Production-induced seismicity indicates a low risk of strong earthquakes in the Groningen gas field

The Groningen gas field, the largest in Western Europe, has experienced seismicity induced by decades of gas production. Determining the maximum possible earthquake magnitude (Mmax) associated with this activity is crucial for seismic hazard assessment and risk management. This question is particularly challenging and controversial due to the long-term nature of hydrocarbon production in Groningen. The paper aims to address this issue by developing a novel approach to differentiate between induced and triggered seismicity. Understanding the distinction is paramount, as induced seismicity, primarily controlled by the technical operations, is more manageable than triggered seismicity, which relies on pre-existing tectonic conditions. Accurate assessment of both induced and triggered Mmax is vital for effective seismic hazard control in Groningen, especially as gas extraction is nearing its end, and post-production seismicity is anticipated. The research explores whether the characteristics of induced seismicity can provide insights into the potential for larger, triggered tectonic earthquakes. This is critical for the safe and sustainable extraction of geo-energy resources worldwide.

The literature extensively documents that both fluid extraction and injection can induce seismicity. However, estimating Mmax for such operations, particularly in the context of long-term hydrocarbon production (as seen in Groningen), remains under-researched and a subject of debate. Existing studies have offered various approaches to discriminate between induced and triggered seismicity, each focusing on different physical aspects of the phenomenon. Some differentiate based on the stress impact of the stimulation, others consider the events' dependence on human-made stress changes, and yet others define the distinction based on the location of the rupture surface relative to the stimulated volume. The challenges of constraining the magnitude bound or taper of the Gutenberg-Richter (GR) statistic from purely statistical analyses of observed seismicity have also been highlighted. Previous work on Groningen seismicity has included statistical analyses of the event-size distribution and the development of ground-motion prediction models. However, a comprehensive approach integrating reservoir geometry, seismicity statistics, and the potential for triggered events has been lacking.

The study employs a multi-faceted approach to estimate Mmax for fluid extraction from a flat underground reservoir. It begins by considering induced seismicity as events entirely contained within the reservoir layer, assuming approximately isometric ruptures. Using the known reservoir geometry, the maximum magnitudes for these reservoir-limited ruptures are calculated using a formula that incorporates effective rupture length and stress drop. The Lower-Bound (LB) frequency-magnitude statistic is applied to the Groningen reservoir to obtain a maximum possible induced magnitude (MI). This statistic describes a truncated Gutenberg-Richter law where larger magnitudes are underrepresented, implying an upper magnitude limit. The LB-statistic accounts for the finite stimulated volume and differs from traditional GR statistics that assume an infinite seismo-tectonic continuum. Explicit equations for the LB-statistic are derived considering critically oriented faults and limiting rupture sizes within the reservoir layer. These equations are then fitted to the observed frequency-magnitude distribution in Groningen. The authors also calculate the Seismogenic Index (SI), which quantifies the normalized seismic response of the seismo-tectonic continuum to a perturbation of the Coulomb Failure Stress (CFS). The SI is computed using two scenarios: one considering pore pressure changes since the start of production, and another considering changes only after the first observed earthquake in 1992. Finally, the Worst-Case Exceedance Probability (WCEP) is estimated to assess the probability of triggering larger-magnitude events in the seismo-tectonic continuum. This calculation considers the upper bound of the SI and the lower bound of the b-value, representing a worst-case scenario. Spatial variations in LB parameters (a-value, b-value, My, and SI) are also mapped using a smoothing radius of 15 km to account for variations in reservoir thickness across the field. The methodology relies on observed data such as annual gas production, pore pressure decrease, annual earthquake rates, and the spatial distribution of earthquakes. It combines theoretical models with statistical analyses of observed seismicity data to estimate the maximum possible magnitude and the probability of triggering larger events.

The analysis reveals a maximum possible induced magnitude (MI) of approximately 4 for the Groningen gas field, based on the LB-statistic. This value aligns with the lower limit of commonly used estimates. The model also indicates that if larger tectonic earthquakes could be triggered, the probability of a Mw 5.5 earthquake would be substantial (up to 30%), implying that multiple Mw ≥ 4 events should have occurred given the field's long production history. This contradicts the actual observations. The spatial distribution of maximum possible magnitudes (My) derived from the LB-statistic shows a good agreement with the observed maximum magnitudes, increasing from southeast to northwest, correlating with reservoir thickness. The LB-statistic fits to the observed frequency-magnitude distribution more accurately than the classical GR statistic, suggesting an upper magnitude limit. The temporal analysis shows that the b-value derived from the LB-statistic initially decreased and then increased after larger-magnitude events, likely due to aftershocks. The Seismogenic Index (SI) exhibits a slight increase over time in a stationary-condition scenario, potentially explained by a gradual involvement of cohesive faults in seismic activation. Spatially, the a-value remains relatively constant, while the b-value shows a clear segmentation between the northwest (low b-values) and southeast (high b-values) parts of the field, consistent with reservoir thickness variations. The WCEP, which estimates the probability of triggering larger magnitude events, suggests a low risk of such events in Groningen.

The discrepancy between the model's prediction of multiple Mw ≥ 4 earthquakes and the lack of such observations in the Groningen field strongly suggests that the field is inherently stable. The physical conditions necessary to trigger large tectonic earthquakes are likely absent. The good agreement between the calculated maximum magnitudes based on reservoir geometry and observed magnitudes further supports the conclusion that seismicity in Groningen is primarily induced and limited by the reservoir's geometry. The application of the LB-statistic, which accounts for the finite nature of the stimulated volume, proves more effective than the classical GR statistic in describing the observed frequency-magnitude distribution. The findings have important implications for seismic hazard assessment in Groningen and similar settings, suggesting that the risk of large triggered earthquakes may be lower than previously thought. This research provides a novel framework for integrating reservoir characteristics and seismicity data to evaluate the likelihood of induced and triggered seismicity.

This study demonstrates a new method for assessing the seismic hazard in gas production fields by differentiating between induced and triggered earthquakes. The results suggest a low risk of strong earthquakes in the Groningen gas field due to its inherent stability. The model's prediction of strong earthquakes that are not observed highlights the limitations of models assuming an infinite seismo-tectonic continuum and reinforces the importance of considering the finite dimensions of the stimulated rock volume. Future research could focus on refining the model to account for more complex geological features and further investigate the underlying mechanisms of induced and triggered seismicity in similar settings. The development of improved monitoring techniques and data analysis methods could also enhance the accuracy of seismic hazard assessments.

The study assumes a simplified reservoir geometry (flat layer) and neglects potential complexities in the subsurface structure that may influence stress distribution and fault behavior. The analysis relies on the accuracy of existing seismic catalogs and pore-pressure data, which may have uncertainties or limitations, particularly in the early years of the gas field's operation. The model's prediction of multiple Mw ≥ 4 earthquakes that are not observed could indicate either an issue with model assumptions or an unexpected resilience of the Groningen field. Future work could focus on improving model accuracy by incorporating higher-resolution subsurface models and more advanced statistical approaches.