Application of Multivariate Statistics and Polarization Analysis in the Study of Volcanic Tremor Data at Stromboli Volcano, Italy

The continuous monitoring of seismic signals recorded on Stromboli, Italy, since 1985 has allowed to gain insight on this peculiar volcanic system. The spectral analysis of the persistent volcanic tremor along with the polarization analysis have highlighted an internal dynamics which evolves in long time spans of the order of months. According to previous studies, we demonstrate that the slow evolution of the internal changes can be followed with the application of techniques of multivariate statistics. In this note, we considered two unsupervised classification methods, i.e., the Principal Component Analysis and the Cluster Analysis. In particular, the Cluster Analysis allows us to identify three different clusters of spectra, according to the spectral amplitude of the signal and its frequency content. We present an overview of the results obtained from the application of statistical analyses we carried out from 1990 to 1999. For the most recent years, we combine these results with those of polarization analysis, considering azimuth and incidence angle of the seismic signal. Our interpretation of the overall data takes into account the relationships with volcanic activity using the images of a permanent video camera of INGV, which monitors the active craters 24 hours a day. Cyclic changes of the spectral features of tremor along with the incidence angle lead us to highlight seismic sources acting at different depth within the volcano feeder. This finding has important consequences for the understanding of: i) changes of intensity of Strombolian activity visible at the surface, and ii) the processes of lateral intrusion of magma, such as that inferred from the data recorded in 1997 and 1998.