Earthquake nucleation constrained from induced seismicity at Groningen: evidence that seasonal stress variations determine seismicity at the annual to multiannual time scale.

Mateo Acosta, Jonathan D. Smith, & Jean-Philippe Avouac

Submitted September 11, 2022, SCEC Contribution #12292, 2022 SCEC Annual Meeting Poster #204

Earthquake nucleation models can in principle be tested by analyzing the seismicity response to stress variations. We use this approach focusing on the case example of seismicity induced by gas extraction from the Groningen gas field. Stress-based seismicity forecasting models have previously been tested at annual to multi-annual timescales assuming either instantaneous nucleation or a nucleation process governed by rate- and-state friction. The success of both model types at hindcasting induced seismicity in Groningen indicates that a sub-annual resolution is needed to detect the effect of a possible nucleation process. However, estimating stress variations at short time scales and detecting eventual seismicity variations associated to such variations is challenging. To address this challenge, we developed a computationally efficient approach to model reservoir pressure and to forecast induced seismicity at sub-annual resolution. The reservoir model combines 1) a vertically integrated fluid diffusion model to resolve the pressure over the reservoir using the historic extraction data, and 2) an analytical poroelastic model to calculate Coulomb stress changes within and outside the reservoir due to reservoir compaction and pore pressure variations. The earthquake nucleation process is modeled using Rate-and-State friction. The model’s optimal parameters are determined using a MCMC inversion procedure. The model fits remarkably well the evolution of seismicity over the 30 years of gas production at Groningen. In addition, due to the seasonal variations of extraction, it predicts seasonal variations of seismicity which are consistent in amplitude and phase with observed seismicity. Because of the non-linearity of the seismicity model, the seismicity at the multiannual time scale is very sensitive to seasonal stress fluctuations. It is therefore incorrect to calibrate a seismicity model by ignoring stress variations at the sub-annual time scale. The forecasting model would be biased, and although it can perform well at the yearly time scale, assuming stationary seasonal variations, it would not correctly predict the seismicity response to sub annual variations of stress. By contrast, stress variations at the time scale less than the typical duration of earthquake nucleation have no effect and can be smoothed out.

Key Words
Short-term forecasting; Induced Seismicity; Seasonal loading

Citation
Acosta, M., Smith, J. D., & Avouac, J. (2022, 09). Earthquake nucleation constrained from induced seismicity at Groningen: evidence that seasonal stress variations determine seismicity at the annual to multiannual time scale. . Poster Presentation at 2022 SCEC Annual Meeting.

Related Projects & Working Groups
Earthquake Forecasting and Predictability (EFP)