InFocus

Gas production can lead to fault slip in the subsurface, which can result in earthquakes. By combining measurements of fault slip in a laboratory with state-of-the-art multi-physics models through data assimilation, we can refine a faults properties and variables. This allows to better understand the physical processes at the fault’s interface. Successful synthetic tests have demonstrated the vast potential Bayesian data assimilation to estimate and forecast fault stress. We now aim to evaluate whether Bayesian data assimilation can reliably constrain and probabilistically forecast fault slip in laboratory experiments. This research could ultimately provide a decision-making tool that through considering different gas-production scenarios helps to minimize the effects of earthquakes for society.

Project description 

Gas production can induce seismicity through changes in normal and shear stress on pre-existing faults. Which processes govern these stresses is presently a subject of debate. Compaction, fluid flow and friction each play a role in the occurrence of seismicity. Only if we can resolve these three aspects, their interactions and the resulting stress field, it may be possible to estimate and ultimately forecast seismicity. 

This is not straightforward, as uncertainties in subsurface conditions, the inherent intermittent occurrence of fault slip, and the limited availability of insitu observations make it difficult to uniquely describe responsible mechanisms, let alone provide probabilistic estimates of seismicity. In InFocus, we do this through coupled simulation of geomechanics, dynamic earthquake rupture and two-phase flow with a code originally developed for large-scale tectonic settings.

Analysis of seismic observations over the Zeerijp area of the Groningen field will help to calibrate the model so it can be used in reservoir-scale studies. Next, we propose to further refine the state or parameter estimate using a statistical, Bayesian approach in data assimilation. To test the validity of this data-assimilation approach we develop a simplified, controlled laboratory setup of a slipping fault monitored with acoustic emission sensors and strain sensors and a parallel numerical simulation of this setup in which we assimilate the lab measurements. This helps to identify what observations would be required to better understand and ultimately forecast seismicity as a result of gas extraction. Extending data-driven simulation tools developed for subduction zones to laboratory and reservoir scales is expected to shed light on the dynamics and occurrence of seismicity.

Background

The recent seismicity induced by natural gas extraction from the Dutch Groningen gas field underlines the importance of understanding and forecasting induced seismicity1,2. At the moment, quantitative estimates of key processes and their uncertainties are lacking. A predictive tool for decision-making requires integration of state-of-the-art experimental, numerical and observational analysis across a range of spatial scales. The integration of existing elements in InFocus will provide a framework to evaluate future and past decisions on gas extraction as, for example, the recent decision to halt gas production in Groningen3. In addition, it provides a first step towards a means to forecast induced seismicity.

Insights are evolving through observational, laboratory and numerical studies. High-resolution observations of seismic activity in the Zeerijp area of the Groningen field have been used for estimation of epicentre location and source characteristics11. Laboratory experiments quantify controlling fault properties and allow to study fault-rupture processes in more detail. Modelling studies address the processes involved in fault rupture and their influence on the occurrence and severity of seismic events in a reservoir setting12-15,5 of both observations and laboratory experiments.

Figure 1. Error diagram illustrating exceptional performance in forecasting large synthetic earthquakes, since the red, solid lines approach the optimal origin (0,0), while the periodic benchmark (dashed red lines) rather performs random16.


InFocus will complement geophysical observations by laboratory studies and numerical simulations. Key in this integrated approach is the use of data assimilation methodologies. Data assimilation methodologies originate from meteorology and oceanography. Bayesian data assimilation has been shown to be very effective in estimating and forecasting fault stress in a perfect model test of seismic cycle models in subduction zones 16 (Figure 1 and 2D). We now aim to evaluate this potential in a real-life laboratory setting in which physics is not represented perfectly.

In addition to constraining model uncertainties, data assimilation can also help to identify which observations provide the most valuable information for estimation and forecasting through value-of-information studies17,18 . It provides a quantitative tool for decision making that provides key information to policymakers to define the requirements for present and future monitoring programs.

Project objectives and approach

The overall aim of InFocus is to prioritise potential mechanisms responsible for induced seismicity due to gas extraction and evaluate the relative importance of observations in constraining state and parameters controlling fault slip occurrence in a laboratory setup. This is achieved through the following objectives:

  • Evaluate processes controlling fault slip using observations, laboratory experiments and multi-physics numerical simulations
  • Develop a data-driven multi-physical modelling framework for induced fault slip
  • Constrain fault state and parameters through assimilation of laboratory measurements
  • Evaluate observational requirements to monitor fault slip behaviour

Given the growing societal requirement for monitoring and forecasting the environmental effects of subsurface activities, these objectives will eventually provide key information to policymakers to prevent any undesired effects of subsurface activities.

Key to the approach of this project is the application of non-linear data assimilation to a state-of-the-art dynamic model of fault rupture. The application will be validated in the controlled setting of laboratory experiments. Using this combination we will address relevant physical and observational questions. The elements of the project, illustrated in Figure 2, will together deliver:

  • a seismological earthquake catalogue for the Zeerijp area (A in Fig 2)
  • a laboratory earthquake catalogue (C in Fig 2)
  • a multi-physics forward model for simulating induced seismicity by adapting an existing tectonic seismic cycle model with poro-elastic mechanics, two-phase fluid flow, and dynamic rupture and seismic wave propagation (B in Fig 2)
  • a data assimilation framework to combine the above model with surface and subsurface observations for data-driven modelling of non-linear fault state and rupture (D in Fig 2)

Figure 2 illustrates the various parts of the project and how they are interconnected. Eventually, this will lead to improved identification of physical mechanisms responsible for induced seismicity due to fluid extraction as well as an assessment of the relative influence of observations in the controlled laboratory setting, which will help to define observational requirements for realistic cases.

Figure 2 Elements of InFocus and their connections. A) Seismological observations of the Zeerijp area. B) Examples from the available poro-visco-plasto-elastic model showing a 3D viscous sinker, seismic cycles, a 3D dynamic rupture. C) pistons for laboratory experiments in the Utrecht High Temperature Laboratory. D) proof of concept of data assimilation for shear stress estimation with the available model.

Project organisation

The integrated approach of InFocus is founded on the involvement of staff of various fields of expertise, employed at six different knowledge institutions (Fig. 3). The involvement of staff from both Delft University of Technology and Utrecht University is critical for the success of this project and will form the central piece in the network of institutions involved.

To reach the research objectives, three people will be recruited: one PhD student at Delft University of Technology, one PhD student at Utrecht University and a post-doc. Strong interaction between all researchers will be ensured through regular meetings in and personal visits to Utrecht, Delft or the KNMI. Yearly visits to collaborators in Sheffield, Reading and Zürich will stimulate the interaction with international leading experts at University of Sheffield, University of Reading and ETH Zürich.

Researchers involved:

Collaborations with: Peter Jan van Leeuwen (University of Reading/Colorado State University), Taras Gerya (ETH Zürich), Caspar Pranger (ETH Zürich), René de Borst (University of Sheffield), Jan Dirk Jansen (Delft University of Technology) and Liviu Matenco (Utrecht University).

Through regular visits of staff to the various knowledge institutions, the research team will be able to work in seamless collaboration, leading to inspired partnership in follow-up projects.

Figure 3 Embedding of InFocus within ongoing research of the institutes involved.

Literature references

  1. Lele, S.P., S.Y. Hsu, J.L. Garzon, N. DeDontney, K.H. Searles, G. A. Gist, P.F. Sanz, E.A.O. Biediger, and B.A. Dale (2016), Geomechanical modeling to evaluate production-induced seismicity at Groningen field, in Proceedings of the Abu Dhabi International Petroleum Exhibition and Conference, SPE-183554-MS, doi:10.2118/183554-MS
  2. Zbinden, D., A. P. Rinaldi, L. Urpi, and S. Wiemer (2017), On the physics-based processes behind production-induced seismicity in natural gas fields, J. Geophys. Res., doi:10.1002/2017JB014003
  3. Reuters, Netherlands to halt gas production at Groningen by 2030, press release, https://www.reuters.com/article/us-netherlands-groningen-gas/netherlands-to-halt-gas-production-at-groningen-by-2030-idUSKBN1H51PN
  4. Bogert, P.A.J. van den (2015), Impact of various modelling options on the onset of fault slip and the fault slip response using 2-dimensional finite-element modelling, Shell Int. Expl. Prod. B.V., https://nam-feitenencijfers.data-app.nl/download/rapport/604ef59b-5ac5-4770-93b1-e501c87da3a9?open=true
  5. Buijze, L., P. van den Bogert, B. Wassing, B. Orlic, and J. ten Veen (2018), Fault reactivation mechanisms and dynamic rupture modelling of depletion-induced seismic events in a Rotliegend gas reservoir, Netherlands J. Geosci. / Geologie en Mijnbouw, doi:10.1017/njg.2017.27
  6. Platteeuw, I. (2018), Initiation of fault reactivation: New insights into the effect of differential compaction, Master’s thesis, Delft University of Technology, http://resolver.tudelft.nl/uuid:d6d83aba-3437-45d5-b099-dd5c1b287d54
  7.  Wees, J.D. van, P.A. Fokker, K. van Thienen-Visser, B.B.T. Wassing, S. Osinga, B. Orlic, S.A. Ghouri, L. Buijze, and M. Pluymaekers (2017), Geomechanical models for induced seismicity in the Netherlands: inferences from simplified analytical, finite element and rupture model approaches, Neth. J. of Geosci., doi:10.1017/njg.2017.38.
  8. Andrews, D. J., and Y. Ben-Zion (1997), Wrinkle-like slip pulse on a fault between different materials, Journal of Geophysical Research: Solid Earth, doi:10.1029/96JB02856
  9. Dalguer, L. A., and S. M. Day (2009), Asymmetric rupture of large aspect-ratio faults at bimaterial interface in 3D, Geophys. Res. Letters, doi:10.1029/2009GL040303
  10. Share, P.E., and Y. Ben-Zion (2016), Bimaterial interfaces in the South San Andreas fault with opposite velocity contrasts NW and SE from San Gorgonio pass, Geophys. Res. Letters, doi:10.1002/2016GL070774
  11. Dost, B., E. Ruigrok, and J. Spetzler (2017), Development of seismicity and probabilistic hazard assessment for the Groningen gas field, Neth. J. Geosci., doi:10.1017/njg.2017.20
  12. Mulders, F.M.M. (2003), Modelling of Stress Development and Fault Slip in and around a Producing Gas Reservoir, Ph.D. dissertation, Delft University of Technology, https://repository.tudelft.nl/islandora/object/uuid%3Abe742135-10d7-4d69-bdee-f808b5926065
  13. Wees, J.D. van, L. Buijze, K. van Thienen-Visser, M. Nepveu, B.B.T. Wassing, B. Orlic, and P.A. Fokker (2014), Geomechanics response and induced seismicity during gas field depletion in the Netherlands, Geothermics, doi:10.1016/j.geothermics.2014.05.004
  14. Buijze, L., B. Orlic, B.B.T. Wassing, and G.-J. Schreppers (2015), Dynamic rupture modeling of injection-induced seismicity: Influence of pressure diffusion below porous aquifers, in Proceedings of the 49th US Rock Mechanics / Geomechanics Symposium, ARMA 15-384, https://www.onepetro.org/conference-paper/ARMA-2015-384 
  15. Wassing, B.B.T., L. Buijze, and B. Orlic (2016), Modelling of fault reactivation and fault slip in producing gas fields using a slip-weakening friction law., in Proc. of the 50th U.S. Rock Mechanics/Geomechanics Symposium, 26-29 June, Houston, Texas, ARMA-2016-658, American Rock Mechanics Association, https://www.onepetro.org/conference-paper/ARMA-2016-658
  16. Dinther, Y. van, H. R. Künsch, and A. Fichtner (2019), Ensemble data assimilation for earthquake sequence: Probabilistic estimation and forecasting of fault stresses, Geophys. J. Int., in press, doi: 10.1093/gji/ggz063, file: https://www.dropbox.com/s/6zp3zva2bzxvydc/ggz063.pdf?dl=0 
  17. Bratvold, R.B., J.E. Bickel, H.P. Lohne (2009), SPE Reservoir Evaluation and Engineering, doi:10.2118/110378-MS
  18. Barros, E., J.-D. Jansen, and P. Van den Hof (2015), Value of information in parameter identification and optimization of hydrocarbon reservoirs, IFAC-PapersOnLine, doi:10.1016/j.ifacol.2015.08.036

Contact information

Dr.ir. Femke C. Vossepoel: f.c.vossepoel@tudelft.nl

For additional positions at Utrecht University, please contact Dr. Ylona van Dinther: y.vandinther@uu.nl