Determining the Uncertainty Range of Coseismic Stress Drop of Large Earthquakes Using Finite Fault Inversion

Friday, 18 December 2015: 14:25
305 (Moscone South)
Mareike Adams, Chen Ji, Cedric Twardzik and Ralph J Archuleta, University of California Santa Barbara, Santa Barbara, CA, United States
A key component in understanding the physics of earthquakes is the resolution of the state of stress on the fault before, during and after the earthquake. A large earthquake’s average stress drop is the first order parameter for this task but is still poorly constrained, especially for intermediate and deep events. Classically, the average stress drop is estimated using the corner frequency of observed seismic data. However a simple slip distribution is implicitly assumed; this assumed distribution is often not appropriate for large earthquakes. The average stress drop can be calculated using the inverted finite fault slip model. However, conventional finite fault inversion methods do not directly invert for on-fault stress change; thus it is unclear whether models with significantly different stress drops can match the observations equally well. We developed a new nonlinear inversion to address this concern. The algorithm searches for the solution matching the observed seismic and geodetic data under the condition that the average stress drop is close to a pre-assigned value. We perform inversions with different pre-assigned stress drops to obtain the relationship between the average stress drop of the inverted slip model and the minimum waveform misfit. As an example, we use P and SH displacement waveforms recorded at teleseismic distances from the 2014 Mw 7.9 Rat Island intermediate depth earthquake to determine its average stress drop. Earth responses up to 2 Hz are calculated using an FK algorithm and the PREM velocity structure. Our preliminary analysis illustrates that with this new approach, we are able to define the lower bound of the average stress drop but fail to constrain its upper bound. The waveform misfit associated with the inverted model increases quickly as pre-assigned stress drop decreases from 3 MPa to 0.5 MPa. But the misfit varies negligibly when the pre-assigned stress drop increases from 4.0 MPa to 50 MPa. We notice that the fine-scale roughness of the inverted slip distributions changes dramatically. Future investigations using velocity records, which should be more sensitive to fine scale roughness on the fault surface, will be conducted.