A spectral-ratio-constrained joint inversion of source parameters and attenuation

Strong trade-offs between earthquake source and attenuation term remain a major challenge in source parameters inversion and attenuation structure. Spectral ratio methods alleviate this problem by using nearby small earthquakes with highly correlated waveforms as empirical Green’s functions (EGF), thereby reducing path and site effects and enabling robust relative estimation of source parameters, particularly corner frequency. However, limited signal-to-noise ratios and spikes at high frequencies significantly affect the estimation of corner frequency. In addition, different choices of EGF may further increase the uncertainty in corner frequency estimations. To reduce the effects of high-frequency spectral instability and EGF selection on spectral ratios, we first perform single-spectrum fitting to obtain physically constrained and smoothed amplitude spectra. These fitted spectra are then used to construct spectral ratios, from which corner frequencies can be robustly estimated. The source parameters constrained by the spectral ratio analysis are then incorporated as prior information, with the introduction of controlled perturbations, a joint inversion of the source parameters (M₀ and f_c) and the attenuation factor t* is carried out using single spectra fitting. We applied this method to earthquakes that occurred in the southern Sichuan Basin. We applied this method to 257 earthquakes with magnitudes ≥1.5 recorded in the Weiyuan area of the southern Sichuan Basin, China, between November 2015 and November 2016. Seismic moments and corner frequencies are obtained through the combined use of spectral ratio analysis and single spectral fitting, from which stress drops are estimated assuming a circular crack model. The resulting t* measurements are subsequently used to invert for the regional attenuation structure, providing an independent evaluation for the robustness of the inferred source parameters. This study was supported by the National Natural Science Foundation of China (42474084).