Geomechanical effects during large-scale underground injection

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239731433 Geomechanical effects during large-scale underground injection CONFERENCE PAPER · JUNE 2013 READS 101 4 AUTHORS: Antonio Pio Rinaldi Pierre Jeanne 48 PUBLICATIONS 314 CITATIONS 27 PUBLICATIONS 130 CITATIONS ETH Zurich Lawrence Berkeley National Laboratory SEE PROFILE SEE PROFILE Jonny Rutqvist Frederic Cappa 251 PUBLICATIONS 4,172 CITATIONS 109 PUBLICATIONS 1,020 CITATIONS Lawrence Berkeley National Laboratory SEE PROFILE University of Nice-Sophia Antipolis SEE PROFILE Available from: Pierre Jeanne Retrieved on: 05 February 2016 ARMA 13-255 Geomechanical effects on CO2 leakage through fault zones during large-scale underground injection Rinaldi A. P., Rutqvist J., and Jeanne P. Lawrence Berkeley National Laboratory, Earth Sciences Division, Berkeley, CA, USA Cappa F. Lawrence Berkeley National Laboratory, Earth Sciences Division, Berkeley, CA, USA GeoAzur, University of Nice Sophia-Antipolis, Côte d'Azur Observatory, Nice, France Copyright 2013 ARMA, American Rock Mechanics Association th This paper was prepared for presentation at the 47 US Rock Mechanics / Geomechanics Symposium held in San Francisco, CA, USA, 23-26 June 2013. This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and critical review of the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of ARMA, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 200 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgement of where and by whom the paper was presented. ABSTRACT: The importance of geomechanics including the potential for reactivating faults associated with large-scale geologic carbon sequestration operations has recently become more widely recognized. However, not withstanding the potential for triggering notable (felt) seismic events, the potential for buoyancy-driven CO2 to reach potable groundwater and the ground surface is more important from safety and storage-efficiency perspectives. In this context, this work extends previous studies on the geomechanical modeling of fault responses during underground carbon dioxide injection, focusing on short-term integrity of the sealing caprock, and hence of potential leakage of either brine or CO2 to shallow groundwater aquifers during active injection. We account for a stress/strain-dependent permeability and study the leakage through a fault zone as its permeability changes during a reactivation, also causing seismicity. We analyze several scenarios related to the injected amount of CO2 (and hence as a function of the overpressure) both involving minor and major faults, and analyze the profile risks of leakage for different stress/strainpermeability coupling functions. We conclude that whereas it is very difficult to predict how much fault permeability could change upon reactivation, this process can have a significant impact on the leakage rate. 1. INTRODUCTION Public concerns always arise when dealing with exploitation of underground natural sources. Ground deformation, induced seismicity and fracture reactivation, as well as sealing capacity of the rocks surrounding a reservoir are general issues for public acceptance of projects involving the injection of fluids (e.g. disposal water or carbon dioxide) into the underground. The potential for injection-induced fault reactivation associated with industrial injection activities is an important issue, not just from a safety point of view, but also from a public acceptance perspective [1]. The correlation between fluids and seismicity is an issue that has been largely studied [2]. Although in natural seismicity is hard to discriminate between fluid contribution and regional regime of stress, there are few examples in literature relating overpressurized fluids to local seismic events, such as the case of Basel, where an earthquake of magnitude 3.4 occurred [3], or during disposal of water into the Ozark aquifer of Arkansas, where few earthquakes of magnitude greater than 4 where felt by the local population [1]. Although the carbon capture and storage (CCS) has been recognized as a promising option for reducing the CO2 emission into the atmosphere [4], there are concerns related to the potential for triggering notable (felt) seismic events and how such events could impact the long-term integrity of a CO2 repository (as well as how it could impact the public perception of geological carbon sequestration) [5]. A state-of-art review can be found in Rutqvist [6], which describes the effects of CO2 injection in a deep sedimentary basin as reservoir stress-strain and microseismicity, caprock sealing performance, and the potential for fault reactivation. Moreover, several studies have recently shown how CO2 injection may produce seismic events both on major faults (i.e. 2 km-long fault with initial offset [7-9]) and on minor faults (undetected, i.e. less than 1 km-long and without detected initial offset [10]). However, not withstanding the potential for triggering notable (felt) seismic events, the potential for buoyancy-driven CO2 to reach potable groundwater and the ground surface is more important from safety and storage-efficiency perspectives. In this context, this work extends previous studies on the geomechanical modeling of the effects of CO2 injection with the TOUGH-FLAC simulator [11]. Here, we study Fig. 1. Simulated scenarios with initial and boundary conditions. (a) Configuration for an undetected fault, 1 km long with no offset [10]. Figure also shows the orientation of the considered stresses in this 2D models; (b) Configuration for a 2 km-long fault, with 100 m offset [7 - 9]. the fault response during active underground CO2 injection, focusing on short-term (5 years) integrity of the CO2 repository, and hence of potential leakage of CO2 to shallow groundwater aquifers. Increased pore pressure of course alters the stress distribution on a fault/fracture zone, and at the same time may produce changes in the permeability, related to the elastic and/or plastic strain, stress, and it may increase when single (or multiple) shear rupture occurs. We account for a stress/strain-dependent permeability and study the leakage through the fault zone as its permeability changes along with strain and stress variations. Simultaneously, we study the reactivation of the fault itself. We analyze several scenarios related to the injected amount of CO2 (and hence as a function of the overpressure) and to the initial fault zone permeability both involving minor and major faults, and analyze the risk of leakage and fault reactivation for different stress/strain-permeability coupling functions. We conclude that whereas it is very difficult to predict how much fault permeability could change upon reactivation, this process can have a significant impact on the leakage rate and induced seismicity. 2. MODEL SETUP In this section we discuss the modeling approach we used to estimate the effects of different CO2 injection rates on the upper aquifer leakage and the potential for fault reactivation for two different scenarios. Scenario #1 is basically the one presented by Mazzoldi et al. [10], with a minor 1 km-long fault zone intersecting a 100 m-thick injection aquifer bounded by a 150 m-thick low permeability caprock (Fig. 1a). This scenario aims to represent a fault that is difficult to be detected by a seismic survey, since it would have a shear offset less than 10 m. Scenario #2 was first presented by Cappa and Rutqvist [7, 8], and it represents a very large fault zone (greater than 2 km-long). Also in this case the fault zone intersects a 100 m-thick reservoir confined on the upper and bottom parts by a 150 m-thick caprock. However, with this scenario we want to simulate a major, "easy" to detect fault zone, which has an offset of 100 m (Fig. 1b). This scenario was also recently used for the simulation of the dynamic behavior of fault reactivation during underground CO2 injection [9]. We analyzed the short-term fault response during active CO2 injection in terms of displacement and as a pathway for fluid leakage, and then we applied seismological theories to estimate the corresponding seismic magnitude [19, 20]. We simulate different rates of CO2 injection for both the scenarios. Moreover, since changes in permeability and porosity may occur, we simulate three different cases of coupling between mechanical and hydraulic properties. Following their approach, changes in porosity are caused by plastic deformation: 2.1. Mechanical effects on porosity and permeability Hydraulic properties of a porous medium may change as the pressure and stresses evolve. Depending on the fault (and fracture reactivation), changes in hydromechanical properties may be isotropic or anisotropic. Isotropic changes can be assumed in a non-fractured fault core or in the damage zone when highly fractured without any preferential direction. In such cases, permeability and porosity changes may be simply related to changes in volumetric strain or mean stress. We will evaluate the effects changes in permeability and porosity on fault reactivation and CO2 leakage after few years of injection accounting for two different isotropic models. The first case relates the porosity (φhm) to the mean stress (σ'M), and then the permeability (κhm) depends on the porosity changes. The formulation was first derived by Davies and Davies [12] and then modified for carbon sequestration application by Rutqvist and Tsang [13]: (1) where subindex 0 refers to the initial unstressed value (for both porosity and permeability, φ0 and κ0), and φr is the residual porosity at high stress. We applied the changes to the fault zone only. The second case relates the porosity to the isotropic volumetric strain variation, and again the permeability is then related to the porosity changes. This model was develop and applied by Chin et al. [14] for modeling of permeability changes in petroleum reservoir and then modified by Cappa and Rutqvist [8] for the permeability changes in a fault zone after reactivation because of CO2 underground injection: (2) where φhm and κhm are the porosity and the permeability at a given stress, φ0 and κ0 are the initial porosity and permeability, respectively, and εV is the volumetric strain. The relation between the porosity and the total volumetric strain accounts for both elastic and plastic behavior that may occur during fault reactivation. In a fractured fault core, changes in permeability and porosity may be extremely anisotropic, depending on fracture direction. Then, the hydraulic parameters depend on anisotropic elasto-plastic properties: the permeability may depend on both the fault normal stress and on the plastic shear and tensile strain. Hsiung et al. [15] derived the relation between these parameters and porosity and permeability of a fracture. Here we apply the same formulation for a fault zone. (3) where Δφfp are the changes in porosity. eftp and efsp are the plastic strains caused by tensile and shear deformation, respectively. ψ represents the fault dilation. The permeability changes are then based on a non-linear function of the normal effective stress (σ'n), as well as depending on the plastic strain: (4) a and c are two empirical constants for normal-closure hyperbola [16] that can be approximated from the initial stiffness (K) and from the initial normal effective stress (σ'n0). In addition to two coupling equation for porosity and permeability, the capillarity pressure (pc) varies according to a function by Leverett [17]: (5) where pc0(Sl) is the unchanged capillarity pressure, which depends on the liquid saturation (Sl). This equation for the capillarity pressure is applied for each of the three cases analyzed. 2.2. Seismic event modeling and magnitude estimation Following previous studies, the fault zone is simulated using a ubiquitous-joint fractured media [8]. This approach allows representing strongly anisotropic behavior, accounting for the presence of an orientation of weakness (fault plane) in a Mohr-Coulomb solid. In a fault with a given orientation, the Mohr-Coulomb criterion for failure can be written as [18]: (6) where τ is the critical shear stress (i.e. shear strength) necessary for slip occurrence, c is the cohesion, and µs is the static friction (µs=tanϕ, where ϕ is the friction angle). For most rocks the static friction ranges between 0.6 and 0.85 [8, and references therein]. In order to allow us to model a sudden slip (i.e. seismic event), we used a Mohr-Coulomb model with strain Table. 1. Mechanical and hydraulic properties used in the numerical modeling for both scenario #1 and #2 for each domain. Listed porosity and permeability for the fault zone represent the initial non-stressed/strained value. Parameters Young’s modulus, E (GPa) Poisson’s ratio, ν (-) Rock density, ρs (kg/m3) Joint peak friction angle, φ (°) Joint residual friction angle φ (°) Dilation angle, ψ (°) Porosity, φ0 (-) Permeability, κ0 (m2) Upper Aquifer 10 0.25 2260 0.1 10-14 softening frictional strength properties, consistent with a seismological slip-weakening fault model. In our model the frictional coefficient varies from a static value of 0.6 to a value of 0.2 when the strain on the fault zone is greater than a certain critical value (10-3), and a rupture occurs. Following the approach used by Cappa and Rutqvist [7-9] and Mazzoldi et al. [10], the magnitude of a seismic event is estimated using seismological theories. The seismic moment (M0) is first estimated for a ruptured patch on a fault following the well-known equation [19] (8) where G is the rigidity of the medium (Pa), A is the rupture area (m2) and d (m) is the average slip along the fault zone occurring when the shear-stress drops and the frictional coefficient changes [7-10]. In our 2D-model, the rupture area (A) is consider to be circular with diameter equal to the rupture length simulated along the fault line in the 2D-plane. The rupture extends along the fault line, and then the rupture area is considered normal to the 2D-plane. The seismic magnitude is estimated by the equation proposed by Kanamori and Anderson [20] as: (9) where the seismic moment M0 is in Nm. In our modeling we are able to distinguish between the coseismic fault slip (d), which is used to give an estimate of the seismic event, and the aseismic slip, which may produce a larger displacement on the fault plane, but is not causing an earthquake. 2.3. Numerical model and conditions Numerical simulations were carried out using the coupled fluid flow and geomechanical simulator TOUGH-FLAC [11] based on the multiphase, multicomponent fluid flow and heat transport simulator TOUGH2 [21] and on the geomechanical simulator FLAC3D [22]. In both the scenarios considered in this work (Fig. 1), the fault zone is simulated as constituted by a fault core bounded by a damage zone, which Central Aquifer 10 0.25 2260 0.1 10-13 Basal Aquifer 10 0.25 2260 0.01 10-16 Caprock Fault zone 10 0.25 2260 0.01 10-19 5 0.25 2260 31 11 10 0.1 10-16 - 10-14 corresponds to a more permeable zone with macroscopic fracture network [7, and references therein]. For the aim of this work we simulated the fault core using a ubiquitous joint model, with finite thickness elements having anisotropic elasto-plastic properties (Table 1) and intensely jointed along a direction parallel to the fault plane, thus permitting the shear failure to occur along the fault itself. The damage zone is simulated as a poroelastic medium (no slip occurs within the damage zone) with finite thickness elements having the same high permeability as the fault core and that can be subject to permeability changes due to the variation of stress and/or strain. Such a fault zone (damage zone plus fault core) intersects a 2D plane-strain multilayer system (2 km x 2 km) with a dip angle of 80˚ and a length of 1 or 2 km (according the selected scenario). The storage formation is 100 m-thick and bounded at top and bottom by a low-permeability 150 m-thick formation, which, in turn, is surrounded by two other aquifers. Hydraulic and mechanical properties for the different layer (aquifers and caprock) are equal for both scenarios and are listed in Table 1. These properties were kept constant during all the simulations performed, with the exception of the fault zone ones. We set the initial conditions as a linear pore pressure and temperature gradient (9.81 MPa/km and 25 ˚C/km, respectively) with atmospheric pressure of 0.1 MPa and temperature of 10 ˚C at the ground surface, resulting then in a pore pressure of 5 MPa and temperature of 22.5 ˚C at the top of our model (at 500 m depth). One of the most important parameter related to the reactivation of a fault zone is the initial in situ stress [7,10]. Mazzoldi et al. [10], showed, for example, as the maximum earthquake magnitude for an undetected fault (1000 m-long) may change from 2.7 to 3.5 when the stress ratio between horizontal (minimum) and vertical (maximum) stress varies in the range of few percent (from 0.7 to 0.65). For all the simulations in this study, we kept a stress ratio (σh/σV) of 0.7, which is already a critical value for a fault striking along the minimum horizontal stress, but may prevent in some cases the rupture to extent for the Fig. 2. Percentage of CO2 leaking into the upper aquifer as a function of permeability and injection rate for a 1 km-long fault with no offset. Contour lines represent the 1% and 5% leakage rate. (a) Case1: fault permeability changes as function of mean stress. (b) Case2: fault permeability changes as a function of the volumetric strain. The white contour indicates the region where the injection pressure exceeded 35 MPa. (c) Case3: fault permeability changes as a function of both fault normal stress and plastic shear and tensile strain. entire length of the fault zone [7]. Hence we set the vertical stress gradient to 22,148 Pa/m and then the corresponding horizontal stress gradient as 15,504 Pa/m. Boundaries were open for fluid flow (i.e. at constant pressure and temperature), except for the left boundary, where no flow occurred. The simulations were conducted in an isothermal mode, implying that the temperature gradient is maintained during the simulation. Null displacement conditions were set normal to the left and bottom boundaries, whereas constant stress was imposed normal to the right and top boundaries (Fig. 1). Two other critical parameters in evaluating the shortterm leakage and fault reactivation are the fault permeability and the CO2 injection rate. For both scenarios, the initial permeability of the fault zone (fault core and damage zone) was varied in the range from 10-16 to 10-14 m2. Although sometimes the permeability of the fault core may result much smaller ranging from 10-17 to 10-21 m2, the values we accounted for are a good estimate of the damage zone permeability [23-24], which mostly affect a leakage along a fault zone. The Young's modulus of the fault zone was set to 5 GPa [23-24]. The second critical parameter is the injection rate. The amount of CO2 injected may vary from site to site. For example, at the In Salah (Algeria) CO2 storage project, the injection occurred over three horizontal wells at a rate of about 0.5-1.0 millions tons/yr, which correspond to an average injection rate per well of about 10-15 kg/s [25]. In our simulations for scenario #1 (Fig. 1a), CO2 is injected as a point source at 1500 m depth, with a constant rate ranging from 0.002 to 0.1 kg/s/m, which for a horizontal well 1000 m-long would correspond to an injection rate ranging from a reasonable low value of 2 kg/s up to a very big value of 100 kg/s (comparable with injection rates used during shale gas hydraulic fracturing [26]). For scenario #2 (Fig. 1b), however, we needed to decrease the range from 2 to 12 kg/s, because the fault offset of 100 m results in a confined CO2 reservoir bonded on the right by the offset part of caprock. A higher injection rate would result in a pore pressure increase up to unrealistic values, and since pressure should be one of the parameter monitored during injection, a project would stop at a very early stage. 3. RESULTS In this section results for the two scenarios (Fig. 1) are analyzed. For each scenario we first analyzed the amount of CO2 that may leak into the upper aquifer as a function of the initial fault permeability and the injection rate. Basically a CO2 storage project will be considered a good one when the amount of CO2 leaking into the upper aquifer after hundred years is less than 0.1%/yr [27]. In our study we focus only on the CO2 that may leak through the fault zone in a shortterm of active injection of 5 years, then considering a maximum leakage rate of 0.1% per year at the end of the 5 years injection our threshold will 0.5% of the total injected CO2 during the entire period. At the same time we analyzed the magnitude of the main seismic event (if any) as resulting from a sudden slip. Also in this case the magnitude was analyzed as a function of the initial fault permeability and the injection rate. 3.1. Scenario 1 Resulting leakage percentage for the scenario with a small, undetected fault (1 km-long) for three different cases are shown in Fig. 2. Fig. 3. Magnitude of a single event due to a sudden slip along the 1 km-long fault zone with no offset as a function of initial fault permeability and injection rate. Contour lines represent the 1% and 5% leakage rate. (a) Case1: fault permeability changes as function of mean stress. (b) Case2: fault permeability changes as a function of the volumetric strain. (c) Case3: fault permeability changes as a function of both fault normal stress and plastic shear and tensile strain. Fig. 2a shows the results for Case1, in which the permeability changes as a function of the mean stress (Eq. 1). The leakage rate is much below the 1% for most of the simulated case. Leakage would start occurring when the initial fault permeability increase and with very high, unrealistic injection rate. For example the fault will be vertically impermeable after 5 years of injection when the initial permeability is low even using a rate of 100 kg/s. However, when the initial permeability of the fault is as high as 10-14 m2 then about 30% of the injected CO2 would reach the upper aquifer within 5 years, and such a project would be a total failure. It is worth noting that even for a very high permeability, the leakage rate would be less than 5% if the injection rate were kept below 10 kg/s, which is still a good injection rate for a CO2 storage project. Fig. 2b shows the resulting leakage percentage for a case in which the permeability changes as a function of the total volumetric strain (Eq. 2). Results for Case2 show a trend similar to Case1 for high initial fault permeability, with a maximum percentage of about 25% when the injection rate is 100 kg/s and basically no leakage when the injection rate is lower than 10 kg/s. However, the trend is totally different for permeability lower than 10-15 m2. In fact, while Eq. 2 may produce up to 4-order magnitude permeability changes within the fault core (where the strain may reach very high values in the worst case), the permeability changes are small in the damage zone, which features small strain. Then, the damage zone permeability has a value very close to the initial one. In this environment the pressure would highly increase for high injection rate, as well as the leakage rate. Fig. 2b shows the region where the injection pressure reaches a value greater than 35 MPa: a value much greater than the minimum principal stress at the same depth, then much more suitable for a shale-gas hydraulic fracturing project, but totally unrealistic for a CO2 storage project. For these values of pressure, the leakage percentage may reach values up to 40%. Finally in Case3 we simulated the permeability and porosity as a function of the normal stress and of the plastic strain (Eq. 3 and 4). Results are shown in Fig. 2c: the percentage of leakage in this case is very similar to Case1, but with slightly higher values for very high injection rates (greater than 70 kg/s). In all simulated cases then, leakage is very likely to happen with a percentage greater than 1% only for very high injection rates greater than 30 kg/s, more suitable for other projects rather than for carbon sequestration. In terms of seismic events, it is very likely to have a sudden slip within this scenario only for high injection rates (Fig. 3). In detail, Case1 features events of at least magnitude 2 only for injection rates greater than 30 kg/s and initial permeability lower than 10-15 m2. As the permeability increases, the fault reactivation requires a higher injection rate. For low injection rates the overpressure never reaches a limit value to induce an earthquake, and when the initial permeability is high then the overpressure distributes much faster avoiding the accumulation on the fault zone (Fig. 3a). For Case2, some events occur for lower injection rates, since as explained earlier, the permeability changes mostly within the fault core and not in the damage zone, allowing a greater pressure for lower injection rates, and increasing the probability for an event to occur (Fig. 3b). Case3 is similar to Case1 also in terms of events magnitude except that more seismic events may occur for permeability up to 10-15 m2 (Fig. 3c). It is of note that magnitude of the seismic events and percentage of leakage are not always correlated, and that when an event occur not always a high short-term leakage is associated with that kind of scenario. For example, in Case1 a seismic event of magnitude 3 Fig. 4. Percentage of CO2 leaking into the upper aquifer as a function of permeability and injection rate for a 2 km-long fault with 100 m offset. Contour lines represent the 1% and 5% leakage rate. (a) Case1: fault permeability changes as function of mean stress. (b) Case2: fault permeability changes as a function of the volumetric strain. (c) Case3: fault permeability changes as a function of both fault normal stress and plastic shear and tensile strain. Fig. 5. Magnitude of a single event due to a sudden slip along a 2 km-long fault zone with 100 m initial offset as a function of initial fault permeability and injection rate. Contour lines represent the 1% and 5% leakage rate. (a) Case1: fault permeability changes as function of mean stress. (b) Case2: fault permeability changes as a function of the volumetric strain. (c) Case3: fault permeability changes as a function of both fault normal stress and plastic shear and tensile strain. occurs for a permeability of 3·10-16 m2, although not leakage is observed after five years of injection at more than 50 kg/s. The poor correlation means that a single event is not enough to change the permeability substantially along the entire fault length, and then, even if some changes in permeability occur, this not means that the fluid will move along the entire fault breaking through the caprock and then degrading the upper aquifer. However, after the first slip the stresses on the fault dissipate, and our slip-weakening model does not permit the stress to accumulate after the first drop. Basically, we are not considering at this time the effects of multiple felt seismic events. 3.2. Scenario 2 This scenario has a 2 km-long fault zone with an initial offset of 100 m, and then the multilayer system results spatially shifted when crossing the fault zone. For this reason, the central aquifer where the CO2 is injected results somehow bounded on the right side by the shifted caprock, and then with a low injection rate it is already possible to increase the pressure much more over the minimum principle stress. Hence we considered a smaller range of injection rates from 2 to 12 kg/s (which is still in the same order of magnitude of the In Salah CO2 storage project [25]). In this range of injection rates it is very unlikely that a notable leakage occur, and only for a very high fault initial permeability some few percent (less than 8%) of the CO2 leaks into the upper aquifer (Fig. 4). The explanation is that for low permeability the CO2 actually starts moving upward along the injection zone, but when it reaches the shifted part of the aquifer it will rather move into the central aquifer rather than keep moving upward for buoyancy, because the permeability gradient between fault zone and reservoir. Then some CO2 will keep moving upward only when the fault zone permeability is comparable to the central aquifer permeability. A little exception occurred for Case2 (Fig. 4b). As explained earlier for the scenario #1, when the permeability changes as a function of the volumetric strain, it will change mostly within the fault zone rather than within the damage zone. Then when considering Case2 for the scenario #2, and when the initial permeability of the fault zone is low (10-16 m2), the fault core will have a very high permeability changes compared to the surrounding damage zone, and the CO2 will keep moving upward along the fault core, like in a channel, and then will leak into the upper aquifer. Much more interesting in this scenario #2 is the relation between magnitude of induced seismic events and leakage percentage (Fig. 5). Since the fault zone is larger compare to the previous scenario, events will have a bigger magnitude. It is of note to say that notable earthquake might be produced even though the leakage rate is very low or null. For example, for all the three simulated cases if the initial fault permeability is relatively low (less than 10-15 m2), then events of magnitude in the range 2-3.5 are really likely to happen even without leakage (or with low in Case2) into the upper aquifer. The results show basically no correlation between earthquake magnitude and leakage percentage. Rather, for most of the simulated cases in this scenario a fault reactivation does not imply changes in permeability that compromise the sealing potential of the caprock through the fault zone. This means that also in the case of a bigger fault zone, the presence of seismic activity does not mean an alteration in sealing properties of the caprock. Again the no correlation means that the permeability changes do not affect the entire length of the fault, hence the fault itself does not behave as a preferential pathway for the fluid to leak into the upper aquifer. In terms of magnitude itself, the three cases show slightly the same values for all the different combination, with reactivation for most of the cases, with the exception of very high initial fault permeability and very low injection rate. In some cases, results also show an event of greater magnitude for a lower injection rate. For example, in Case1 the magnitude estimation for the simulation with injection rate of 6 kg/s and permeability of 10-16 m2 is around 2, while for the same permeability the simulation with 4 kg/s will produce an event of magnitude 2.7. This effect may be explained in terms of timing: for lower injection rates (4 kg/s) the system will require a longer time to increase the pressure up to the critical value for reactivation, and then the pressure itself will distribute much more along the fault, producing at the time of reactivation a larger rupture with respect to the case with higher injection rate (6 kg/s). 4. DISCUSSION AND CONCLUSIONS In this paper, we studied the fault reactivation and the CO2 leakage through a fault zone during geological carbon sequestration activities. We addressed the shortterm capability (i.e. during 5 year of active injection) of a fault zone to act as a pathway for CO2 moving upward by overpressure and buoyancy. We carried out a high number of simulations relating different injection rates, fault permeability, and how the permeability changes as function of geomechanical parameters in two different scenarios. The first scenario represented a small, undetected fault zone (i.e. 1 km-long) with no offset [10], while the second scenario represented a larger fault zone (i.e. 2 km-long) with an initial offset of 100 m. For scenario #1, results showed that a substantial amount of CO2 may leak through the fault zone only for very high, unrealistic injection rate (more than 50 kg/s), or when the initial fault permeability was set to very high values (10-14 m2). Fault reactivation also occurred for high injection rate (more than 30 kg/s), but fault with initial low permeability (10-16 m2) were facilitated. For scenario #2, results never showed substantial leakage, even though the injection rate range was reduced to prevent an unrealistic pressurization of the aquifer. Most of the simulation performed never showed CO2 leakage into the upper aquifer during the 5 year of injection, although some few percentage of CO2 (around 8%) may leak when high injection rate (12 kg/s) and high initial fault permeability (10-14 m2) were chosen. Although no notable leakage occurred, most of the simulations in scenario #2 were characterized by fault reactivation, producing seismic event of magnitude in the range 2-4. Therefore, our results show that a seismic reactivation may occur without affecting the potential for leakage through a fault zone. This is true for a small fault, even though a high injection rate is needed for reactivation, since the permeability changes does not allow the pressure to accumulate. The seismic reactivation without leakage is more evident in the scenario #2, in which is possible to simulate an induced earthquake even using a relatively small injection rate. Moreover, inclusion of rock heterogeneities in the model will decrease the risk of leakage, and will help the CO2 to be confined within the injection reservoir. Indeed, Jeanne et al. [28] demonstrated that for the same injection rate, fault length and dip, and boundary and initial conditions, although the pressure increase will be the same for homogeneous and heterogeneous model, hence producing the same fault slip, the amount dioxide. Proceedings of the National Academy of Sciences. of CO2 leaking is definitely lower for a heterogeneous fault zone. However, although our model is the most up-to-date one, it still presents few approximations. The first (and probably the most important) is that we are only using a 2D model, what will change when considering a full 3D formulation? A second major approximation is that we can basically simulate only an induced event followed by mostly aseismic deformation: will a series of notable (felt) earthquakes compromise the integrity of the system allowing the fluids to move faster (and better) along the fault zone? The use of a slip-weakening friction law is also an approximation. In fact, after initial slip the deformation deriving from the current formulation may lead to an over-estimation of the permeability changes along the fault. Anyway, the results from the current analysis still hold, because it means that even considering a larger estimation of the permeability changes, a system can still have seismic events without substantially altering the sealing properties of the caprock. While the current analyses are very useful and expansive in scope, for future analyses a more stable rate-and-state formulation should be used. Other aspects to be fully addressed are (1) the size of the reservoir, and (2) the effects of a system with multiple caprocks and reservoirs. ACKNOWLEDGMENTS The work presented in this paper was financed by the Assistant Secretary for Fossil Energy, Office of Natural Gas and Petroleum Technology, through the National Energy Technology Laboratory, under the U.S. Department of Energy Contract No. DE-AC0205CH11231. REFERENCES 1. 2. 3. 4. 5. Kerr, R.A., 2012. Learning how to NOT make your own Earthquakes. Science 335(6075). National Research Council, 2012. Induced Seismicity Potential in Energy Technologies, National Academies Press, Washington D.C., pp. 300. Bachmann, C.E., Wiemer, S., Woessner, J., Hainzl, S., 2011. Statistical analysis of the induced Basel 2006 earthquake sequence: Introducing a probabilitybased monitoring approach for Enhanced Geothermal Systems. Geophys. J. Int. 186(2). Pacala, S., Socolow, R., 2004. Stabilization wedges: solving the climate problem for the next 50 years with current technologies. Science 133(5686): 968–972. Zoback, M.D., Gorelick, S.M., 2012. Earthquake triggering and large-scale geologic storage of carbon 6. Rutqvist, J., 2012. The Geomechanics of CO2 Storage in Deep Sedimentary Formation. Geotech. Geol. Eng. 30(3): 525–551. 7. Cappa, F., Rutqvist, J., 2011a. Impact of CO2 geological sequestration on the nucleation of earthquakes. Geophys. Res. Lett. 38(L17313). 8. Cappa, F., Rutqvist, J., 2011b. Modeling of coupled deformation and permeability evolution during fault reactivation induced by deep underground injection of CO2. Int. J. Greenh. Gas Contr. 5(2): 336–346. 9. Cappa, F., Rutqvist, J., 2012. Seismic rupture and ground accelerations induced by CO2 injection in the shallow crust. Geophys. J. Int. 190: 1,784–1,789. 10. Mazzoldi, A., Rinaldi, A.P., Borgia, A., Rutqvist, J., 2012. Induced seismicity within geological carbon sequestration projects: Maximum earthquake magnitude and leakage potential from undetected faults. Int. J. Greenh. Gas Control 10: 434–442. 11. Rutqvist, J., 2011. Status of TOUGH-FLAC simulator and recent applications related to coupled fluid flow and crustal deformations. Comput. Geosci. 37: 739– 750. 12. Davies, J.P., Davies, D.K., 2001. Stress-dependent permeability: characterization and modeling. SPE Journal 6: 224–235. 13. Rutqvist, J., Tsang, C.F., 2002. A study of caprock hydromechanical changes associated with CO2injection into a brine formation. Environ. Geol. 42: 296–305. 14. Chin, L.Y., Raghavan, R., Thomas, L.K., 2001. Fully coupled geomechanics and fluid-flow analysis of wells with stress-dependent permeability. SPE Journal 5: 32–45. 15. Hsiung, S.M., Chowdhury, A.H., Nataraja, M.S., 2005. Numerical simulation of thermal-mechanical processes observed at the Drift-Scale Heater Test at Yucca Mountain, Nevada, USA. Int. J. Rock Mech. Min. Sci. 42(5–6): 652–666. 16. Bandis, S.C., Lumsden, A.C., Barton, N.R., 1983. Fundamentals of rock joint deformation. Int. J. Rock Mech. Mining Sci. Geomech. Abs. 20(6): 249–368. 17. Leverett, M.C., 1941. Capillary behavior in porous media. Trans. Am. Inst. Mining Eng. 142: 341–358. 18. Jaeger, J.C., Cook, N., 1979. Fundamentals of Rock Mechanics. 3rd ed. New York: Chapman & Hall, pp. 28–30. 19. Kanamori, H., Brodsky, E.E., 2004. The physics of earthquakes. Rep. Prog. Phys. 67(8): 1,429. 20. Kanamori, H., Anderson, D.L., 1975. Theoretical basis of some empirical relations in seismology. Bull. Seism. Soc. Am. 65(5): 1073–1095. 21. Pruess, K., Oldenburg, C.M., Moridis, G., 2011. TOUGH2 User’s Guide, Version 2.1. Lawrence Berkeley Natl. Lab., Berkeley, CA, USA: Paper LBNL-43134 (revised). 22. ITASCA, 2009. FLAC3d v4.0, Fast Lagrangian Analysis of Continua in 3 Dimensions, User’s Guide. Minneapolis, MN, USA: Itasca Consulting Group. 23. Gudmusson, A., 2000. Active fault zones and groundwater flow. Geophys. Res. Lett. 27(18): 2,9932,996. 24. Faulkner, D. R., Mitchell, T. M., Healy, D., Heap, M., 2006. Slip on 'weak' faults by the rotation of regional stress in the fracture damage zone. Nature 444(7121): 922-925. 25. Rinaldi, A. P., Rutqvist, J., 2013. Modeling of deep fracture zone opening and transient ground surface uplift at KB-502 CO2 injection well, In Salah, Algeria. Int. J. Greenh. Gas Control 12: 155–167. 26. Rutqvist, J., Rinaldi, A. P., Cappa, F., Moridis, G. J., 2013. Modeling of fault reactivation and induced seismicity during hydraulic fracturing of shale-gas reservoirs. J. Petrol. Sci. Eng., submitted. 27. Hepple, R. P., Benson, S. of carbon dioxide as a strategy: performance implications of surface 47(4):576-585. M., 2005. Geologic storage climate change mitigation requirements and the seepage. Environ. Geol. 28. Jeanne, P., Rinaldi, A. P., Rutqvist, J., Cappa, F., Guglielmi, Y., 2013. Relation between fault zone architecture, earthquake magnitude and leakage associated with CO2 injection in a multilayered sedimentary system. In Proceedings of the 47th US Rock Mechanics, Geomechanics Symposium, San Francisco, 23 – 26 June 2013.