Understanding the Non-Stationary Nature of Human-Induced Earthquakes and its Impact on Geothermal Energy Production

Anthropogenic activity in the subsurface causes stress perturbations which can lead to the onset of seismicity. One of the notorious examples is the Groningen gas field in the northeast part of the Netherlands which is among the largest in Europe. Hydrocarbons have been produced there since 1963 until the field’s ultimate shutdown in October 2023. From December 1991 until January 1^st, 2026, total of 1561 events have been recorded in this area, with magnitude ranging from  to . The  events caused extensive damage to buildings and quite a societal unrest as well as scepticism towards subsurface operations in general. Considering, it is crucial to identify an envelope for safe utilization of the subsurface to be able to continue its usage for energy transition while limiting the risk of induced seismicity.

To be able to limit the risk of seismicity from subsurface operations, it is necessary to understand the non-stationary nature of induced seismicity, meaning the underlying physical causes of the observed spatial and temporal variations in event locations and frequency-magnitude distribution. This research is based on the hypothesis that the fault spatial distribution and geometry (dip angle, offset) in conjunction with operational parameters (pressure history, rates, injection temperatures) are the causal processes of the temporal and spatial variations in the Gutenberg-Richter parameters.

I will present the results from modelling production induced seismicity using the Groningen field as a study area. The results include synthetic earthquake catalogues obtained by modelling the event nucleation and magnitudes using a semi-analytical approach of slip weakening faults. For this model, fault geometry and pressure history serve as input. In order to obtain multiple catalogues spanning the full range of uncertainty, a Monte Carlo sensitivity analysis is conducted for different reservoir and fault properties. Subsequently, several statistical comparison tests of the simulated catalogue with the observed seismicity allows us to derive posterior estimates for our properties and provide crucial insight into how we are doing solving the puzzle of what is causing the observed spatiotemporal behaviour of induced earthquakes.