Fault Zone Imaging: Past, Recent, Future

Session: Seismic Imaging of Fault Zones
Type: Oral
Date: 4/29/2020
Time: 03:30 PM
Room: 215 + 220
Description: 

Many early local earthquake tomography studies targeted fault zones due to abundant earthquake sources. One-step inversions for Vp in the 1970s gave way to iterative solutions in the 1980s once efficient methods for determining ray paths and travel times were developed. As 3-component seismic data became more prevalent, joint inversions for Vp and Vs or Vp and Vp/Vs became more common. Typical findings for fault zones are large velocity differences across the fault, reduced velocity centered on the fault, and along-strike variations correlated with fault behavior. Inspired by the DD location method, the development of DD tomography in the 2000s was a major step forward. Its development was timely, as its use was critical for defining the SAFOD borehole trajectory to reach the San Andreas fault zone within 100 m of the target repeating earthquakes. Recently, much effort has gone into joint inversions to reduce model nonuniqueness. Example combinations include seismic and gravity data, seismic and MT data and body-wave and surface-wave data. New joint inversion methods need to be developed, for example body-wave, surface-wave and trapped-wave joint inversions. Greater effort to jointly model passive and active source data is desirable. One example is joint inversion of earthquake and explosion first arrival times and reflected wave travel times for 3D velocity structure, reflector geometry and earthquake locations, work I am pursuing in New Zealand. There are other important directions for future research. More work is needed on flexible or adaptive model parameterization for constructing multi-scale models, especially for joint inversions. In the longer term, more dense array studies are needed to push resolution to finer scale. Machine learning methods can help deal with massive datasets. A great challenge is full waveform tomography of faults zones, for which accurate sensitivity kernels need an initial model that can generate all the important wave arrivals.

Presenting Author: Clifford H. Thurber

Authors