A GIT algorithm for simultaneous estimation of seismic source, site response and regional-distance dependent attenuation parameters: application to synthetic and real data

Abstract

The generalized inversion technique (GIT) of earthquake recordings is a useful tool in retrieving seismic source spectra, attenuation path parameters and site transfer functions (TF). Further understanding of these three fundamental factors which control seismic ground motion constitutes one of the main challenges in seismology. Especially, the site TF estimation is an important objective that can significantly contribute to seismic hazard assessment. In this study, a parametric GIT algorithm (in MATLAB (The MathWorks, Inc., Natick, Massachusetts, United States, 2017)), based on the one proposed by Drouet et al. (Bull Seism Soc Am 98:198–219, 2008a), is developed, by introducing distance- and regional-dependent attenuation parameters for geometrical spreading and anelastic attenuation terms, respectively. This step is aimed at a more detailed investigation of attenuation path, anticipating to improve the knowledge of all three fundamental factors controlling seismic ground motion. The algorithm is based on a Gauss-Newton iterative inversion method, using initial reasonable model parameters. Source term is parametrically investigated for seismic moment and corner frequency, which can be simultaneously controlled with respect to stress drop. A synthetic dataset, approximating a simplified real dataset was inverted by the proposed inversion algorithm examining its computational validity. Four tests were implemented, with or without reference condition, providing encouraging results. The applicability of the algorithm is supported by the inversion of a real dataset which was also examined by Grendas et al. (Bull Earthq Eng 16:5061–5094, 2018) using a uniform attenuation model in GIT. Approximately 9% reduction of the misfit between real and computed data from the inverted model data is achieved, showing the effectiveness of the proposed algorithm.