Revisiting the Timpson Induced Earthquake Sequence With Deep-Learning

Session: Mechanisms of Induced Seismicity: Pressure Diffusion, Elastic Stressing and Aseismic Slip
Type: Oral
Date: 4/29/2020
Time: 09:15 AM
Room: 215 + 220
Description: 

On May 17th, 2012 a Mw 4.8 earthquake struck Timpson, eastern Texas. It is the largest known earthquake in this region. The Timpson earthquake sequence is associated with wastewater injection through two nearby, high-volume injection wells that started operation five years before the main shock. Many of the aftershocks are recorded by a temporary network installed in response to the Mw 4.8 event. Frohlich et. al. (2014) located the aftershocks occurring after May 26 using VELEST and found them to lie along a NW-SE trend, which agrees with a mapped basement fault. However, Lund Snee and Zoback (2016) observed N68^oE to N80^oE S_Hmax direction using borehole breakout measurements. This stress orientation is highly unfavorable for slip on the fault indicated by the aftershock locations while the auxiliary plane is well-oriented for slip. According to their stress model, the fault should only slip in the presence of high pore pressure and in the absence of well-oriented faults. To understand this discrepancy, we reanalyzed the Timpson earthquake sequence to refine the locations of the mainshock and aftershocks that occurred before local station were installed. We first apply PhaseNet, a deep-learning-based arrival time picker to get P and S wave picks. Those picks are associated using REAL, a grid-search based associator. The associated picks are used with VELEST to generate primary locations. We also add cross-correlation differential arrival time measurements as input in addition to the absolute phase arrival measurements to produce high-precision relative locations with HypoDD. Our preliminary results show that the refined locations support the NW-SE trending fault plane of Frohlich et al. (2014). Our next step is to run the detection, association and location workflow on continuous recordings to enrich the earthquake catalog and refine the fault structure. Potential explanations for the discrepancy might involve a well-oriented foreshock triggering the misoriented mainshock, as recently seen in the Ridgecrest sequence or local rotation of the stress field.

Presenting Author: Kaiwen Wang

Authors