GNGTS 2019 - Atti del 38° Convegno Nazionale

568 GNGTS 2019 S essione 3.1 Ground-Roll; the static corrections; the application of weights computed evaluating the spatial coherency by means of the cross-correlation to the data; finally, the use of a moderate prestack F-XY deconvolution. Presentation of the Data. The survey area is approximately 80 km 2 and the total number of recorded sources is 1474 with spacing equal to 80 m, while the source-line spacing is 500 m. The total number of employed receivers is 2664 and the receiver-line spacing is 480 m. The receiver interval is 50 m and the maximum number of active receivers per shot is 720. The offset ranges between 50 m and 3.8 km distributed in a split spread configuration. The recording length is 8 s adopting a sample rate of 2 ms. Fig. 1a shows an example of raw shot gather composed by five receiver lines. As can be noted, the seismogram is affected by a low S/N ratio, it is strongly contaminated by Ground-Roll and the occurrence of reflections is evident at approximately 2.5 s. Fig. 1b shows the same shot gather of Fig. 1a after the processing presented in the next section. Fig. 1 - a) example of shot gather before the processing and b) same shot gather after the processing sequence illustrated in Fig. 2. Seismic Data Processing. Fig. 2 illustrates the flow chart with the processing steps used to calculate the time stack volume. We adopt different colours to evidence in blue the steps adopted to reduce the noise, in gray the operations of static corrections, in green the steps used to enhance the coherency of the reflected events and in orange colour the computation and application of the weights used to improve reliability of the stack volume. The denoising applied is composed by a trace editing step such as top muting, despiking, and killing of the bad traces. In order to identify the noisy traces, we estimate the root mean square (RMS) amplitude on three temporal windows. All the traces that exhibit an anomalous decay were flagged as noisy and then removed. We killed more than 5000 traces. In order to reduce the surface waves we tested a simple band-pass filter (BP), a time and offset variant filtering (TOV), and a filtering in the continuous-wavelet domain (WF) (Tognarelli 2018a, 2018b). The outcomes produced by WF and TOV were very similar and we held the latter, after splitting the data into more than 10 offset ranges and time windows, applies to each of them a dedicated band-pass filter whose frequencies are chosen to adapt the changing characteristics of the Ground-Roll. A Predictive deconvolution is applied only to the near-offset traces to remove the residual Ground-Roll. The surface-consistent amplitude corrections consider the RMS amplitude computed in a time window of 1.5 s length and we corrected the term shot and receiver. The statics was corrected considering a velocity model with one refractor and a final datum of 200 m. Despite the relevant results provided by the static corrections, a strong effort was dedicated to the estimation and application of the surface consistent residual static corrections. Starting from the picking of the 3D layers observed on the stack volume, compute the time shifts after four iterations