My research focuses on time series analysis applied to monitoring and forecasting active volcanoes. Within the European project MULTIMO (2001-04) I followed spectral, dynamical and stochastic approaches. Later, empirical relationships from material failure theory were used to estimate the onset of an eruption by using the inverse event rate of seismic signals. The main problem of classifying the state of an active volcano still remains unsolved, as only indirect parameters can be estimated with common monitoring networks. The widespread use of Bayesian Belief Networks (BBN) and Bayesian Event Trees (BET) is reflecting this balance of knowledge and ignorance. One of the most informative data that can be (easily) acquired at a volcano is seismic noise / volcanic tremor. At Erta Ale, Stromboli, Dallol, Ambrym its analysis led to the highlighting of different alternating volcanic regimes, a phenomenon common to several and quite different volcanoes. Moreover, a non trivial influence of (volcano)tectonic events on these regimes was also recognized at many volcanoes. The characterization of such regimes can be based on classical intensity or spectral parameters, but can also profit from the use of more sophisticated analyses, such as Singular Value Decomposition, Self Organizing Maps or Nonnegative Matrix Factorization.

During my stay at ERI I propose to continue this line of research from the methodological point of view and from the applicative point of view, using continuous seismic data recorded at Japanese volcanoes and made available by ERI. The volcanic regime changes will be also tentatively related to the occurrence of (volcano)tectonic seismic events.