Extending Seismic Tomography along the San Andreas Fault to the Lower Crust with Low Frequency Earthquakes

Similarities within families of low frequency earthquakes (LFE's) occurring within non-volcanic tremor (NVT) beneath the San Andreas fault (SAF) in central California facilitates applying high-precision location techniques to the LFE's and tomographic imaging of the tremor zone. In turn, this will allow us to examine the geometry and character of the SAF deep in the crust and evaluate the lithologies and physical conditions (e.g., fluid content) surrounding the tremor zone. We build on the work of Shelly and coworkers (Shelly et al., 2009; Shelly and Hardebeck, 2010) to stack and pick LFE "tremorgrams" for temporary array stations and other stations not previously analyzed. Our initial work focused on the 2001-2002 Parkfield Area Seismic Observatory (PASO) array and 15 LFE families directly beneath it. Augmenting our existing PASO earthquake and explosion dataset with the LFE picks allows us to extend our PASO tomographic model deeper to include the tremor zone, where we find slightly reduced Vp and more sharply reduced Vs near the LFE locations. We are now expanding our work to include PASO records of more distant LFE families and other seismic stations. We find that the high amplitudes and more frequent recurrence of LFE's to the southeast of PASO results in high quality stacks for most PASO stations. We can also produce good stacks for weaker, less frequent LFE's northwest of PASO. We will present examples of our new tremorgrams along with preliminary LFE relocations. There are three main underlying goals for this project. The first is to extend the existing Vp model and develop the new Vs model to cover the depth range of the NVT present beneath the SAF in the Parkfield region. The presence of ambient and triggered NVT in this area has mainly been attributed to the presence of fluids (Ghosh et al., 2008; Peng et al., 2008, 2009; Nadeau and Guilhem, 2009; Thomas et al., 2011; Hill et al., 2013). The velocity models we will develop will also help constrain the lithologies and conditions present in the NVT zone. Our second goal is to locate all well-recorded LFE families in the region using new data from the multiple networks and arrays. Previous studies have yielded different results for the depths of LFE's and NVT, their degree of spatial clustering, and their epicenters relative to the surface trace of the SAF. These differing results lead to different conclusions about the origin and nature of SAF NVT. A third goal is to develop a 3D regional-scale Vs model that can be used for determining improved earthquake locations and for carrying out waveform modeling and geodetic calculations. For example, strong motion modeling of the 2004 Parkfield earthquake can be done more accurately with a 3D Vs model, and a reasonable 3D Vs model can be a starting point for spectral element waveform inversion. Additionally, a 3D model of elastic moduli derived from Vp and Vs can be used for finite-element modeling of geodetic data covering the Parkfield earthquake cycle.