Statistical moments of power spectrum: a fast tool for the classification of seismic events recorded on volcanoes

Abstract

Spectral analysis has been applied to almost thousand seismic events recorded at Vesuvius volcano (Naples, southern Italy) in 2018 with the aim to test a new tool for a fast event classification. We computed two spectral parameters, central frequency and shape factor, from the spectral moments of order 0, 1, and 2, for each event at seven seismic stations taking the mean among the three components of ground motion. The analyzed events consist of volcano-tectonic earthquakes, low frequency events and unclassified events (landslides, rockfall, thunders, quarry blasts, etc.). Most of them are of low magnitude, and/or low maximum signal amplitude, therefore the signal to noise ratio is very different between the low noise summit stations and the higher noise stations installed at low elevation around the volcano. The results of our analysis show that volcano-tectonic earthquakes and low frequency events are easily distinguishable through the spectral moments values, particularly at seismic stations closer to the epicenter. On the contrary, unclassified events show the spectral parameters values distributed in a broad range which overlap both the volcano-tectonic earthquakes and the low frequency events. Since the computation of spectral parameters is extremely easy and fast for a detected event, it may become an effective tool for event classification in observatory practice.