GNGTS 2018 - 37° Convegno Nazionale

GNGTS 2018 S essione 3.3 735 capacità del metodo di ottenere ottime stime delle componenti della velocità e del parametro di anisotropia ε , anche se utilizzando un valore fisso del parametro δ , che comunque è risultato poco influente sui risultati finali per valori minori di 0.01. Ringraziamenti. Desidero ringraziare Flavio Poletto, Biancamaria Farina e Giuliana Rossi per le utili osservazioni e commenti fatte duramte le discussioni su questo argomento. Bibliografia Böhm G., Carrion P., Marchetti A., Pettenati F. and Vesnaver A.; 1993: Reflection tomography in complex structures. Expanded Abstracts of 55th EAEG Meeting (Stavanger), D014. Thomsen L.; 1986: Weak elastic anisotropy. Geophysics, 51, 1954-1966. Stewart R.; 1993: Exploration Seismic Tomography: Fundamentals. Course note series, vol. 3, S. N. Domenico, Editor. SEG - Society of Exploration Geophysicists. Vesnaver A., Böhm G., Madrussani G., Petersen S. and Rossi G.; 1999: Tomographic imaging by reflected and refracted arrivals at the North-Sea. Geophysics , 64(6) 1852-1862. IMPROVING MAGNETOTELLURIC IMPEDANCE TENSOR ESTIMATES BY SELF-ORGANIZING MAPS R. Carbonari 1 , R. Di Maio 1 , E. Piegari 1 , L. D’Auria 2 , A. Esposito 3 , Z. Petrillo 3 1 Dipartimento di Scienze della Terra, dell’Ambiente e delle Risorse, Università di Napoli Federico II, Naples, Italy 2 Instituto Volcanológico de Canarias, San Cristobál de la Laguna, Spain 3 Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Napoli Osservatorio Vesuviano, Naples, Italy Introduction. In the last decades, the magnetotelluric (MT) method has been proved to be a useful geophysical tool in different contexts, from geothermal reservoir characterization to crustal structures studies (Simpson and Bahr, 2005). Nevertheless, the MT method is very sensitive to the presence of noise. Indeed, as the method is based on the measurement of both the electric and magnetic components of the natural electromagnetic (EM) field, it fails when these components are affected by noises of different nature. In particular, in industrialized and urbanized context, the MT time series may be strongly affected by man-made noise and, as a consequence, the impedance tensor estimates, given by the ratio between the electric and magnetic components of the MT field, could be unreliable. To improve the reliability of these estimates, most of the proposed approaches rely on the robust evaluation of the impedance tensor (e.g., Sutarno and Vozoff, 1989; Egbert, 1997) as well as on the use of remote reference MT stations (Gamble et al. , 1979) or on the combination of both approaches (Jones et al. , 1989; Egbert, 1997). However, these methods are not always effective. The robust methods fail when most of the data are affected by noise, giving as result a biased impedance tensor, while the remote reference approach is ineffective when the noise is correlated between reference and local MT station (Ritter et al. , 1998). In recent years, alternative procedures have been proposed to obtain reliable estimates of the MT impedance tensor. In particular, Weckmann et al. (2005) proposed a pre-selection scheme of the Fourier transform coefficients followed by a robust processing, while Escalas et al. (2013) and Carbonari et al. (2017) proposed a polarization analysis of the MT signals in the wavelet transform domain. The use of the wavelet transform is motivated by the resolution that it provides in both time and frequency domain, thus allowing to deal with transient components of the MT signal, as the man-made noise usually occurs. In the present work, a different approach based on the use of Discrete Wavelet Transform (DWT) and Self-Organizing Map (SOM) neural network analysis (D’Auria et al. , 2015) is proposed for improving the magnetotelluric impedance tensor estimates. The approach has