Group B, Poster #036, Seismology

Evaluating Relative Relocation Methods in a Synthetic Setting: The Case of the 2019 Ridgecrest Earthquake Sequence

Yifan Yu, William L. Ellsworth, & Gregory C. Beroza
Poster Image: 

Poster Presentation

2023 SCEC Annual Meeting, Poster #036, SCEC Contribution #13103 VIEW PDF
Relative earthquake relocation algorithms, such as HypoDD (Waldhauser & Ellsworth, 2000) and GrowClust (Trugman & Shearer, 2017), have emerged as crucial tools for enhancing the accuracy and resolution of earthquake catalogs by reducing earthquake location errors. Although these methods employ distinct approaches to minimize different norms of misfits, both have proven effective in resolving fault structures through improved hypocentral determination. The advent of machine learning-based workflows for building earthquake catalogs, and the dramatic increase in the number of detected earthquakes that results, raises questions when two methods give different locations. In this study we ...use a controlled experiment to explore differences in these location methods with synthetic arrival times based on the setting of the 2019 Ridgecrest earthquake sequence. We calculated travel times using the Fast Marching Method, with synthetic hypocenters randomly shuffled relative to the SCEDC catalog. The 3D velocity model is extracted from the SCEC Community Velocity Model with a von Karman model superimposed to represent unmodeled small scale structure. We use the two relocation methods with a 1D velocity model to recover the hypocenters. Our preliminary results indicate that station coverage significantly influences the performance of both methods, particularly in terms of depth accuracy. Our results suggest that HypoDD displays greater resiliency to both unmodelled Earth structure and incorrect initial absolute locations, while GrowClust demonstrates higher computational efficiency and robustness when dealing with poorly constrained events.