Signal enhancement with generalized ICA applied to Mt. Etna volcano, Italy

Independent Component Analysis (ICA) is an emerging new technique in the blind identification of signals recorded in a variety of different fields. ICA tries to find the most statistically independent sources from an observable random vector, with the only restriction that all sources, but at the most one, are non-Gaussian; no other a priori information on sources and mixing dynamic system are needed. The applications of this technique to the analysis of volcanic time series are until today relatively few. In this paper, we show that ICA is a suitable technique to separate a volcanic source component from ocean microseisms in a seismic data set recorded at the Mt. Etna volcano, Italy. The encouraging results obtained with this methodology in the presented case study support its wider applicability in the volcano seismology context. The separation and consequent elimination of noise components from the continuous seismic signal can in fact facilitate tasks such as the characterization of volcanic regimes, their relationship with tectonic activity and the identification of possible precursors of paroxysmal phases.