A systematic study of seismic moment tensor inversions using a 3-D structural model: application to Central and Southern Mexico

Abstract:

Advances in seismic imaging techniques such as ambient noise tomography and full-waveform inversion (FWI) have allowed to obtain increasingly accurate regional 3D tomographic Earth models. It is therefore now feasible to use more realistic 3D regional models to retrieve seismic source parameters with higher accuracy, accounting for crustal and upper mantle heterogeneities. Nevertheless, the use of 1D instead of 3D models is employed routinely to constrain seismic moment tensor solutions, due to its efficiency and near real-time application even though results may be affected by small-scale heterogeneities.

We develop a systematic study to quantify the impact of surface topography and 3D regional velocity models on seismic moment tensor inversions. We precompute three different sets of receiver side strain Green's tensors (SGT) in time and space. The first SGT data set is computed using a discrete wave number (AXITRA) code of Cotton and Coutant [1997] for a layer media. The second and third are computed by spectral-element simulations (SPECFEM3D), employing a layered media with surface topography and a new 3D tomographic model of Central and Southern Mexico. We perform a series of inversions for every earthquake in an assembled local earthquake catalog, using body waves and surface waves for a grid search over the moment tensors and source depths to minimize the waveform misfit. We compare the solutions between the three seismic moment tensor data sets and study the advantages of using a 3D structural model. Based on our analysis, we recommend accounting for a 3D tomographic model to yield more accurate seismic moment tensor solutions. Finally, the ultimate goal for this research is an automated seismic moment tensor inversion workflow for Mexico.