The concept of mining heat from hot, low-permeability rock at great depth was first proposed at Los Alamos National Labs in the 1970s and was called hot dry rock system (Brown et al., 2012). They initially envisioned creating a reservoir by applying oil and gas reservoir hydro-fracture technology to build a heat exchanger between two boreholes. Subsequent field tests have demonstrated that hydraulic stimulation injections are effective in enhancing the permeability of a rock mass by several orders of magnitude by producing irreversible fracture opening, whilst also increasing the connectivity of the fracture network (Kaieda et al., 2005; Evans et al., 2005b; Häring et al., 2008). Two different “end-member” mechanisms commonly appear in discussions of permeability creation processes through hydraulic injections: (1) hydraulic fracturing as the initiation and propagation of new tensile fractures and (2) hydraulic shearing, i.e., the reactivation of existing discontinuities in shear with associated irreversible dilation that is often referred to as the self-propping mechanism. Hydraulic shearing is of particular relevance for EGS as it has been shown that slip along fractures can generate a permeability increase by up to 2–3 orders of magnitude (Jupe et al., 1992; Evans et al., 2005a; Häring et al., 2008). If the rock mass in the reservoir is stressed to a critical level (e.g., Byerlee, 1978), then a relatively small reduction in effective normal stress would be sufficient to cause shearing along preexisting discontinuities that are optimally oriented for failure (Hubbert and Rubey, 1959; Raleigh et al., 1976; Zoback and Harjes, 1997; Evans et al., 1999; Evans, 2005). Thus, shearing and the associated permeability enhancement can occur at large distances from the injection point, even though the causal pressure increases may be low (Evans et al., 1999; Saar and Manga, 2003; Husen et al., 2007). In contrast, hydraulic fracture initiation and propagation (i.e., the original concept of EGS to connect two boreholes) requires high pressures exceeding the minimum principal stress to propagate hydro-fractures away from the wellbore. The high pressure in the fracture may interact with natural fractures and stimulate them, leading to leak-off (i.e., the extent of hydro-fractures is influenced by pressure losses and the existence of preexisting fractures). Therefore, hydraulic fracturing is often only considered relevant in the near field of a wellbore, where it improves the linkage between the borehole and the natural fracture system. Rutledge et al. (2004) showed that shear activation of existing fractures and creation of new fractures can occur concomitantly, dependent on the in situ stress conditions, injection pressure, initial fracture transmissivity, fracture network connectivity and fracture orientation (e.g., McClure and Horne, 2014). Regardless of which process is dominant, the direction of reservoir growth, and therefore the geometry of the stimulated volume, depends to a considerable degree on the in situ stress gradient, stress orientation and the natural fracture network.
Pressurized fractures may open due to a reversible, compliant response to pressure (Rutqvist, 1995; Rutqvist and Stephansson, 2003; Evans and Meier, 1995) or due to largely irreversible shear dilation (Lee and Cho, 2002; Rahman et al., 2002). As a consequence of the coupling between pressure, fracture compliance and permanent fracture aperture changes, the pressure field does not propagate through the reservoir as a linear diffusive field, but rather as a pressure front (Murphy et al., 2004). The fracture normal and shear dilation that occurs in response to elevated fluid pressure thus has a major influence on the magnitude and profile of the propagating pressure perturbation in the rock mass during hydraulic stimulations (Evans et al., 1999; Hummel and Müller, 2009). As a consequence, fracture compliance and normal or shear dilation characteristics have an impact on the size and geometry of the reservoir created during hydraulic stimulation.
Although the aforementioned processes are conceptually well understood, the quantification and detailed understanding required for designing stimulations and truly engineering geothermal reservoirs are insufficient. There remains considerable uncertainty as to how the above processes interact and what rock mass characteristics and injection metrics control the dominant mechanisms (Evans et al., 2005a; Jung, 2013). Thermo-hydromechanically coupled numerical models have become widely used for analyzing relevant aspects of reservoir stimulation in retrospective (e.g., Baujard and Bruel, 2006; Rutqvist and Oldenburg, 2008; Baisch et al., 2010; Gischig and Wiemer, 2013) or as prospective tools for predicting reservoir behavior or alternative stimulation strategies (e.g., McClure and Horne, 2011; Zang et al., 2013; Gischig et al., 2014; McClure, 2015; Yoon et al., 2014). The fact that such numerical models must be parameterized from sparse quantitative field-scale data is a major limitation of all those studies. In the following we present an overview of the experimental observations of hydromechanical coupling that are relevant to the parameterization of numerical models. These stem from reservoir-scale (i.e., hectometer) stimulation operations, such as in EGS demonstration projects or oil and gas reservoirs, intermediate-scale (i.e., decameter) in situ experiments and small-scale laboratory experiments.
The paucity of high-quality data on the stimulation process from reservoir-scale projects is largely a result of the considerable depth of typical geothermal resources (e.g., several kilometers), which prohibits the observation of hydromechanical processes from instrumentation installed within the reservoir. In the geothermal domain, such projects constitute expensive experiments and thus are relatively few in number, whereas, in the oil and gas domain, where hydro-fracture operations are frequent and routine, the data tend to be proprietary. Nevertheless, some notable datasets have been acquired for deep brine injection projects (Ake et al., 2005; Block et al., 2015), deep scientific drilling projects such as the German KTB project (Zoback and Harjes, 1997; Emmermann and Lauterjung, 1997; Jost et al., 1998; Baisch and Harjes, 2003), and hydraulic fracturing for oil and gas production enhancement (Warpinski, 2009; Das and Zoback, 2011; Dusseault et al., 2011; Pettitt et al., 2011; Vermylen and Zoback, 2011; Boroumand and Eaton, 2012; van der Baan et al., 2013; Bao and Eaton, 2016) and during the stimulation of deep geothermal boreholes (Parker, 1989; Jupe et al., 1992; Cornet and Scotti, 1993; Tezuka and Niitsuma, 2000; Asanuma et al., 2005; Evans et al., 2005a; Häring et al., 2008; Brown et al., 2012; Baisch et al., 2015). Well-documented hydraulic stimulation datasets generally include microseismic observations as well as injection pressures and flow rates and, occasionally, tilt monitoring (Evans, 1983; Warpinski and Teufel, 1997). Although much information can be gained from these datasets, including imaging of microseismic structures (Niitsuma et al., 1999; Maxwell, 2014), energy balance between injected fluids and seismic energy release (Boroumand and Eaton 2012; Zoback et al., 2012; Warpinski, 2013), and source mechanisms (Jupe et al., 1992; Deichmann and Ernst, 2009; Warpinski and Du, 2010; Horálek et al., 2010), the constraints placed on the processes are insufficient to resolve details of the hydromechanical processes that underpin permeability enhancement, flow-path linkage, channeling or the interaction with natural fractures. Many of these processes possibly also depend on rock type. For instance, case studies analyzed by Evans et al. (2012) support the notion that injection into sedimentary rock tends to be less seismogenic than in crystalline rock. Moreover, it is likely that a significant part of the permeability creation processes take place in an aseismic manner (Cornet et al., 1997; Evans, 1998; Guglielmi et al., 2015b; Zoback et al., 2012), implying that seismic monitoring may only illuminate parts of the stimulated rock volume. In many deep hydraulic stimulation projects the rock mass is only accessed by one or at most a few boreholes, and the structural and geological models of the reservoir are not well defined. In general, the displacements on fractures arising from the injection can only be directly measured where they intersect the boreholes, and deformation occurring within the rock mass is poorly resolved.
Despite limitations in reservoir characterization and monitoring, significant
insights into the stimulation process can be gleaned from the experience from
the EGS projects that have been conducted to date. Two examples in
crystalline rock are studies of stimulation-induced fault slip and changes in
flow conditions in the fracture network associated with the permeability
creation processes at the Soultz-sous-forêt (Cornet et al., 1997; Evans
et al., 2005b) and the Basel EGS projects (Häring et al., 2008). At both
sites, it has been shown that permeability in the near-wellbore region
increased by 2–3 orders of magnitude. At Basel, a single
initially impermeable fracture has been shown to take at least 41 % of
the flow during the 30 L s
An additional, important lesson from deep stimulation projects is that the stress conditions in reservoirs may be strongly heterogeneous and that this influences the flow field (e.g., Hickman et al., 2000). For instance, profiles of horizontal stress orientation defined by wellbore failure observations commonly show significant fluctuations whose amplitude varies systematically with scale (Shamir and Zoback, 1992; Valley and Evans, 2009; Blake and Davatzes, 2011), even though that may have an average trend consistent with the tectonic stress field. Strong deviations may occur in the vicinity of faults, indicating past fault slip and complex fault zone architecture (Valley and Evans, 2010; Hickman et al., 2000). Similarly, the hydromechanical properties of faults depend on the fault architecture, which itself depends on lithology and the damage history accumulated over geological time (Caine et al., 1996; Faulkner and Rutter, 2008; Guglielmi et al., 2008; Faulkner et al., 2010; Jeanne et al., 2012). Within a fault zone, permeability and compliance contrasts can vary by several orders of magnitude (Guglielmi et al., 2008), thus complicating the predictability of hydromechanical responses to stimulations. In some EGS projects, it was observed that the hydraulic communication between injection and production boreholes may be unsatisfactory for efficient exchange of heat, either because of high flow impedance, such as in granite rock at Ogachi, Japan, (Kaieda et al., 2005), or because of flow channeling, as inferred from early thermal drawdown in granitic rock at Rosemanowes, UK (Nicol and Robinson, 1990) and in granodiorite at Hijiori, Japan (Tenma et al., 2008).
On the laboratory scale, considerable effort has been devoted to experiments
that address the role of effective stress changes in normal fracture opening
and closure, shear dilatancy and related permeability changes (Goodman, 1974;
Bandis et al., 1983; Yeo et al., 1998; Esaki et al., 1999; Gentier et al.,
2000; Olsson and Barton, 2001; Samuelson et
al., 2009). These experiments have demonstrated that the relationships
between fluid pressure change, fracture opening and flow within rough natural
fractures are strongly nonlinear. Even though significant progress has been
made on defining permeability changes during normal opening and shear slip on
the laboratory scale, the nonlinear relationships between fracture opening,
changes in effective normal stress, shearing and the resulting permeability
are not yet well constrained (Esaki et al., 1991; Olsson and Barton,
2001;
Vogler et al., 2015). One common approach is to represent the fracture as two
parallel plates whose separation, the hydraulic aperture, gives the same flow
rate per unit pressure gradient as would apply for the natural fracture. For
parallel plates and laminar flow, the flow rate per unit pressure gradient is
proportional to the cube of hydraulic aperture. However, for rough-walled
fractures, the hydraulic aperture,
Dilatancy associated with shearing is often expressed in terms of a dilation angle, which is a property describing the relationship between mean mechanical aperture and slip. Dilation angle depends on the fracture surface characteristics, the effective normal stress and the amount of slip. The dependence of dilation on effective normal stress is particularly important within the stimulation context, the dilation angle tending to decrease at higher effective normal stress, in large part because shorter wavelength asperities are sheared off (Evans et al., 1999). Thus, shearing-induced dilation is likely to be more effective at low effective normal stress, such as in the near field of the injection where fluid pressures are relatively high. Clearly, insights from laboratory experiments into the relationships describing fracture dilation and permeability changes are important for understanding field observations in EGS reservoirs (e.g., Robinson and Brown, 1990; Elsworth et al., 2016; Fang et al., 2018) and also for parameterizing numerical models.
In situ experiments on the intermediate scale (i.e., decameter scale) serve as a vital bridge between laboratory and reservoir scales. As such, they can contribute to an improved understanding of reservoir behavior during stimulation and enable upscaling of hydromechanical information obtained from laboratory experiments (Jung, 1989; Martin et al., 1990; Rutqvist, 1995; Schweisinger et al., 2007; Cornet et al., 2003; Murdoch et al., 2004; Cappa et al., 2006; Derode et al., 2013; Guglielmi et al., 2014, 2015a, b). Much experience has been gained from stress testing using the hydraulic methods of hydro-fracturing (HF), hydraulic testing of preexisting fractures (HTPF) (Haimson and Cornet, 2003) and hydro-jacking (Evans and Meier, 1995; Rutqvist and Stephansson, 1996). Hydraulic tests have been commonly used to quantify pressure-sensitive permeability changes (Louis et al., 1977) and normal stiffness in natural fractures or faults (Rutqvist, 1995). Evans and Wyatt (1984) estimated the closure of a fracture zone from observed surface deformations induced by drilling-related drainage of fluid pressure within the structure. Similarly, Gale (1975), Jung (1989), Martin et al. (1990), Guglielmi et al. (2006) and Schweisinger et al. (2009) used borehole caliper sondes to monitor changes in fracture aperture and pressure during hydraulic jacking tests. The resulting displacements and the flow and pressure responses allowed relationships between mechanical and hydraulic aperture changes to be established and helped to constrain the fracture/fault normal compliance on larger scales.
Irreversible permeability increases arising from slip-induced dilation of natural fractures are particularly relevant for the stimulation of EGS and hydrocarbon reservoirs. To study the phenomenon in situ, Guglielmi et al. (2014) developed a novel double packer system (Step-Rate Injection Method for Fracture In-Situ Properties, SIMFIP) that allows the simultaneous measurement of pressure, flow rates and three-dimensional relative displacements occurring across a fracture isolated within the interval in response to injection. The device was successful in reactivating a fault zone in a limestone formation in southeast France (Derode et al., 2013; Guglielmi, et al., 2015a, b). Pressure, injection rate and 3-D displacements in the SIMFIP interval were measured, together with microseismic activity, tilt and fluid pressure in the vicinity of the injection borehole. The dataset is unique and provided quantitative insights into the relationships between (i) fault dislocation including shear and permeability changes, (ii) fault normal compliance and static friction, and (iii) slip velocities and magnitudes and their relation to aseismic and seismic slip. Recently, a similar experiment was conducted in a series of interacting complex fault zones in shale (Guglielmi et al., 2015a, b). Distributed pore pressure and strain sensors across the faults allowed the evolution of the pressurized and slipped areas to be constrained, which was not previously possible. Such experiments provide a useful methodology for advancing our understanding of the hydromechanically coupled processes in complex faults.
Experience gained from large-scale stimulation of EGS reservoirs in crystalline rock suggests that hydraulic shearing is the dominant mechanism for permeability creation, at least several tens of meters distance from the injection point (e.g. Evans, 2014). However, the initiation and propagation of hydraulic fractures may be an important mechanism in the near field of the wellbore to connect the wellbore to the preexisting fracture network in the reservoir (Cornet and Jones, 1994). Considerable effort has been devoted to understanding the initiation and propagation of hydraulic fractures on both the laboratory and intermediate field scale.
Many well-controlled, small-scale laboratory experiments on hydro-fracture are documented in the literature (Jaeger, 1963; Zoback et al., 1977; Warpinski et al., 1982; Bruno and Nakagawa 1991; Johnson and Cleary 1991; Song et al., 2001; Jeffrey and Bunger, 2007; Bunger et al., 2011). For such experiments, samples of various shapes (e.g., hollow cylinders and perforated prisms) are loaded along their boundaries and the internal fluid pressure is increased until a hydraulic fracture initiates and propagates. For some tests, transparent materials like polymethyl methacrylate (PMMA) were used to image fracture growth. Some experimental setups include multi-material “sandwiches” to study the effect of stress contrast on hydraulic fracture containment (Jeffrey and Bunger, 2007; Warpinski et al., 1982). Others study the interaction of propagating hydro-fractures with preexisting fractures (Zoback et al., 1977; Meng, 2011; Hampton et al., 2015) or rock textures (Ishida, 2001; Chitrala et al., 2010), the impact of injection fluids with different viscosities (Bennour et al., 2015) or the role of stress anisotropy (Doe and Boyce, 1989) on the geometry and orientation of generated fractures or the interaction between multiple fractures (Bunger et al., 2011). These laboratory studies provide important results relevant for EGS. For instance, in the common situation where a family of natural fractures in not normal to the minimum principal stress, injections with high-viscosity fluids (viscosity-dominated regime) may help maintain tensile fracture propagation normal to the minimum principal stress despite the presence of cross-cutting fractures (Zoback et al., 1977), whereas low-viscosity fluids (toughness-dominated regime) such as water will promote leak-off into the cross-cutting natural fractures, whose permeability may be increased by shear (Rutledge et al., 2004). This leak-off will tend to limit hydro-fracture propagation. Laboratory studies also give insights into the influence of shear stress shadow and transfer on hydraulic fracture growth (Bunger et al., 2011). Laboratory tests have also been essential for providing well-controlled fracture initiation and propagation datasets to benchmark hydraulic fracture simulation codes (Bunger et al., 2007).
Intermediate-scale experiments have been performed to study the initiation and propagation of hydraulic fractures. Typically, they are conducted from boreholes drilled from excavations to facilitate dense near-field instrumentation and secure good experimental control. An early example is the series of experiments that took place at the Nevada Test Site in soft, bedded volcanic tuff with high porosity and high permeability (Warpinski, 1985; Warren and Smith, 1985). The pressure, flow and fracture aperture were monitored during the experiments, and the fractures were mined back at the end of the experiments. The mine-back revealed that stress contrasts were the predominant influence on hydraulic fracture containment and that the fractures consisted of multiple fracture strands and thus differed significantly from simple shapes assumed in theoretical studies. This complexity of the fracture shape impacts the flow and pressure distribution within the propagating hydraulic fractures. Another notable series of in situ tests on hydraulic fracture propagation within the context of coal-seam mining and block cave mine preconditioning have been performed by the hydraulic fracture group of CSIRO (Chacón et al., 2004; Jeffrey et al., 1993, 1992, 2009; Jeffrey and Settari, 1995; van As et al., 2004; van As and Jeffrey, 2002, 2000). The block cave mining experiments were performed in hard rock media and thus are more relevant to EGS. Those conducted in the quartz monzonite porphyries at the Northparkes mine in Australia are probably the most detailed and densely instrumented tests executed to date and included tiltmeter monitoring, a microseismic network and pore pressure sensors as well as detailed rock mass and stress characterization (Jeffrey et al., 2009). Hydro-fractures were formed with water and cross-linked gels, with colored plastic proppants added in order to facilitate their identification once the test volume was mined back. The mapped trajectories of the hydraulic fractures exhibited complex geometries, sometimes with multiple branching and crossing of joints, veins and shear zones, with and without offset. Subparallel propped sections accounted for 10 to 15 % of the total fracture extent, which microseismic activity indicated was more than 40 m from the injection point. The results demonstrate that the geometry of the fractures is much more complex than typically obtained in small-scale laboratory experiments in a homogeneous material and uniform stress field. The complexity close to the injection point is controlled by the near-well stress perturbation and the interaction with natural fractures and rock mass fabric.
Natural fractures also have a strong influence on the propagation of hydraulic fractures. The propagation regime (i.e., viscosity-dominated or toughness-dominated; Detournay, 2016) can be controlled by the injection rate and injected fluid rheology and will have likely a strong influence on the interaction with natural fractures and the final complexity of the hydraulic fractures, although this has not been validated by in situ experiments. Another relevant aspect that has not been investigated with in situ tests is the problem of proppant transport and distribution within the created fractures. Indeed, in the case of hydraulic fractures, the self-propping mechanism, which results in a permanent aperture increase, is unlikely to be effective, and so proppant placement is necessary for ensuring permanent permeability enhancement. Finally, the nature of the microseismicity generated by hydraulic fracturing is not adequately understood. Moment tensor analyses can offer insight into the nature of the failure in a microseismic event (Warpinski and Du, 2010; Eyre and van der Baan, 2015). For example, they can help resolve whether the seismic radiation is primarily generated by shear on preexisting fractures that are intersected by the propagating fracture, with relatively little energy generated by the advancing mode 1 tip of the hydraulic fracture (Sileny et al., 2009; Horálek et al., 2010; Rutledge et al., 2004).
Injection of fluid into a rock mass invariably leads to the deformation of
the surrounding rock mass due to poroelasticity (Biot, 1941) or slip-related
stress changes (McClure and Horne, 2014). Numerical studies have suggested
that stress interaction between adjacent fractures can have a significant
impact on the stimulation results (e.g., Preisig et al., 2015; Gischig and
Preisig, 2015). In most reservoir stimulations, the microseismic clouds
exhibit an oblate shape, due primarily to the interaction between the
strongly anisotropic stress field with the natural fracture population. This
tendency to form an oblate ellipsoidal shape instead of a sphere may also be
promoted by stress transfer from slipped fractures, which tends to inhibit
slip on neighboring fractures (Gischig and Preisig, 2015). Schoenball et
al. (2012) and Catalli et al. (2013) have demonstrated that induced
earthquakes preferably occur where stress changes generated by preceding
nearby earthquakes render the local stress field to be more favorable for
slip. Similar effects have been observed for natural earthquakes (Stein,
1999). The effect becomes more important during stimulation as time goes on,
especially at the margin of the seismicity cloud. Direct observation of
deformation associated with fluid injection has been observed in several
intermediate-scale in situ experiments. Evans and Holzhausen (1983) report
several case histories of using tiltmeter arrays to observe ground
deformation above high-pressure hydraulic fracturing treatments. The results
show clear evidence of the self-propping of the induced fractures (van As et
al., 2004). Jeffrey et al. (2009) used a tiltmeter array to monitor a
hydro-fracturing treatment at the Northparkes mine in Australia. The pattern
of tilting indicated that the induced fracture was subhorizontal, which was
confirmed by excavating the fracture traces. Evans and Wyatt (1984) modeled
strains and tilts occurring around a well during air drilling and found that
the deformation was due to the opening of a preexisting fracture zone in
response to fluid pressure changes. Derode et al. (2013) observed tilts of
10
Rock mass deformation during stimulation injections necessarily leads to stress changes in the rock mass. Small but nonzero residual stress changes induced by hydraulic fracturing were measured using a stress cell by van As et al. (2004). Stress changes during injections are recognized as playing a potentially important role in determining the pattern of fracture and slip that develops during the injection (e.g., Preisig et al., 2015; Catalli et al., 2013).
A significant fraction of the slip that occurs on fractures within a reservoir undergoing stimulation may be aseismic, depending upon in situ stress and geological conditions. That aseismic slip has occurred is often inferred indirectly from changes in the hydraulic characteristics of a reservoir without attendant microseismicity (Scotti and Cornet, 1994; Evans, 1998). Direct detection of aseismic slip is difficult as it requires relative displacements across fractures to be resolved from borehole or near-field deformation measurements (e.g., Maury, 1994; Cornet et al., 1997; Evans et al., 2005b). For example, Cornet et al. (1997) compared borehole geometry from acoustic televiewer logs run before and after the 1993 stimulation at the Soultz-sous-forêt site and found that 2 cm of slip had apparently occurred across a fracture. The cumulative seismic moment of events in the neighborhood of the fracture was insufficient to explain the observed slip magnitude, thereby suggesting that a large portion of the slip had occurred aseismically. Indeed, almost all fracture zones that were hydraulically active during the stimulation showed evidence of shear and opening-mode dislocations of millimeters to centimeters (Evans et al., 2005b).
The transition from aseismic to seismic slip was directly observed by
Guglielmi et al. (2015a) during fluid injection into a well-instrumented
fault in limestone in a rock laboratory at 280 m depth. Some 70 % of a
20-fold permeability increase occurred during the initial aseismic slip
period. The transition to seismic slip coincided with reduced dilation, and
the inference is that slip zone area exceeded the pressurized area,
suggesting that the events themselves lay outside the pressurized zone.
Modeling the observed slip as occurring on a circular fracture with total
stress drop gave a radius of 37 m and a moment release of
Apart from these observations, aseismic slip has been mostly discussed from the perspectives of semi-analytical or numerical models. Garagash and Germanovic (2012) used a slip-weakening model to show that aseismic slip depends on the stress conditions and injection pressure. Zoback et al. (2012) used McClure's (2012) rate-and-state friction model to show that aseismic slip becomes more prominent for stress states farther from the failure limit. Using the same model, Gischig (2015) demonstrated that slip velocity depends on fault orientation in a given stress field. For non-optimally oriented faults, aseismic slip becomes more prominent and the seismicity is less pronounced for lower slip velocities and shorter rupture propagation distances. These model results suggest that aseismic slip and low slip velocities may be promoted by avoiding the stimulation of optimally oriented critically stressed faults. Clearly, a more detailed understanding of the conditions that result in aseismic slip may be a basis for less hazardous stimulations.
Keeping induced seismicity at levels that are not damaging or disturbing to
the population continues to be a major objective for EGS (Giardini, 2009;
Bachmann et al., 2011; Majer et al., 2012; Evans et al., 2012) and other
underground engineering projects (oil and gas extraction, liquid waste
disposal, gas and CO
There are numerous analyses of induced seismicity at other EGS sites. Pearson (1981) and Phillips et al. (1997) analyzed microseismicity generated during the stimulation of the 2930 m deep “large Phase 1” and the 3460 m deep Phase 2 reservoirs, respectively, at the Fenton Hill EGS site, New Mexico. Batchelor et al. (1983) summarize microseismicity observed during the stimulation injections into the Phase 2a and 2b reservoirs at Rosemanowes in Cornwall, UK. Tezuka and Niitsuma (2000) examined clusters of microseismic events generated during the stimulation of the 2200 m deep reservoir at the Hijiori EGS site in Japan. Baisch et al. (2006, 2009, 2015) analyzed data from different stages of the stimulation of the Habanero EGS reservoir in the Cooper Basin, Australia. Calò et al. (2011) used microseismicity generated during the stimulation of the 5 km deep EGS reservoir at Soultz-sous-forêt to perform time-lapse P-wave tomography to infer pore pressure migration during injection. Various authors also explored the vast induced seismicity dataset of > 500 000 events recorded since the 1960s at the Geysers geothermal site, where recently also an EGS demonstration stimulation has been performed (Garcia et al., 2012; Jeanne et al., 2014). The observed seismicity was partly related to injections (Jeanne et al., 2015) and thermoelastic stress changes (Rutqvist and Oldenburg, 2008). Here, local variability in the stress field (Martínez-Garzón et al., 2013) and volumetric source components (Martínez-Garzón et al., 2017) were inferred from detailed analysis of injection-induced seismicity.
Another major focus of induced seismicity research has been the development of hazard assessment tools for injection-related seismicity. The primary goal of these efforts is to develop a dynamic, probabilistic and data-driven traffic light system that can provide real-time hazard estimates during injections (Karvounis et al., 2014; Király-Proag et al., 2016), as opposed to the traditional, static traffic light system (Bommer et al., 2006). Bachmann et al. (2011) and Mena et al. (2013) developed several statistical models and tested them in a pseudo-prospective manner using the Basel seismicity dataset. More complex models including physical considerations and stochastic processes (so-called hybrid models) were developed to include information on the reservoir behavior and from the spatiotemporal evolution of seismicity (Goertz-Allmann and Wiemer, 2013; Gischig and Wiemer, 2013; Kiràly et al., 2018). Mignan et al. (2015) evaluated reported insurance claims arising from the Basel induced seismicity in order to infer procedures for evaluating risk based on induced seismic hazard estimates.
The Gutenberg–Richter
Whilst the hazard associated with induced seismicity is clearly an important factor for reservoir engineering, it should not be forgotten that the shearing of fractures and fracture zones, which is the source of the seismicity, is a key process in the irreversible permeability enhancement that is the objective of the stimulation injections. Furthermore, precise mapping of the 3-D distribution of events provides an indication of the direction of fluid pressure propagation and hence the geometry (i.e., size, shape, degree of anisotropy) of the distribution of permeability enhancement – information that is vital for drilling a subsequent well (Niitsuma et al., 1999). Managing induced seismic hazard also requires considering the design of reservoir attributes such as size, system impedance and heat exchanger properties that control system longevity (e.g., Gischig et al., 2014). Currently, few case studies consider both seismicity and the related changes that occurred in the reservoir (e.g., Evans et al., 2005a), and relatively few studies even report both permeability changes or well injectivity (e.g., Häring et al., 2008; Evans et al., 2005b; Kaieda et al., 2005; Petty et al., 2013). More work is needed to quantitatively link the spatial, temporal or magnitude distribution of seismicity with the thermo-hydraulic–mechanical properties of the rock mass under stimulation conditions. We believe controlled experiments on the intermediate (in situ test site) scale supported by laboratory-scale experiments could be key in making progress towards this end.
Research on reservoir stimulation for deep geothermal energy exploitation has been largely performed through laboratory observations, large-scale projects and numerical models. Observations of full-scale reservoir stimulations have yielded important observations. However, the difficulty in observing the processes occurring within the reservoir under stimulation conditions severely limits the understanding of the permeability creation processes in a way that aids future stimulation design.
Laboratory experiments are attractive because they are controllable and readily repeatable, but they suffer from two main limitations. (1) Upscaling results to the field scale is affected by large uncertainties (Gale, 1993). Although there is evidence that the roughness of fresh fracture surfaces obeys well-defined scaling over many orders of magnitude (Power and Tullis, 1991; Schmittbuhl et al., 1995), complications arise in upscaling the aperture distribution and hence permeability of two semi-mated rough surfaces due to the effects of damage and wear of the asperities during shearing and gouge formation (Amitrano and Schmittbuhl, 2002; Vogler et al., 2016). (2) Laboratory tests are typically performed on single fractures in relatively homogeneous materials and uniform stress conditions, which makes upscaling to structures with multiple fractures such as fracture zones challenging. Similarly, hydraulic fracture propagation behavior is usually studied with homogeneous rock samples under uniform stress, and this can lead to an oversimplistic fracture flow and/or hydraulic fracture propagation behavior. In an EGS reservoir, for example, the stress may be heterogeneous on the meter to decameter scale (Evans et al., 1999; Valley and Evans, 2009; Blake and Davatzes, 2011) and the rock mass may contain various heterogeneities such as stiffness contrasts, fractures or faults (Ziegler et al., 2015).
Because of the large uncertainties in upscaling, many numerical studies make direct (i.e., not upscaled) use of laboratory results to parameterize HM-coupled models for EGS because so few field-scale relationships are available (e.g., Rutqvist, 2011; McClure, 2012; Gischig et al., 2014). This impacts the reliability of the numerical simulation studies because the descriptions of the processes and the input parameter values may be inappropriate for the scale of the simulation.
Clearly there is a need for field-scale hydraulic stimulation experiments
that bridge the various scales and are performed with the target rock mass
equipped with a comprehensive monitoring system to capture details of the
processes. Recently several intermediate-scale hydro-shearing and
hydro-fracturing experiments have been performed in a densely instrumented
rock mass (i.e., Guglielmi et al., 2008, 2014, 2015a, b; Jeffrey et al.,
2009). The hydro-shearing experiments by Guglielmi et al. (2008) have all
been in sedimentary rock types at shallow depth. No such densely instrumented
experiments have been performed in fractured and faulted crystalline basement
rocks faults, the target rocks for most EGS, where a variety of complex fault
architectures and stress-fracture system configurations need to be
investigated. The In situ Stimulation and Circulation (ISC) experiment
addresses these research gaps, with a focus on the following research
questions (RQs).
RQ1: What is the relationship between pressure, effective stress, fracture
aperture, slip, permeability and storativity (i.e., the hydromechanically coupled response of fractures)? RQ2: How does the transient pressure field propagate in the reservoir during
stimulation? RQ3: How does the rock mass deform as a result of rock mass pressurization,
fracture opening and/or slip? RQ4: How does stress transfer inhibit or promote permeability enhancement and
seismicity along neighboring fractures? RQ5: Can we quantify the transition between aseismic and seismic slip and the
friction models (such as rate-and-state friction) describing slip evolution
and induced seismicity? RQ6: How do hydraulic fractures interact with preexisting fractures and faults
and how can the interaction be controlled? RQ7: How does induced seismicity evolve along faults and fractures of different
orientation? RQ8: How does induced seismicity along stimulated faults compare to induced
seismicity along newly created hydraulic fractures? RQ9: Can we quantify the link between spatial, temporal and magnitude
distribution of induced seismicity and HM-coupled properties of fractures
and faults?