Multivariate stochastic modelling: towards forecasts of paroxysmal phases at Stromboli

The assessment of potential volcanic eruptions is a critical aspect when evaluating the safety of populated areas. A stochastic approach has been developed for the analysis and simulation of data sampled at active volcanoes. This approach allows for the detection and quantification of time correlation, cross correlation and volcanic tremor decomposition in order to identify potential precursors of major eruptions. The stochastic approach has been applied to data gathered at Stromboli volcano. Multivariate kriging analysis of the tremor intensity and the number of events per day has enabled time component estimation which could furnish additional monitoring variables for the forecast of paroxysmal phases at Stromboli, i.e., phases of significantly increased energy output. This multivariate approach is sufficiently general to be applicable to the analysis of multiparametric datasets recorded at volcanoes with different types of eruptive behaviour.