hazard. Recently, researchers attempted to manually classify lava fountains at Etna on the basis of volcanological and geophysical data. However, manual classification is time consuming and prone to subjective biases.We propose an automatic procedure to cluster the lava fountain events that occurred in 2020-2022 at Etna using unsupervised machine learning techniques. The clustering algorithm is applied on high precision strain signals recorded by the borehole dilatometer network deployed to monitor volcano deformation processes. In particular, the analysis focuses on the strain variations recorded during the lava fountain events to highlight similarities and differences among the eruptions in terms of induced ground deformation. The results disclose the main features of the strain signal effective to group the lava fountain events. Four well-separated and coherent clusters are identified improving the manual classifications performed by the experts. Moreover, the analysis reveals that the lava fountains clusters are grouped also over time showing possible transitions in the eruptive style.