Empirical reference site and generalized inversion technique for seismic inversion in Yunnan, China: validation through stochastic simulation of the 2009 Ms 6.3 Yaoan earthquake

This study divides the Yunnan block in China into three regions based on the spatial distribution of historical earthquakes and active faults: Region A (Baoshan-Puer block), Region B (Western Central Yunnan block), and Region C (Eastern Central Yunnan block). These areas, situated `within the Sichuan-Yunnan rhombic block (SYRB) and its adjacent territories, are key seismic hotspots due to the interactions between the Eurasian and Indian plates. Given the difficulty in identifying traditional reference sites, we developed V_s30 velocity profile models for Yunnan Province using regional borehole data. Additionally, we established regional empirical reference site amplification models using the quarter-wavelength method. Using the generalized inversion technique (GIT), we performed joint inversions on 24, 40, and 40 earthquakes in Regions A, B, and C, respectively. Obtaining source parameters for 104 earthquakes, regional quality factors (Q) for the three regions, and local site amplification effects for 124 stations. The stress drop ranged from 0.20 to 6.94 MPa. The average stress drop in Region A (1.61 MPa) is greater than in Region B (1.10 MPa), and Region C (0.77 MPa). Low stress drop areas exhibited a strong spatial correlation with regions of high heat flow, suggesting that high heat flow areas may lead to lower stress drops. These results are consistent with previous studies. The quality factor Q models for Regions A, B, and C are 194.48f^0.418, 156.80f^0.537 and 382.66f^0.322, respectively. The Q value in Region C, near the Sichuan Basin, is significantly higher than in Region B, highlighting notable lateral heterogeneity. The resonant frequencies (f_res) of GMX-A, GMX-B, GMX-C, and GMX-D across 124 stations are 7.75, 6.20, 4.69, and 2.15Hz, with corresponding amplification factors of 3.02, 2.57, 6.62, and 6.50. The average amplification factors for GMX-A and GMX-B were similar, as were those for GMX-C and GMX-D. As the site conditions became softer, the peak amplitude plateau shifted to lower frequencies, consistent with the general observation that stiffer sites exhibit higher resonant frequencies. Finally, the parameters obtained from the GIT were used for stochastic finite-fault simulation of the 5% damped PSA, FAS, and acceleration time series of the 2009 M_s 6.3 Yaoan mainshock and two aftershock sequences. The simulation results were consistent with the observed results, validating the reasonableness of the inversion parameters for the Yunnan block.