GNGTS 2017 - 36° Convegno Nazionale

586 GNGTS 2017 S essione 3.1 wavelet spectra of three traces extracted from the gather of Fig. 1a and located at offset 20 m (Fig. 2a), 40 m (Fig. 2b) and 90 m (Fig. 2c) respectively. In red colour, the contour of Fig. 2 puts in evidence the time-frequency coordinates related to the stronger events observed in the time-offset domain (Fig. 1a). Analysing the wavelet spectra of Fig. 2, it is possible to localize both in frequency and in time the GR allowing to design a mute mask aimed at removing it. The mute masks applied for the spectra presented in Fig. 2 are highlighted in light red. Since the effects of the GR on the seismogram are offset dependent, different masks are required to adapt the characteristics of the GR. In this specific case, the offsets of the data are gathered into ten groups and a mask for each group is defined. The masks are drawn directly on the spectra by a user defined picking. Note that the spectrum in Fig. 2c exhibits the contamination of GR and of an high frequency noise that is included in the mute mask. In Fig. 1b is shown the same gather of Fig. 1a after the filtering in the wavelet domain and Fig. 1c illustrates the difference between them. As can be noted, the filtering produces a significant suppression of the surface waves. The improvement of the data quality can be also appreciated on the stack section (Fig. 3c) where the reflected events appear more energetic and continuous and where a general S/N enhancement can be observed. Fig. 3 show the raw stack section (Fig. 3a), the stack section with a traditional band- pass filter (15-25-130-150 Hz) applied, and the stack section obtained after filtering the data in the wavelet domain (Fig.3c). Conclusions. In this work is presented the application of the continuous wavelet transform aimed at attenuating the Ground Roll on seismic data. The results shown demonstrate that filtering in the wavelet domain produces an improved data quality if compared with traditional band-pass filtering. In addition, the wavelet spectrum provides a time-frequency representation of the data where the components related to the GR can be easily detected allowing to design a mute mask in a simple way. With respect to many traditional methods, the showed procedure is not time consuming, is not computationally expensive and it doesn’t generate artefacts. The continuous wavelet transform constitutes a powerful tool for analysing seismic data and its use can have many applications in the field of the seismic reflection. Acknowledgements The code used in this work is implemented in Matlab® (University of Pisa Campus Licence). The seismic data processing is carried out using the Promax® software of Landmark Graphics Corporation that is gratefully acknowledged. References Daubechies I.; 1990: The wavelet transform time-frequency localization and signal analysis . IEEE Transaction on Information Theory, 36 , 961-1005. DOI 10.1109/18.57199 Fig. 3 - Comparison between the raw stack section (a)), the stack section obtained after a band-pass filter (b)) and the stack section obtained filtering the data in the wavelet domain(c)). The sections are represented using the same amplitude range.