HR: 1340h
AN: S33A-0290    [Abstracts]
TI: Analysis of Wide-Band Signals Using Wavelet Array Processing
AU: Nisii, V
EM: nisii@ov.ingv.it
AF: Istituto Nazionale di Geofisica e Vulcanologia -Osservatorio Vesuviano, Via Diocleziano, 328, Napoli, 80124 Italy
AU: * Saccorotti, G
EM: gilberto@ov.ingv.it
AF: Istituto Nazionale di Geofisica e Vulcanologia -Osservatorio Vesuviano, Via Diocleziano, 328, Napoli, 80124 Italy
AB: Wavelets transforms allow for precise time-frequency localization in the analysis of non-stationary signals. In wavelet analysis the trade-off between frequency bandwidth and time duration, also known as Heisenberg inequality, is by-passed using a fully scalable modulated window which solves the signal-cutting problem of Windowed Fourier Transform. We propose a new seismic array data processing procedure capable of displaying the localized spatial coherence of the signal in both the time- and frequency-domain, in turn deriving the propagation parameters of the most coherent signals crossing the array. The procedure consists in: a) Wavelet coherence analysis for each station pair of the instruments array, aimed at retrieving the frequency- and time-localisation of coherent signals. To this purpose, we use the normalised wavelet cross- power spectrum, smoothed along the time and scale domains. We calculate different coherence spectra adopting smoothing windows of increasing lengths; a final, robust estimate of the time-frequency localisation of spatially-coherent signals is eventually retrieved from the stack of the individual coherence distribution. This step allows for a quick and reliable signal discrimination: wave groups propagating across the network will manifest as high-coherence patches spanning the corresponding time-scale region. b) Once the signals have been localised in the time and frequency domain,their propagation parameters are estimated using a modified MUSIC (MUltiple SIgnal Characterization) algorithm. We select the MUSIC approach as it demonstrated superior performances in the case of low SNR signals, more plane waves contemporaneously impinging at the array and closely separated sources. The narrow-band Coherent Signal Subspace technique is applied to the complex Continuous Wavelet Transform of multichannel data for improving the singularity of the estimated cross-covariance matrix and the accuracy of the estimated signal eigenvectors. Using synthetic multichannel data generated for different signal types, the resolution in time-frequency-slowness domains is estimated by a direct comparision with the Windowed Fourier Transform MUSIC algorithm. The main advantages of our Wavelet Transform MUSIC algorithm (WTM) consists in: a) The simplicity of the procedure, as different frequency bands are processed at once without sacrifying time resolution; b) Its ability to selectively process only those data windows which depict significant coherence throughout the network, thus ensuring the physical meaning of the solution. We applied this metodology to the study of the wavefield characteristic of seismo-volcanic activity recorded by a dense array of short-period seismometers deployed at Stromboli volcano during its 2002-2003 eruption. WTM gives precise descriptions of the distribution in time and frequency of the different wavefield components, in turn providing precise estimates of the corresponding wavevectors. These achievements represent a crucial step toward a better understanding of the seismic wavefields associated with volcanic activity.
DE: 7219 Seismic monitoring and test-ban treaty verification
DE: 7280 Volcano seismology (8419)
DE: 7290 Computational seismology
DE: 7299 General or miscellaneous
SC: Seismology [S]
MN: Fall Meeting 2005