GNGTS 2019 - Atti del 38° Convegno Nazionale

GNGTS 2019 S essione 3.3 763 interfaces separating a top water layer, a sedimentary layer, a thick (ca 600-m) basaltic bed, followed by three sedimentary strata, for a total model depth of ca 4500 m. This model is designed to cause an overlap at a near-zero offset of the top basalt reflection with the first sea- bottom multiple at about 3 s. Note the high attenuation value (α=0.91) ascribed to the basalt layer. The seismogram are generated by propagating a 20-Hz central frequency wavelet and a 2-ms sample rate. Fig. 2 shows the simulated seismogram with the addition of Gaussian noise. The black arrows indicate the time location of the simulated sub-basalt reflections. Figure 2 compares the velocity spectra computed by the Semblance functional, by the CMA where the wavelet used as matching filter is a Ricker with central frequency of 15-Hz, and by the wavelet-based functional (WBF). Only the scales corresponding to 10-Hz, 15-Hz and 20- Hz are shown. The simulated case demonstrates that the Semblance is not able to detect the sub-basalt reflections. The CMA functional as well as the wavelet-based functional locate the sub-basalt reflections with a high degree of accuracy. The main difference between CMA and WBF is that the former requires a fixed set of parameters for the analysis of a seismogram and, in particular, it requires the estimation of a wavelet representing of the reflections of interest or the definition of a central frequency of a synthetic wavelet. Instead, WBF does not require any information a priori and it automatically scans the coherencies along common frequency panels. Comparison of the CMA spectrum with the WBF spectrum at 15-Hz (Fig. 2) shows that the latter is characterized by higher resolution and more energetic coherencies. It also suggests that the wavelet transform produces more efficient filtering (in this example I have employed a complex Morlet mother wavelet) with respect to the filtering produced by the Ricker wavelet. Fig. 2 - Left: synthetic seismogram with added noise. From left to right: velocity spectra computed by the Semblance, by the Complex Matched Semblance and by the Wavelet-based functional (only the scales corresponding to 20, 15 and 10-Hz are illustrated). The second example (see Fig. 3) is the application to a field data case pertaining to an off- shore acquisition. The sea-bottom is located at 1.5 s approximately and the top basalt reflection is around 3 s. No reflections are clearly evident in the seismogram below the top basalt. Figure 3 also shows that the velocity spectrum computed by the Semblance is no able to detect the sub-basalt primaries. By adopting a 15-Hz Ricker wavelet as matching filter, the CMA identi- fies the weak sub-basalt reflections highlighted by the occurrence of coherencies below 3 s in the velocity range between 2 and 3 km/s (see the CMA spectrum in Fig. 3). Again, WBF evi- dences its capability to explore the trial velocities over trial scales with no information a priori,. In this specific case, WBF clearly detects the sub-basalt reflections by considering common scale panels corresponding to 10-5-Hz.