GNGTS 2017 - 36° Convegno Nazionale

GNGTS 2017 S essione 3.2 633 frequency band is selected as reference and inverted using a Monte Carlo Inversion (MCI) (Socco and Boiero, 2008). In the inversion, layer thickness, S-wave velocity and Poisson�� ����� ’s ratio are randomly sampled from wide uniform distributions. This allows a VS model to be estimated without bias on Poisson�� ����� ������ ����� �� � ������������ ��� ��� ������� ������������ ��� ’� ����� ������ ����� �� � ������������ ��� ��� ������� ������������ ��� s ratio values which is a prerequisite for the adopted methodology. The reference W/D relationship is then computed and used to determine the time-average VS of all profiles in the cluster. By exploiting the W/D relationship sensitivity to Poisson�� ����� ��� ����� ’� ����� ��� ����� s ratio the time- average VP of the whole cluster is then determined. Results. The data have been acquired during a student-team work at a test site located at CNR headquarter in Torino. A squared area (5x5m) was excavated and filled with uniform non-compacted sand at the site. The bottom of the sand volume goes down to 2.5 m depth. A seismic line was acquired with 72, 4.5 Hz vertical geophones with 0.3 m spacing and with multiple shots by a hammer source. The data were recorded using three Geode modules with 512 ms record duration and 0.125 ms sampling interval. The data were processed in f -k domain using a moving window to estimate 13 DCs along the seismic line (Teodor et al. , 2017). Fig. 1a shows the DCs numbered according to their position along the line and with different colors to highlight curves located inside the sand (green), outside the sand (blue) or close to the sand boundaries (red). The clustering method produced the dendrogram reported in Fig. 1b. The horizontal axis of the dendrogram reports the DC numbers and the vertical axis reports the distance between linked clusters. The higher the vertical nodes the further the clusters are from each other. The dendrogram shows two major clusters corresponding to the curves that are inside and outside the sand body. For the two curves close to the boundaries it shows that curve 10 could be considered as a single cluster (outlier). Curve 4 is connected to its cluster with a very high distance. Curves 4 and 10 are then considered outliers that should be processed individually and they were not included in the velocity model estimation. A DC from each cluster was used to estimate the VS model through MCI and to obtain the reference W/D. The application of the W/D to the other DCs of the cluster provided the time-average VS models reported in Fig. 2a. The Poisson�� ����� �������� ���� ��� ��� ������������ �� ���� ������� ��� ���� ������� �� ��� ’� ����� �������� ���� ��� ��� ������������ �� ���� ������� ��� ���� ������� �� ��� s ratio estimate from the W/D relationship of each cluster was then applied to the VS model to transform it into a time-average VP model shown in Fig. 2b. Estimated time-average VS and VP clearly show a low velocity zone in the middle of the profile related to the presence of the sand body. This low velocity zone is partially laterally continued outside of the sand body in the left part of the sand body. This is due to the limited Fig. 1 - a) the dispersion curves along the seismic line from CNR dataset b) the dendrogram showing the cluster system of the CNR’s dispersion curves.