Probabilistic forecasting of induced seismicity in the Groningen gas field

Jonathan D. Smith, Meyer Hadrien, Jean-Philippe Avouac, Robert S. White, & Stephen Bourne

Published August 12, 2019, SCEC Contribution #9428, 2019 SCEC Annual Meeting Poster #032

The Groningen gas field, situated in the north-east of the Netherlands, has been in production since 1963. Prior to gas extraction, the region was considered aseismic with no historical earthquakes recorded. However, since the late 1980s small magnitude earthquakes have been detected, with these shallow events causing non-structural damage and public concern. In January 2014, production across the reservoir was reduced from 53.8 billion cubic metres (bcm) to 42 bcm, in response to the increasing frequency of earthquakes across the region. Since 2014 reservoir extraction rates have been declining (27.5 bcm in 2015, 21.6 bcm in 2018) with extraction rates expected to reach values of 12 bcm in the years 2018-2023, and cease by 2030.

A simple analytical pressure diffusion model is produced by supplementing the more in-depth pressure depletion models generated by the regulator, with a more computationally efficient procedure. We demonstrate within similar margin of errors (<5%) that a simple diffusion model is able to reconcile the observed reservoir pressure depletion. Using these pressure depletion models in conjuction with our geodetic derived surface subsidence models (a combination of Optical Levelling, PS-InSAR and GPS) we are able to determine that the reservoir has a pressure-invarient compressibility deforming poroelastically, with values varying within <1% of the mean compressibility across all geodetic acquisitions.

Through the combination of the analytical pressure diffusion models and compressibility models we are able to determine the three-dimensional spatial redistribution of stress from continued gas production across the region. By combining these stress redistribution models and our refined earthquake hypocentral locations, we are able to determine a pre-stress distribution required to reconcile the time delay in observed seismicity. We use the pre-stress to model future spatial distributions and maximum magnitudes of earthquakes with future production scenarios, formulating an inversion procedure that minimizes both of these quantities for the expected regulatory future production scenarios up to 2030. Using these methods we determine that the maximum magnitude does not exceed a MLN=4.5 even for the worst scenario of a unconstrained maximum magnitude, outlining production scenarios that could minimize the earthquake production.

Key Words
Induced Seismicity, Probabilistic Forecasting, Reservoir Dynamics, Geodetic Inversion

Citation
Smith, J. D., Hadrien, M., Avouac, J., White, R. S., & Bourne, S. (2019, 08). Probabilistic forecasting of induced seismicity in the Groningen gas field. Poster Presentation at 2019 SCEC Annual Meeting.

Related Projects & Working Groups
Earthquake Forecasting and Predictability (EFP)