GNGTS 2019 - Atti del 38° Convegno Nazionale

618 GNGTS 2019 S essione 3.2 does not include a proper inversion but phase slownesses (reciprocals of velocities) are directly computed from an average of the phase propagation time gradient over different shots. Tests were performed both using synthetic and real data. Results with synthetic data. A test phase velocity distribution for one single mode and one frequency (f = 5 Hz) was defined . A realistic acquisition geometry, including 10 source lines and 10 receiver lines, for a total number of 110 sources and 1010 receivers, was then designed. Travel times between source-receiver couples were computed using the straight rays approximation. Different percentages of Gaussian noise were added to the data. Finally, an offset selection was applied. For each of the three algorithms described above, several tests were performed in order to find the best parametrization for the given acquisition geometry. Then, the three algorithms were run using data with different noise levels. FMST and RJMCMC demonstrated to be robust and not affected by problems related to heterogeneous acquisition geometries, typical of active data. Eikonal tomography, instead, is sensitive to coarse spatial sampling in one direction and to the presence of noise. However, both RJMCMC and FMST need much longer computational time compared to Eikonal tomography, which still makes the latter an interesting tool. Results with real data. A real 3D dataset, originally acquired for exploration purposes in the Po plain, has been used to test surface wave tomography. The acquisition geometry covers an area of about 7 km x 12 km, with 3777 sources and 5551 receivers. The active receiver spread for each shot is formed by a maximum of 6 lines, each with 96 geophones. Most of the data have been acquired using dynamite, while only 18% of the shot positions were energized by vibroseis: considering the poor low frequency content of the latter data, only dynamite shots were used for surface wave analysis. First, we needed to extract the phase of the signal from the time domain records, which is not a trivial problem for active seismic data: heterogeneous spatial sampling (dense receiver spacing along cables, much wider spacing between cables) often generates cycle skipping when unwrapping the phase of the signal, especially in the cross-line direction; in addition, higher modes have to be filtered out in order to work only with the fundamental mode. In this work we analysed the signal at 2 Hz, where the surface wave propagation is dominated by one single mode (likely, the fundamental mode). A linear moveout correction (LMO) is an important processing step before unwrapping the phase in two dimensions, since it flattens the phases and dramatically reduces phase jumps. However, the choice of an LMO velocity representative of the whole investigation area is quite challenging. For this reason, we performed a shot by shot velocity analysis to extract a 2D LMO velocity distribution: the approach consists in scanning over different LMO velocities and selecting the one minimizing the phase variance (i.e. better flattening the events). Then, we performed tomographical tests using phases extracted with the 2D LMO velocity distribution described above. Models resulting from FMST, RJMCMC and Eikonal tomography runs, together with the 2D LMO velocity plot, were finally compared: despite small differences in the overall velocity values, the different methods highlight the presence of the same high/low velocity zones. The good correspondence between the different methods gives us confidence in interpreting the retrieved spatial velocity variations “seen” by the selected frequency as real geological structures. References Bodin T. and Sambridge M.; 2009: Seismic tomography with the reversible jump algorithm . Geophys. J. Int., 178 , 1411–1436. Boschi L. and Dziewonski A. M.; 1999: High- and low-resolution images of the Earth’s mantle: Implications of different approaches to tomographic modelling . J. Geophys. Res., 104 , no. B11, 25,567-25,594. Duret F., Bertin F., Garceran K., Sternfels R., Bardainne T., Deladerriere N. and Le Meur D.; 2016: Near-surface velocity modeling using a combined inversion of surface and refracted P-waves . The Leading Edge, 35 , no. 11, p. 926-1008.