Automatic detection and characterization of Very Long-period seismic events for volcanic monitoring applications.

Real-time seismological applications are now crucial for the monitoring and surveillance of active volcanoes, as they are useful tools for the early detection of volcanic unrest. In open-vent active volcanoes,  Very Long Period (VLP) seismicity, typically associated with mild and persistent explosive activity, is of crucial importance for volcano monitoring, as its variations in occurrence rate and magnitude may prelude a phase of unrest.  Here we show a new method for the automatic real-time detection and characterization of  VLP seismicity at Stromboli active volcano (Italy).

The detection algorithm is based on the Three-Component Amplitude (TCA) obtained from waveform polarization and spectral analysis of the continuous recording, providing time of detection,  azimuth,  incidence,  amplitude, and frequency of the detected VLP events. The VLP amplitudes derived at all stations of the monitoring network, provided as peak-to-peak amplitudes and mean square amplitudes, are also used to perform an automatic localization of VLP source.

VLP detections and characterizations derived from our automatic detection algorithm are compared with detection derived from manual and automatic inspections of the seismic record and with VLP time histories from available published VLP datasets.

From this comparison, it turns out that the VLP detection time series produced by the automatic algorithm tracks fluctuations in the  VLP activity well,  as manually detected by the operators over a  ~20-year period, thus allowing us to include it into the real-time processing framework operating at Stromboli for volcano surveillance.