Unified Regional Tomography and Source Moment Tensor Inversions Based on Finite-Difference Strain Green Tensor Databases

摘要

Seismic monitoring requires accurate source characterization in real time. Accurate 3D earth models are essential for accurate predictions of seismic observables and source characterization. While recovering the true earth structure has always been the goal of tomographic inversions, methods based on full wavefields in 3D reference models have several distinct advantages. Made possible by advances in high-performance computation, this full-wave approach accounts for complex wave propagation in 3D heterogeneous earth, enables fuller utilization of seismic records, and allows us to linearize the inverse problem by iteratively updating the 3D reference model. An important benefit of physically realistic and accurate modeling of wavefields in 3D models is the consistency of the system of equations in inversion. This is particularly important for the integration of different types of observations (P, S, and surface waves including empirical Green's functions derived from ambient noise) in inversion. Two methods have been developed in this field to carry out full-wave tomography iteratively with 3D reference models. One is the adjoint-wavefield (AW) method, which back- propagates the data from the receivers to image the structure. The other is the scattering-integral (SI) method, which constructs the strain Green tensor (SGT) databases and calculates sensitivity kernels for each data functional. Both methods are based on the full-wave theory, and the main differences are in the computational approaches. In general, the adjoint method is CPU intensive, while the SGT-database approach requires a large disk space and fast network. To date, the AW and SI methods use two different approaches in calculating synthetic waveforms, i.e. the spectral-element method and finite-difference (FD) method, respectively. One notable advantage of the SI method is that the SGT databases make it possible to use 3D synthetics in real time seismic monitoring.