Short-term forecasting of Mmax during hydraulic fracturing

Abstract

Previous studies of injection-induced earthquake sequences have shown that the maximum magnitude (Mmax) of injection-induced seismicity increases with the net injected volume (V); however, different proposed seismic-hazard paradigms predict significantly different values of Mmax. Using injection and seismicity data from two project areas in northeastern British Columbia, Canada, where hydraulic fracturing induced seismicity was observed, we test the predictive power and robustness of three existing and one novel method to estimate Mmax. Due to their vastly different values of seismogenic index (Σ), these two project areas represent end-member cases of seismogenic response. Our novel method progressively adjusts the Mmax forecast under the assumption that each recorded event embodies an incremental release of fluid-induced stress. The results indicate that our method typically provides the lowest upper bound of the tested methods and it is less sensitive to site-specific calibration parameters such as Σ. This makes the novel method appealing for operational earthquake forecasting schemes as a real-time mitigation strategy to manage the risks of induced seismicity.

Physics-informed deep learning to forecast $${\widehat{{\varvec{M}}}}_{{\varvec{m}}{\varvec{a}}{\varvec{x}}}$$ during hydraulic fracturing

“…Occurrence of HF-induced seismicity in the presence of a TLP show that these assumptions are not universally applicable 37 . Adaptive (or advanced) TLPs have been proposed 9 , 10 , 24 , 38 , but these methods require real-time access to stimulation data, such as the injection rate. Since this type of data is not always available in real time to an independent observer, we focus here on a purely data-driven approach.…”

“…In summary, our findings provide a proof-of-principle that a fixed-window PIDL approach could serve as a basis for an adaptive TLP system. While many studies have explored methods to forecast induced seismicity associated with industrial activity, including hydromechanical models that combine fluid pressurization and rate-and-state friction 40 or related mechanisms 41 , 42 , models to predict maximum seismic magnitude using injection data 12 , 15 , 24 , 25 and machine learning models to forecast induced seismicity rates using highly related features 43 , these models require access to injection data and/or geomechanical parameters, e.g. poroelastic stress, stress rate and rate-state friction parameters, which are typically not available at all or at least not in real-time.…”

“…Li et al 24 developed a method for real-time operational forecasting of during stimulation. Their method estimates seismic efficiency ratio (SER), an empirically determined fraction of the maximum expected cumulative seismic moment based on injected volume 25 .…”

Induced Seismicity Controlled by Injected Hydraulic Energy: The Case Study of the EGS Soultz‐Sous‐Forêts Site

“…By introducing this new approach, they more accurately predict the total seismic moment associated with fluid injections, as well as the maximum possible magnitude. Li et al (2022) have revisited the approach of Hallo et al (2014) by considering that at any time during injection, the ratio between the predicted and the observed cumulative seismic moment constitutes a stored energy that can be released in a single large event. By using this method, they are able to update this ratio continuously during injection and deduce the maximum expected magnitude.…”

“…Li et al. (2022) have revisited the approach of Hallo et al. (2014) by considering that at any time during injection, the ratio between the predicted and the observed cumulative seismic moment constitutes a stored energy that can be released in a single large event.…”