Simulations for the development of a ground motion model for induced seismicity in the Groningen gas field, The Netherlands

Abstract

We present simulations performed for the development of a ground motion model for induced earthquakes in the Groningen gas field. The largest recorded event, with M3.5, occurred in 2012 and, more recently, a M3.4 event in 2018 led to recorded ground accelerations exceeding 0.1 g. As part of an extensive hazard and risk study, it has been necessary to predict ground motions for scenario earthquakes up to M7. In order to achieve this, while accounting for the unique local geology, a range of simulations have been performed using both stochastic and full-waveform finite-difference simulations. Due to frequency limitations and lack of empirical calibration of the latter approach, input simulations for the ground motion model used in the hazard and risk analyses have been performed with a finite-fault stochastic method. However, in parallel, extensive studies using the finite-difference simulations have guided inputs and modelling considerations for these simulations. Three approaches are used: (1) the finite-fault stochastic method, (2) elastic point-and (3) finite-source 3D finite-difference simulations. We present a summary of the methods and their synthesis, including both amplitudes and durations within the context of the hazard and risk model. A unique form of wave-propagation with strong lateral focusing and defocusing is evident in both peak amplitudes and durations. The results clearly demonstrate the need for a locally derived ground motion model and the potential for reduction in aleatory variability in moving toward a path-specific fully non-ergodic model.

A Probabilistic Model to Evaluate Options for Mitigating Induced Seismic Risk

“…In view of the shallow focus of the Groningen earthquakes, the travel paths to the surface that traverse the high-velocity Zechstein salt layer (Figure 3), and the thick layers of soft deposits at the surface, it was considered necessary from the outset to develop a locally calibrated ground-motion prediction model (Bommer et al 2016). Source, path and site parameters for the reference rock horizon at 800 m depth were estimated from inversions of recorded motions and these were then used as input to finite-fault simulations of spectral accelerations (for periods from 0.01 to 5 s) and peak ground velocity for events from ML 2.5 to greater than 7 (Edwards et al 2018). The upper limit is determined by the Mmax distribution, whereas the lower limit of magnitude considered in the risk calculations depends on the desired risk metric, with lower values being more appropriate for damage estimation than those used for calculations in terms of casualties (Bommer and Crowley 2017).…”

Liquefaction Hazard in the Groningen Region of the Netherlands due to Induced Seismicity

“…Each of the distributed subfaults in this technique is assumed to be a point source (effectively a small magnitude earthquake) and can be characterized using the seismological parameters observed in events recorded in the Groningen gas field. This process is detailed by Bommer et al (2017b) and Edwards et al (2019) and briefly summarized in the following section.…”

Ground-motion networks in the Groningen field: usability and consistency of surface recordings

“…It must be noted that for applications making direct use of the Fourier spectra of the records, such as the inversions discussed by Bommer et al (2017b) and Edwards et al (2019), the usable frequency range of the records can be defined very differently. The window of usable frequencies in this case is selected such that there is a minimum signal-to-noise amplitude ratio of 3 in all frequencies within the window.…”

“…The basic framework of the GMM are equations for predicting motions at a rock horizon located at a depth of about 800 m combined with frequency-dependent non-linear amplification factors assigned to zones defined over the field as well as a 5-km onshore buffer zone (Bommer et al 2017b). The equations predicting accelerations at the reference rock horizon are obtained from regressions on the output from finite fault simulations (Edwards et al 2019) using source, path and site parameters obtained from inversion of the Fourier amplitude spectra of recorded motions in the field. While the extrapolation to the largest magnitudes currently considered in the hazard and risk calculations (see Bommer and van Elk 2017) is inevitably associated with large epistemic uncertainty, the model is well calibrated to local conditions by virtue of the database of recorded motions from the field.…”