Modeling Induced Seismicity on Dormant Faults: Application to the Groningen Gas Field

Abstract:

Constitutive laws for earthquake production based on rate-and-state friction, e.g., the Dieterich 1994 theory, generally make the fundamental assumption that the crust is actively producing seismic events at a constant rate. However, many regions of induced seismicity are in relatively stable areas, where stressing rates are low and have no or very little historical activity at the start of injection or production. The Groningen gas field in the Netherlands is the prime example of such a system, where gas production and significant compaction of the field occurred over 30 years before the onset of any detectable seismic activity. Groningen challenges the validity of the assumption of a constant background seismicity rate and calls for a physical mechanism that can explain the delay a possible observed delay in the onset of seismicity. Here we propose a new constitutive law where we assume the seismic sources in the crust are initially well-below steady state. In other words, the sources are gradually healing, and no earthquakes are produced initially. This model predicts a simple stress threshold condition for activating the seismic source, and once the critical stress is reached, the seismic sources begin to accelerate towards instability. We argue that the stress threshold can be approximated spatially a constant for a specific region due to logarithmic dependence on fault parameters that suppresses variability. We test this new model on the Groningen gas field data set using a well-constrained compaction model to compute stress changes in time and space. We compare the new threshold model to the Dieterich 1994 model, and a standard Coulomb failure model with an optimized initial stress distribution, finding that the new model substantially outperforms these alternative formulations. The model correctly simulates the time-evolution of seismicity since the onset of production, including the initial lag and the recent decrease of seismicity that followed the decline of production after 2014. We suggest that the commonly observed lag between the start of injection/production until the onset of induced seismicity is entirely consistent with under-stressed faults hosting inactive seismic sources.