Exploring finite-slip inversion with near-field seismic data: Analysis of the 2004 Parkfield earthquake, California
Abstract
Most common finite-slip inversion schemes are model-searching techniques conducted in the time domain. In this study, a frequency-based inversion method is proposed. It makes fewer assumptions than most current methods. Due to spatial-temporal trade-offs in the time domain, rupture velocity can bias the estimation when it is considered constant. Our approach does not prescribe the rupture velocity and slip rate functions. The model is also over-parameterize to avoid bias from too few unknowns. By dividing the fault plane into subfault patches, slip rate spectra and recorded spectra are formulated into a linear relationship. To provide good spatial resolution (~2 km), slip rate spectra are fit up to 3 Hz. Our objective function and constraints are convex, thus the global minimum can be found by convex optimization, which assures our solution is always unique. Physically plausible regularization constraints are used to stabilize the inversion and suppress the errors introduced by a simplified seismic velocity structure. Based on a Green's function sensitivity analysis, the model is also weighted during inversion to overcome spatially uneven resolution due to limited data coverage. After obtaining the spectrum for each fault patch, results are inverse Fourier transformed into the time domain. To demonstrate our approach, we analyze the well-studied 2004 Parkfield earthquake and compare our results to previous slip models and dynamic simulation results. Ultimately, we hope to achieve an improved understanding of the resolution limits and uncertainties in kinematic source inversions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2013
- Bibcode:
- 2013AGUFM.S43A2486F
- Keywords:
-
- 7215 SEISMOLOGY Earthquake source observations