This paper suggests utilizing the trust region reflective (TRR) algorithm to address the inverse problem related to seismic spectrum analysis of Pn waves. By applying this approach, it becomes feasible to simultaneously deduce the seismic moment, corner frequency, tomographic model of Pn-wave attenuation, and site responses while incorporating appropriate constraints based on prior knowledge to ensure solution accuracy. The efficacy of this method has been validated using synthetic data. Specifically, the recorded Pn waves from four underground explosions in North Korea were employed to evaluate the performance of the proposed technique. The derived seismic moment and corner frequency for these events were found to align with the characteristics of the seismic data, earthquake magnitude calculations, and empirical data. The resulting tomographic model of Pn-wave attenuation revealed a spatial distribution of high and low attenuation, indicating significant lithospheric heterogeneity in northeastern China. Areas with low Q0 and low η, such as the Xialiaohe Basin and southwestern Changbai Mountain, suggest a high-temperature environment in the upper mantle cap layer or the presence of molten substances. Conversely, regions with high Q0 and high η, like the northern Changbai Mountain and the vicinity of the eastern Tanlu Fault Zone, characterized by high Pn-wave velocity and thick crust, indicate minimal modification or destruction in the lithosphere. Furthermore, site responses were determined for 75 seismic stations in northeastern China, with their characteristics preliminarily analyzed and interpreted in conjunction with geological context. The iterative process based on the TRR algorithm for Pn-wave spectral inversion proposed in this study demonstrates robust convergence and enhances the fitting accuracy of the observed spectrum. The inversion outcomes from this spectral technique yielded synthetic Pn-wave spectra that closely matched the observed spectra of the four underground explosions in North Korea across various frequency bands for over 90 % of the stations.
Read full abstract