GNGTS 2013 - Atti del 32° Convegno Nazionale

Attenuation tomography in Friuli Venezia Giulia region. To invert the k data we used the same grid used by Bressan et al. (2012) in 3D tomographic inversion of velocity in the area. The grid extends 114 km in E-W direction and 55 km in N-S direction. The grid centre has latitude 46.33 N and longitude and 13.08 E. The W-E grid nodes are at X: -60, -50, -35, -25, -15, -7, 0, 7, 15, 25, 36, 45, 54 km; the S-N grid nodes are at Y: -35, -20, -10, -5, 0, 5, 10, 20 km; the Z nodes are at depth 0, 2, 4, 6, 8, 10, 12, 15, 22 km with a layer at negative 3 km depth to account for the Earth’s topography. The X-axis is positive to the east, the Y-axis is positive to the north. The velocity of S waves in the medium is parameterized by assigning velocity values obtained in Bressan et al. (2012) paper at the nodes of the grid. Bressan et al. used 394 events from 1988 to 2004 with duration magnitude between 1.4 and 5.1 and performed an iterative simultaneous inversion of hypocentral parameters and 3D velocity structure with a damped least squares technique using a previous version of the code SIMULPS2000, i.e. SIMULPS12 (Evans et al. 1994). We used 156 events from 1994 to 2011 with duration magnitude between 3.0 and 5.7 for a total of 980 3-D records. The smaller number of earthquakes analyzed in this study is due on the constraints on the minimum magnitude we selected. Like in tomographic applications in which velocity is inverted, an adequate starting model for the inverted parameter is important before attenuation tomographic inversion, due to the non-uniqueness of the solution of linearized inverse problem. A wrong starting model can lead to blunders and biases in the result. No previous knowledge for a valuation of the Q s field in the Friuli area are available; so for finding the 1D starting crustal model a search heuristic was employed, that does not use linearization method. We adopted a genetic algorithm technique (Goldberg, 1989). In particular, the David L. Carroll GAFORTRAN code with MICRO-GA enabled (Carroll, 2001). A micropopulation of five individuals was utilized using as cost function the same SIMULPS2000 code minimizing the total weighted RMS; 600 generations were generated. Fig. 2 – Checkerboard test for depth 4-10 km. Circle size correspond to the percentage difference respect to the mean value. 72 GNGTS 2013 S essione 1.1