Studying the San Andreas-Garlock branch fault system by applying a machine learning approach driven by an earthquake simulator

Abhijit Ghosh, Shankho Niyogi, Evan O. Marschall, Roby Douilly, & David D. Oglesby

Published September 8, 2024, SCEC Contribution #13952, 2024 SCEC Annual Meeting Poster #150

Determining the direction of rupture propagation, or whether it would propagate at all, in a branch fault system is challenging. This task involves understanding the evolution of several critical parameters like stress, strength, and friction at different parts of the branch fault system and their relative importance in rupture propagation. In this study, we have used a machine learning approach to investigate rupture propagation in a branch fault system driven by an earthquake simulator. We have selected the San Andreas and Garlock fault junction in southern California for this study. We use the Rate-State Earthquake Simulator (RSQSim) to produce a sizeable synthetic earthquake catalog containing thousands of earthquakes in this branch fault system. We use the Uniform California Earthquake Rupture Forecast, Version 3 (UCERF3) for realistic input parameters, including fault geometry. In the next step, we use the model results to train our machine learning model. We use simulation results, including ~9800 suitable simulated events, and employ gradient boosting and random forest algorithms to investigate the distribution of parameters related to specific rupture scenarios. We demonstrate that the machine learning model can determine the direction of rupture propagation with reasonable precision and recall with enough secondary and primary input features. In addition, we identify the parameters critical to determining rupture propagation and demonstrate that the conditions at the fault branch where the earthquake is nucleated are the most crucial in determining the path of rupture propagation. Overall, this innovative approach provides a promising tool to shed new light on the complex dynamics of fault systems and their rupture propagation.

Key Words
San Andreas, Garlock, branch, rupture, machine learning

Citation
Ghosh, A., Niyogi, S., Marschall, E. O., Douilly, R., & Oglesby, D. D. (2024, 09). Studying the San Andreas-Garlock branch fault system by applying a machine learning approach driven by an earthquake simulator. Poster Presentation at 2024 SCEC Annual Meeting.

Related Projects & Working Groups
Fault and Rupture Mechanics (FARM)