Dynamic Rupture Modeling of Induced Earthquakes in Oklahoma

Since 2008, seismicity has increased in Oklahoma due to high-volume fluid injection. We examine the impact of pore pressure and tectonic factors on the magnitude and the potential for larger earthquakes Dynamic rupture models are performed on 10 km faults with varying input parameters. A neural network trained to approximate the results of the models predicts the moment magnitude of a rupture with 74% accuracy. Fault roughness and normal stresses are negatively correlated with moment magnitude, while the pore pressure has no correlation. We use the trained neural network with a Markov Chain Monte Carlo (MCMC) to find distributions of tectonic conditions. Using the inverse method on induced seismicity data from the Fairview earthquake sequence provides information about frictional parameters and fault geometry that cannot be found using other methods. The results suggest that induced earthquakes are tectonic earthquakes that have been advanced in time, rather than new events