This study utilizes unsupervised machine learning methods to analyze continuous seismic signals recorded during Iceland's 2021 Geldingadalir eruption. The deep embedded clustering (DEC) technique identified distinct seismic signal clusters corresponding to various eruption phases (unrest, lava extrusion, varying lava fountaining intensities). A precursory volcanic tremor sequence three days before the eruption was detected, potentially signifying impending eruptive activity. The identified seismicity patterns and their temporal changes offer insights into the eruption's evolution, particularly the transition from vigorous outflows to lava fountaining, potentially linked to an increase in discharge rate.

Zahra Zali, S. Mostafa Mousavi, Matthias Ohrnberger, Eva P. S. Eibl, Fabrice Cotton