A frequency-domain parameter-reduction framework for Bayesian coseismic slip inversion: Application to the 1973 M_S7.5 Luhuo (China) earthquake

The Bayesian inversion framework is a powerful framework for solving nonlinear inversion problems, but its application to fault slip inversion is often hindered by high computational costs. In this study, we propose a new approach that approximates the spatial-domain slip distribution using only its low-frequency components in the frequency-domain. This strategy substantially reduces the number of model parameters under the physically reasonable assumption that coseismic slip distributions are smooth and dominated by low-frequency features. We evaluate the method using three synthetic fault slip models with different characteristics, demonstrating that it achieves high data-fitting accuracy with markedly improved computational efficiency. When applied to the 1973 M_S7.5 Luhuo (China) earthquake, our approach yields a coseismic slip model that broadly agrees with previous studies in its first-order features, while differences in small-scale details likely reflect variations in the dataset used. This work demonstrates an efficient and robust strategy for Bayesian coseismic slip inversion, enabling broader applications in earthquake source studies.