GNGTS 2013 - Atti del 32° Convegno Nazionale

Starting from the previous 1D Q s Earth model we performed the 3D inversion using the program SIMUL2000 (Thurber, 1993; Eberhart-Phillips, 1993, Thurber and Eberhart-Phillips, 1999), starting from the values of k. The code applies an iterative damped least squares method. In the inversion, as suggested in SIMUL2000 code, the earthquakes hypocenters and origin times were kept fixed. The Q s crustal values are inverted using a damping value of 0.002, chosen evaluating the trade-off curve of the data variance versus model variance achieved after single iterations (Eberhart-Phillips, 1986, 1993). Results. A refined estimate on the data reliability was performed through the comparison between the checkerboard test and a spread function test (Michelini and McEvilly, 1991).We performed the checkerboard test with a mean Q s of 200 with variations of ±40%; we added to synthetic t* a noise comparable to the residual level of real data. We choose such a value because it’s near to the RMS obtained by the genetic run (Chen and Clayton, 2012). Fig. 2 shows the shows the checkerboard test, for depth 4-7 km, superimposed an isoline of the spread function S value of 7. Fig.3 – Vertical cross sections 3-5 of the 3D Q S distribution. 73 GNGTS 2013 S essione 1.1