Post shut-in hazard for hydraulic-fracturing-induced earthquakes: analysis using data from the Guy-Greenbrier earthquake sequence

This project evaluates the existing statistical models to describe post-shut-in seismicity for hydraulic-fracturing-induced earthquakes and studies the importance of post-shut-in seismicity on decision-making. We focus on the short-term hazard based on the seismicity during and after the injection. We consider the Omori model by Langenbruch and Shapiro (Geophysics 75(6):MA53-MA62, 2010), the exponential model, and the stretched exponential model from Mignan et al. (Sci Rep 7(1):1-10, 2017). In particular, we evaluate their performance on nine earthquake clusters that occurred in 2010 near the Guy-Greenbrier fault in Arkansas (using data from Yoon et al., Solid Earth 122(11):9253-9274, 2017). While many of the post-shut-in sequences could be described by a single decay process, there is an increase in the post-shut-in seismicity in some clusters. Results show that the Omori model performs the best for the former case, while the stretched exponential model could capture the latter situations. We then use the Omori model to explore the effect of shut-in timing on the short-term hazard. Results show that the post-shut-in seismicity could affect the decision significantly for a slower decay in seismicity and longer injection duration. We perform a sensitivity analysis considering the uncertainties in the Omori model and the Gutenberg-Richter distribution. Results show that their relative importance depends on the injection duration and intensity thresholds of interest. Finally, we propose a logic tree model to incorporate the uncertainty in model selection and parameter estimation. The logic tree assembles the Omori model and the stretched exponential model to consider the possibility of increasing post-shut-in seismic hazard. We also show that each branch's weight could be updated in a Bayesian manner with new data.