Simona Petrosino

Automatic Classification of Seismic Signals at Mt. Vesuvius Volcano, Italy, Using Neural Networks

We present a new strategy for reliable automatic classification of local seismic signals and volcano-tectonic earthquakes (vt). The method is based on a supervised neural network in which a new approach for feature extraction from short period seismic signals is applied. To reduce the number of records required for the analysis we set up a specialized neural classifier, able to distinguish two classes of signals, for each of the selected stations. The neural network architecture is a multilayer perceptron (mlp) with a single hidden layer. Spectral features of the signals and the parameterized attributes of their waveform have been used as input for this network. Feature extraction is done by using both the linear predictor coding technique for computing the spectrograms, and a function of the amplitude for characterizing waveforms. Compared to strategies that use only spectral signatures, the inclusion of properly normalized amplitude features improves the performance of the classifiers, and allows the network to better generalize. To train the mlp network we compared the performance of the quasi-Newton algorithm with the scaled conjugate gradient method. We found that the scaled conjugate gradient approach is the faster of the two, with quite equally good performance. Our method was tested on a dataset recorded by four selected stations of the Mt. Vesuvius monitoring network, for the discrimination of low magnitude vt events and transient signals caused by either artificial (quarry blasts, underwater explosions) and natural (thunder) sources. In this test application we obtained 100% correct classification for one of the possible pairs of signal types (vt versus quarry blasts). Because this method was developed independently of this particular discrimination task, it can be applied to a broad range of other applications.

Changes in the Coda Decay Rate and Shear-Wave Splitting Parameters Associated with Seismic Swarms at Mt. Vesuvius, Italy

We study the time changes of (1) the b -value of the Gutenberg-Richter distribution, (2) the inverse coda (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(Q(Q_{\mathrm{C}}^{-1})\) \end{document}, and (3) the shear-wave splitting parameters (i.e., the time delay T d between qS 1 and qS 2 phases and the polarization direction of the qS 1 wave) for small-magnitude volcano-tectonic earthquakes of Mt. Vesuvius, Italy. We used for (1) the seismic catalog of Mt. Vesuvius seismicity starting from January 1994, for (2) a selected (on the basis of the best signal-to-noise ratio) set of data with hypocentral distances smaller than 4 km recorded at station BKE (analogical) with a 1-Hz vertical seismometer during the period from January 1994 until the present, and for (3) a set of data recorded at two digital, high dynamical range, portable short-period seismic stations. These stations (BKE and BKN) were in operation in two periods, BKE (digital) from January 1999 to the middle of 2000 and BKN from January 1999 to the end of 1999; the hypocentral distances were not greater than 4 km. We found evidence of time changes of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(Q_{\mathrm{C}}^{-1}\) \end{document} measured at high frequency (6, 12, and 18 Hz). The changes seem to be correlated with the occurrence of two swarms with largest magnitudes of 3.4 and 3.6, respectively in April 1996 and October 1999. The earthquake with the largest magnitude in the second swarm appears to be the largest event since the latest eruption in 1944. An increase in \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(Q_{\mathrm{C}}^{-1}\) \end{document} starts (msec/strain units) for after the occurrence of both swarms, reaching a maximum after more than 1 yr for the first swarm and after 6 months for the second swarm. These two changes were not accompained by any corresponding variation of the b -value, which shows an almost constant (inside the statistical uncertainty) pattern. The last swarm ( M 3.6) was preceeded by an increase of T d at both stations, indicating a possible change of the stress state before the M 3.6 earthquake. The qS 1 polarization direction also shows a variation in correspondence to the same earthquake, which was interpreted as generated by an increase of the differential stress acting at a regional scale in the north-south direction shortly before the M 3.6 event. The strain change associated to this earthquake was estimated to be of the order of 10-9 using data from the straingram recorded at a Sacks-Evertson dilatometer located about 3 km from the epicenter. The given information allows us to estimate the sensitivity of the the measured parameters to the strain change induced by the M 3.6 earthquake. The sensitivity is of the order of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(1.4{\times}10^{9}{\ }(Q_{\mathrm{C}}^{-1}{/}\mathrm{strain\ units})\) \end{document} for \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(Q_{\mathrm{C}}^{-1}\) \end{document} and is of the order of 2 × 1010 (msec/strain units) for T d. Manuscript received 15 July 2003.

Location of the Source and Shallow Velocity Model Deduced from the Explosion Quakes Recorded by Two Seismic Antennas at Stromboli Volcano

Abstract The seismic wavefield associated to the ongoing eruptive activity at Stromboli volcano (Italy) is investigated using data from two small-aperture, short-period seismic arrays deployed on the northern and western flanks, located at about 1.7 km from the active craters. Two distinct approaches are used to analyze the recorded signals: 1. 1) the zero-lag cross-correlation method is used to analyze the explosion quakes data, to estimate slowness and backazimuth as a function of lapse time; 2. 2) multiple filter technique and phase matched filtering are used to estimate Rayleigh wave dispersion, to obtain a shallow velocity model of the two sites. Estimates of slowness vectors at the two different array sites show a primary (volcanic) source located at shallow depth beneath the crater region. Secondary sources associated with path effects are located in close proximity of the sector graben of Sciara del Fuoco and of the old parasitic cone of Timpone del Fuoco. The shallow velocity structure derived for the western flank depicts striking resemblance with that previously inferred for the northern flank of the volcano.

Seismic Attenuation and Shallow Velocity Structures at Stromboli Volcano, Italy

The surface-wave field associated with the explosive activity at Stromboli volcano is investigated using data recorded by two short-period seismic arrays, deployed on the north and west flanks of the volcano. The group-velocity dispersion curves for Rayleigh waves are derived using the multiple filter technique. The phase-velocity dispersion curves are recovered using a phase match filter and compared with that inferred from zero-lag cross-correlation analysis applied to the array data. These analyses indicate Rayleigh-wave group velocities ranging from 0.29 to 0.24 km/sec in the 1.5- to 8.0-Hz frequency band, and phase velocities ranging from 1 km/sec at 1.5 Hz to about 0.3 km/sec at frequencies above 5 Hz. In addition, the dispersive properties of the attenuation coefficient (γ) for Rayleigh waves are inferred from application of the multiple filter technique to seismograms recorded at different distances from the source. These results are validated through examination of the spectral amplitude decay with distance for both body and Rayleigh waves. The values of the body-wave quality factor thus obtained are Q α = 20 and Q α = 6 for the north and west side of the island, respectively. The velocity and attenuation dispersion curves are inverted for the shear-wave velocity and Q β structures down to a depth of about 200 m. Shear-wave velocities for the west flank range from about 0.3 km/sec for the uppermost 17-m-thick layer to 1.9 km/sec at depths greater than 200 m. Comparison with previous studies indicates a similar velocity structure for the north and west flanks. The attenuation structure for the west flank is described by a shallower, 36-m-thick layer with Q β = 9, underlain by a half-space with Q β = 50. On the north flank, Q β = 40 for the shallower 30-m-thick layer and Q β = 44 for the underlying half-space. Residuals from analysis of the spectral decay with distance are used to quantify site effects affecting the different array elements on the west flank. Local amplifications at that array are interpreted in terms of an edge effect associated with concave topography. Velocity similarities observed at the north and west flanks are compatible with surface geologic data. Discrepancies in attenuation properties at the two sites are interpreted in terms of different degrees of heterogeneity and crack density controlling the scattering quality factor Q s. Manuscript received 15 April 2001.

A local-magnitude scale for Mt. Vesuvius from synthetic Wood-Anderson seismograms

The Local-Magnitude scale actually in use at Vesuvius Observatory is basedon the measure of seismogram coda duration, and calibrated with data fromIrpinia aftershocks. A recent study on local seismic attenuation at Mt.Vesuvius reveals coda shapes highly different from those from Irpiniaaftershocks, and a very low quality factor, if compared to the average Qof the region, indicating the necessity of the revision of the Magnitudescale, in order to better compare the seismic energy associated to the localseismicity of Mt. Vesuvius to that of other active volcanoes. Being theseismic attenuation parameters known in the area, we could correct theseismic amplitudes for the path effect to obtain precise estimates of theamplitude level of the displacement spectrum. Hence we estimated theMoment-Magnitude, MW, for a set of well recorded micro-earthquakes.To use the Richter formulaML =log10Amax −log A0(R)we estimated the log Amplitude-Distance correction curve, - log A0(R),numerically synthesizing an S-wave-packet and letting it propagate in aearth medium with the same attenuation properties of those measured at Mt.Vesuvius. Then we synthesized the Wood-Anderson equivalent seismogram forthe same data set and used the distance correction in order to calculate theWood-Anderson Magnitude.This Magnitude scale was normalized in order to fit the Richter formulavalid for Southern California at a distance Δ of 10 km, and resultsto be MWA =log A + 1.34log(R) −1.10. The comparison of the Wood-Anderson scale with the Duration-Magnitude scalein routine use at Vesuvius Observatory indicates that care must be takenwhen the estimate of the Duration-Magnitude is carried out for smallearthquakes recorded at a site characterized by a high level of seismicnoise.

Shallow velocity model of the northern flank of Stromboli Volcano, deduced by high frequency surface wave dispersion

The wavefield produced by the Stromboli volcano explosion quakes shows a significant amount of surface waves. Rayleigh waves recorded by a linear array have been investigated to infer the shear-wave velocity model of the Stromboli northern flank. The group velocity dispersion curve was obtained using the multiple filter technique, while the phase velocity dispersion curve was calculated both by phase-matched filtering and performing a p–ω stack on the observed waveforms. Through the inversion of these curves we were able to recover the shear-wave structure to a depth of about 190 m.

Tidal and hydrological periodicities of seismicity reveal new risk scenarios at Campi Flegrei caldera

The volcano-tectonic seismicity occurring at Campi Flegrei caldera during its present unrest phase, started in 2005, is distributed into time-clustered events emerging from a background composed of earthquakes with higher inter-arrival times. Here, we show that clustered seismicity is cyclically recurrent at time scales from semidiurnal to annual, matching tidal and hydrological periodicities. These results suggest that volcano-tectonic seismicity at Campi Flegrei caldera is driven by both variations in the deep magmatic feeding system and exogenous phenomena, as rainfall or global inflation/deflation cycles of the Earth’s crust, controlled by the lunisolar interaction. Consequently, the role of exogenous triggers in the evolution of the present unrest phase should be properly considered in the elaboration of volcanic risk scenarios, presently limited to the study of surface indicators of deep phenomena.

Experimental study for evaluation of a suitable ground displacement monitoring system: Pilot hole Campi Flegrei Deep Drilling Project case

The paper presents an experimental study carried out in 2012 during the drilling activity for a pilot hole performed in the framework of the Campi Flegrei Deep Drilling Project. A monitoring network has been installed to test and choose a suitable ground deformation system for the subsequent deep drilling of about 3.5 km in the Campi Flegrei Caldera (Italy). We describe the seismic network installed to characterize the structure of the pilot hole area and the ground deformation monitoring system chosen for the small drilling area. Data analysis and results obtained indicate that Total Station is a suitable tool for this case.