PSP16 Towards Numerical Modelling of Triggered ...

PSP16 Towards Numerical Modelling of Triggered and Induced Seismicity D. Kuehn* (NORSAR), T. Dahm (GFZ German Research Centre For Geosciences), M. Wangen (IFE) & S. Heimann (GFZ German Research Centre For Geosciences)

SUMMARY Human-induced earthquakes are more and more brought into the focus of public attention. E. g. Ellsworth (2013) attracts notice to the dramatic increase of the number of earthquakes in the central and eastern United States over the past few years. Environments prone to induce or trigger seismicity are numerous, e. g. oil and gas exploration sites, large-scale surface quarries and mines, Enhanced Geothermal Systems (EGS), dam sites and injections of e.g. CO2 or waste water. Since the nature of induced and triggered earthquakes implies their occurrence near engineering activity, even earthquakes of small magnitude are a cause for concern. We present several numerical methods to enhance the understanding of the spatial and temporal occurrence of seismicity, which has been triggered or induced by human operations: a) poroelastic modelling employing elementary Green’s functions, b) analytical fracture model combined with a rate- and state- dependent constitutive model and c) a 3D FEM able to handle both heterogeneous rocks and branched fractures.

Fifth Passive Seismic Workshop From Wish-List to To-Do List 28 September - 1 October 2014, Lisbon, Portugal

Introduction Human-induced earthquakes are more and more brought into the focus of public attention. Ellsworth (2013) attracts notice to the dramatic increase of the number of earthquakes in the central and eastern United States over the past few years: ZKHUHDVWKHDYHUDJHUDWHRI0•HDUWKTXDNHVSHU\HDUZDVDV low as 21 in the period between 1967 and 2000, the rate increased to more than 300 earthquakes with 0•in the period 2010 to 2012, whereof 188 events took place in 2011 alone. Although the activity seems to be related to the exploration of tight shale formations, the origin of earthquakes is probably rather connected to the injection of wastewater in deep wells than the fracking itself (Ellsworth 2013). Environments prone to induce or trigger seismicity are numerous, e.g. oil and gas exploration sites, large-scale surface quarries and mines, Enhanced Geothermal Systems (EGS), dam sites and injections of e.g. CO2 or waste water. In some of these projects, seismicity is intentionally induced to enhance reservoir permeability by hydrofracturing, e.g. to increase oil or gas production or explore geothermal reservoirs. Since the nature of induced and triggered earthquakes implies their occurrence near engineering activity, even earthquakes of small magnitude are a cause for concern. In addition, they usually occur at very shallow depths and thus, even weak events may pose a seismic hazard or be felt by the population. The Basel geothermal exploration (Switzerland) was stopped in 2006 due to a magnitude M 3.4 event (Häring et al. 2008) and led to 2300 damage claims (Aegerter and Bosshardt 2007). The youngest damaging gas-field induced earthquake in Northern Europe is the 16 August 2012 M 3.6 Groningen field event. Induced seismicity The terms “induced” and “triggered” seismicity are used inconsistently. In the following, we will use the definition given by Dahm et al. (2012a): the occurrence of an “induced” earthquake is considered as entirely controlled by the human-induced stress change, whereas in the case of a “triggered” earthquake, the human-induced stress change causes only the event nucleation, but not the whole failure process. Until now, the discrimination between natural, triggered and induced earthquakes is challenging and no clear rules or scientific methods are established (Dahm et al. 2010a). The problem of induced seismicity was first recognized for the mining industry in Germany and South Africa at the beginning of the 20th century, for petroleum production in the 1920s, impoundment of water reservoirs in the 1930s and high-pressure fluid injection in the 1960s. Early examples of induced seismicity at oil fields stem from the Wilmington oil field (California), where oil production caused a dramatic subsidence up to 8.8 m accompanied by damaging earthquakes in the period between 1947 and 1961 (Kovach 1974). At the Gazli gas field (Uzbekistan), two magnitude 7 earthquakes are discussed to be triggered by production in 1976 and 1984; if so, then the second event is the largest triggered seismic event ever due to gas withdrawal (Simpson and Leith 1985). The largest known injection-induced earthquakes (0” 5.5) were associated with long-term well injection of wastewater at the Rocky Mountain Arsenal in 1966/67 (Healy et al. 1968), and more recently, a wastewater injection may have triggered the 2011 MW 5.7 central Oklahoma earthquake (Kerenan et al. 2013). Both the impoundment of the Koyna water reservoir (India), the Kariba reservoir (Zimbabwe), and the Kremasta reservoir (Greece) led to earthquakes with magnitudes above 6 (Gupta et al. 1972). The largest mining induced events were a mb 5.5 rock burst in a potash mine near Völkershausen, Germany (Knoll 1990) and an equally sized event in a mine in Newcastle, Australia (Klose 2007). Induced seismicity may be caused by the impact of engineering operations on e.g. stress and strain fields, pore pressure, fluid migration, fluid saturation, and rock strength (Dahm et al. 2010a). Induced stress changes on the order of only a few bars may trigger seismicity on pre-existing faults, if these are close to failure due to a high ambient tectonic stress field (Simpson 1986). Fluid pressure effects Fifth Passive Seismic Workshop From Wish-List to To-Do List 28 September - 1 October 2014, Lisbon, Portugal

from injection operations can extend well beyond the range of actual fluid migration, and such the occurrence of seismicity, once started, may not be controlled completely or easily (Nicholson and Wesson 1992). Especially the importance of hydraulically conductive faults for transmitting the pore pressure within pre-existing fault systems should not be underestimated (Ellsworth 2013). The relation between operational parameters and seismicity can be obscure, resulting in a need for hydro-/geomechanical modelling to enhance understanding of the processes (Ellsworth 2013). In the following, we will present two methods of numerical modelling, which may be applied in order to better understand the origin of triggered and induced seismicity. Poroelastic modelling is appropriate in cases where the physical interaction between rock matrix and pore fluid becomes important. On the other hand, fracture models may be related to seismicity models in order to better understand seismicity generated during fluid injection. Poroelastic modelling To understand or model the physical interaction between rock and fluid in cases where fluid-saturated rock is deformed, elasticity theory is insufficient and (linear) poroelasticity theory has to be applied. Originally, it was developed to calculate the consolidation of porous media in response to loading and has since then developed further to model a variety of processes like for instance seismotectonically induced groundwater flow, the propagation of seismic waves in porous media, land subsidence due to fluid extraction from subsurface reservoirs and production-induced strain near wells. Analytical solutions exist only for elementary configurations (e.g. a point source in a homogeneous full space) and for strong coupling between diffusion and deformation processes (Rice and Cleary 1976). We developed an approach to model the deformation of porous rock and the diffusion of pressure and fluids in layered media using elementary Green’s functions. This approach may be used to model the following problems: x to compute stress changes and deformation in the volume around an injection well and during the depletion of a gas field; x to compute the time-dependent variation of pore pressure within a horizontal aquifer and its effect on deformation and stress distribution in the rock volume; x to compute the time-dependent deformation and stress distribution in a rock volume after dislocation on a fault. Once the stress distribution surrounding such volumes has been calculated, it can be used to estimate the earthquake triggering potential on surrounding faults of known location and size.

Figure 1 Representation of reservoir by elementary Green’s functions; computation of displacement at observation point by superposition. Fifth Passive Seismic Workshop From Wish-List to To-Do List 28 September - 1 October 2014, Lisbon, Portugal

To compute the elementary Green’s functions, we employed the Fortran code POEL by Wang and Kümpel (2003). This code computes transient and steady-state poroelastic solutions for excess pore pressure and displacements in a homogeneous or multilayered half-space caused by forcing through a point or vertical line-source. We implemented routines in the Pyrocko framework (http://emolch.github.com/pyrocko) to not only compute and store Green’s functions, but also to retrieve and combine them to efficiently solve problems of time-dependent, extended sources of arbitrary shape (Figure 1). The Kiwi Tools (Heimann, 2011) as well as ready to use pre-computed

Green's functions (Green's function stores) can be downloaded from the web page of the KINHERD project (http://kinherd.org/). Fracture modelling From early on, hydraulic fracturing models were developed to estimate the shape of the hydrofracture, but in general, they do not explain the pattern of induced seismicity. Other models assume that seismicity is triggered by a transient pore pressure pulse to explain the seismic back front, but do not consider fracturing (Shapiro et al. 1997). A recent study links both approaches by considering fluid leakage from newly formed fracture into the porous formation (Shapiro et al. 2006). An unexpected observation was the identification of asymmetric bi- and unidirectional growth of the seismic foreand back front, indicating asymmetric fracture growth. Dahm et al. (2010b) developed an analytical hydrofracture model to explain these observations, considering the effect of injection pressure, volume rate, stress gradients, viscosity and elastic modules of the rock. Recently, Dahm et al. (2012b) associated their hydrofracture model with a rate- and state-dependent constitutive model (Dieterich 1994) to simulate induced seismicity in space and time. However, their model disregards fracture complexity due to the heterogeneous nature of the rock. Wangen (2011; 2013) presents a finite element model of hydraulic fracturing of a poroelastic rock. The model simulates the pressure build-up from an injection well, how the fluid flows in the fracture and how it leaks off into the host-rock. The model accounts for the displacement field, the strain and the stress that result from the pressure gradients in the host rock. Fracture events are triggered when the stress (or strain) in an element exceeds a fracture threshold. The rupture of elements propagates the fracture and an event where one or more elements break is followed by a pressure drop. The pressure drop following an event, as well as the elastic energy released in the event, is computed. This approach to the modelling hydraulic fracturing handles heterogeneous rocks with branched fractures, since the fracture criterion is an element property. The model was first developed in 2D (Wangen, 2011; see Figure 2) and then extended to 3D (Wangen, 2013).

a)

b)

Figure 2 a) Fluid injection is simulated in a heterogeneous rock using the FEM code by Wangen et al. (2013). The background is coloured by the effective stress; note the irregular and branching growth of the fracture. b) Spatio-temporal distribution of microseismic events related to real-data hydraulic fracturing. The general growth of the microseismic cloud (blue to red) is first developing along one direction, but later branching into two fault segments (Albaric et al. 2012). Fifth Passive Seismic Workshop From Wish-List to To-Do List 28 September - 1 October 2014, Lisbon, Portugal

Acknowledgements This work was supported by the Norwegian Research Council (SafeCO2, project no. 189994) and by the industry partners of the SafeCO2 project (Statoil, Lundin, Octio, and READ). References Albaric, J., Oye, V., Langet, N., Hasting, M., Lecomte, I., Iranpour, K., Messeiller, M., Reid and P. [2013]. Monitoring of induced seismicity during the first geothermal reservoir stimulation at Paralana, Australia. Geothermics, in press, doi:10.1016/j.geothermics.2013.10.013. Aegerter, A. and Bosshardt, O. [2007]. Schadenfälle Erdbeben vom 8.12.2006, Perimeterdefinition. Technischer Bericht, 13.07.2007, Ingenieurbüro A. Aegerter & Dr. O. Bosshardt AG, Basel, Möhlin. Dahm, T., Becker, D., Bischoff, M., Cesca, S., Dost, B., Fritschen, R., Hainzl, S., Klose, C.D., Kühn, D., Lasocki, S., Meier, T., Ohrnberger, M., Rivalta, E., Wegler, U. and Husen, S. [2012a]. Recommendation for the discrimination of human-related and natural seismicity. J. Seismol., doi: 10.1007/s10950-012-9295-6. Dahm, T., Hainzl, S. and Fischer, T. [2012b]. Stimulation induced seismicity: fracture combined with rate-and-state model. 3rd Annual Meeting AIM, 10-12 Oct 2012, Smolenice, Slovakia (Abstract). Dahm, T., Hainzl, S., Becker, D., Bischoff, M., Cesca, S., Dost, R., Fritschen, R., Kühn, D., Lasocki, S., Klose, C.D., Meier, T., Ohrnberger, M., Rivalta, R., Shapiro, S. and Wegler, U. [2010a]. How to discriminate induced, triggered and natural seismicity. In: Ritter, J. and Oth, A. (Eds.) Proceedings of the workshop “Induced seismicity”, Nov. 1-17, 2010, Luxembourg, Cahier du Centre Europeen de Geodynamique et de Sismologie, 30, 69-76. Dahm, T., Hainzl, S. and Fischer, T. [2010b]. Bidirectional and unidirectional fracture growth during hydrofracturing: role of driving stress gradients. J. Geophys. Res., 115(B12322). Dieterich, J. [1994]. A constitute law for rate of earthquake production and its application for earthquake clustering. J. Geophys. Res., 99, 2601-2618. Ellsworth, W.L. (2013). Injection-induced earthquakes. Science, 341, 1225942, doi:10.1126/science.1225942. Gupta, H.K., Rastogi, B.K. and Narain, H. [1972]. Some discriminatory characteristics of earthquakes near the Kariba, Kremasta, and Koyna artificial lakes. Bull. Seismol. Soc. Am., 62(2), 493-507. Häring, M.O., Schanz, U., Ladner, F. and Dyer, B.C. [2008]. Characterization of Basel 1 enhanced geothermal system. Geothermics, 37, 469-495. Healy, J.H., Rubey, W., Griggs, D.T. and Raleigh, C.B. [1968]. The Denver Earthquakes. Science, 161, 1301-1310. Heimann, S. [2011]. A Robust Method to Estimate Kinematic Earthquake Source Parameters. PhD Thesis, University of Hamburg, Hamburg, Germany. Kerenan, K.M., Savage, H.M., Abers, G.A. and Cochran, E.S. [2013]. Potentially induced earthquakes in Oklahoma, USA: links between wastewater injection and the 2011 Mw 5.7 earthquake sequence. Geology, 41, 699–702. Klose, C.D. [2007]. Geomechanical modeling of the nucleation process of Australia’s 1989 M5.6 Newcastle earthquake. Earth Planet. Sci. Lett., 256(3-4), 547–553. Knoll, P. [1990]. The fluid-induced tectonic rock burst of March 13, 1989 in Werra potash mining district of the GDR (first results). Gerlands Beiträge zur Geophysik, 99, 239-245. Kovach, R.L. [1974]. Source mechanisms for Wilmington oil field, California, subsidence earthquakes. Bull. Seismol. Soc. Am., 64, 699-711. Nicholson, C. and Wesson, R.L. [1992]. Triggered earthquakes and deep well activities. Pure Appl. Geophys., 139(3/4), 561578. Rice, J.R. and Cleary, M.P. [1976]. Some basic stress diffusion solutions for fluid-saturated elastic porous media with compressible constituents. Rev. Geophys. Space Phys., 14, 227–241. Shapiro, S.A., Dinske, C. and Rothert, E. [2006], Hydraulic-fracturing controlled dynamics of microseismic clouds. Geophys. Res. Lett., 33, L14312, doi:10.1029/2006GL026365. Shapiro, S.A., Huenges, E. and Borm, G. [1997]. Estimating the crust permeability from fluid-injection-induced seismic emission at the KTB site. Geophys. J. Int., 131, F15–F18, doi:10.1111/j.1365-246X.1997.tb01215.x. Simpson, D.W. [1986]. Triggered earthquakes. Ann. Rev. Earth Planet. Sci., 14, 21-42. Simpson, D.W. and Leith, W. [1985]. The 1976 and 1984 Gazli, U.S.S.R. earthquakes – were they induced? Bull. Seismol. Soc. Am., 75, 1465-1468. Wang, R. and Kümpel, H.-J. [2003]. Poroelasticity: efficient modelling of strongly coupled, slow deformation processes in a multilayered half-space. Geophysics, 68(2), 705-717. Wangen, M. [2011]. Finite element modelling of hydraulic fracturing on a reservoir scale in 2D. J. Petrol. Sci. Engin., 77, 274-285. Wangen, M. [2013]. Finite element modeling of hydraulic fracturing in 3D. Comp. Geosci., 17(4), 647-659.

Fifth Passive Seismic Workshop From Wish-List to To-Do List 28 September - 1 October 2014, Lisbon, Portugal