Physics-Based Forecasting of Induced Seismicity at Groningen Gas Field, The Netherlands: Post Hoc Evaluation and Forecast Update

Induced seismicity due to natural gas extraction from the Groningen reservoir, The Netherlands, has been occurring since the early 1990s. Prospective forecasts of the pos-sible maximum earthquake magnitude for different gas extraction scenarios have pre-viously been published. However, their accuracy and continued relevance are unknown. Here, we evaluate our prior Groningen seismicity forecast and propose a revised model. We compare the forecast to earthquakes that occurred after its publication, between 1 February 2017 and 1 July 2022. We used volume scaling to account for actual gas extrac-tion being 35% less than considered in the forecast. After this correction, we found that the forecast slightly underestimated the number of M >= 2.5 events. However, it was consistent with the largest event in that period, an M 3.4. Then, we updated the fore-cast model for a new depressurization scenario, testing its pseudoprospective (out-of -sample) performance using L-test and N-test. This revision considered two possible magnitude-frequency distributions (MFD): (1) tapered Gutenberg-Richter (GR), in which induced ruptures are confined within the reservoir, and (2) unbounded GR, in which ruptures can propagate into the underburden. A tapered GR is the more likely scenario due to clear rollover in the MFD and a lack of events beneath the reservoir. However, we cannot exclude the possibility of an unbounded GR, because direct stress measurements in the underburden are not currently available. For the period 2021-2030, this new fore-cast suggests a 1% likelihood that the largest event will exceed M 3.8 for the tapered model and M 5.4 for the unbounded model.

Automated summary

Induced seismicity from gas extraction in the Groningen reservoir in the Netherlands has been occurring since the 1990s. The accuracy and relevance of previous earthquake magnitude forecasts for different extraction scenarios are unknown. This study evaluated a prior forecast and proposed a revised model, which suggested a 1% likelihood of the largest event exceeding M 3.8 (tapered model) or M 5.4 (unbounded model) in the period 2021-2030.