Sequestration of large amounts of CO2 within deep underground reservoirs has been proposed as a potential approach for reducing atmospheric emissions of greenhouse gases. A CO2 sequestration project should address the associated environmental and safety issues and, in this respect, the importance of geomechanics has recently been widely recognized. Geomechanics is even more important when fluid injection is planned in faulted reservoirs. How much CO2 can be safely injected into multi-compartment reservoirs? Are geomechanical constraints more restrictive than flow-dynamic constraints? These and other questions are addressed in the present study using a three-dimensional finite element-interface element geomechanical model. We simulate the possible mechanical failure in both the injected formation and the caprock, the fault/thrust reactivation, and the ground surface displacement in a faulted reservoir of the offshore northern Italy, where seismic surveys provided an accurate characterization of the faulted geological structure. Based on reliable petrophysical/geomechanical properties from well logs and pore overpressure as predicted by a fluid-dynamic model, the results show that the injection of 1 × 106 ton/a of CO2 may be performed over a few years only. Thereafter, part of the injected formation fails by shear stress. A number of parametric scenarios are investigated to address the major uncertainties on the geomechanical response to CO2 injection. The modeling outcome suggests that shear failure and faults/thrusts reactivation can occur much before attaining the hydraulic fracturing pressure, hence representing two major constraints for a safe and permanent containment.
 It is widely recognized that fossil fuel power plants are bound to play an important role in energy supply for a large number of countries in the medium to long term. The implementation of suitable carbon capture and storage (CCS) technologies is a key requirement for reducing the greenhouse gas (GHG) emission into the atmosphere and obtaining a sustainable power generation from fossil fuels in a carbon-constrained world. At present, carbon dioxide sequestration in saline aquifers is indicated as one of the most promising techniques [Pacala and Socolow, 2004; Intergovernmental Panel on Climate Change, 2005]. However, geological storage of CO2 requires a multidisciplinary effort involving a number of complex hydrogeological, geomechanical, geothermal, and geochemical processes that act on widely different temporal and spatial scales [Celia and Nordbotten, 2011].
 Although carbon dioxide has long been injected into declining oil fields to enhance oil recovery (EOR), with the technology to pump CO2 underground mature and commercially available, there is relatively little field experience with geological storage of CO2. A major difference between EOR and CCS is that the objective with EOR is to make profit by minimizing the amount of CO2 injected and using it again after it has been produced from the reservoir dissolved into the oil, while the goal of sequestration is to maximize the storage of CO2 and keep it permanently in place. The issues involved in sequestering CO2 underground have been addressed in the scientific literature over the last decade mainly from a fluid-dynamical and geochemical perspective [e.g., Pruess and Garcia, 2002; Nordbotten et al., 2005; Juanes et al., 2006; Mouche et al., 2010], but relatively fewer works have been devoted to the related geomechanical processes [e.g., Rutqvist and Tsang, 2002; Hawkes et al., 2005; Rutqvist et al., 2008; Cappa and Rutqvist, 2011a], and especially so where real geologic settings have been investigated [e.g., Ferronato et al., 2010; Rutqvist et al., 2010; Vasco et al., 2010]. As recently reviewed by Rutqvist , there are a number of geomechanical issues at both the injected formation and the ground surface level that are worth considering in a project of CO2 disposal in active or depleted gas/oil fields and saline aquifers. These include the following:
 the effects of hydraulic fracturing in the injection zone with the aim at increasing CO2 injectivity, along with the stress state induced in the caprock and the risk for the generation of new fractures that may jeopardize the sealing capability of the unit;
 the reactivation of existing faults crossing or bounding the injected formation with the generation of potential paths for the CO2 leakage and the related risk of induced seismicity;
 the prediction of the expected land vertical uplift and horizontal displacement with the corresponding impact on the stability of the existing ground structures and infrastructures, and the injection plant.
 The present work deals with a geomechanical modeling study of CO2 geological sequestration in a multi-compartment aquifer located in northern Italy [Donda et al., 2011]. The reservoir is tested as a possible site of CO2 disposal for one of the six CO2 CCS demonstration projects that have been selected by the European Energy Programme for Recovery (EEPR). The project has been performed in collaboration with the Italian National Institute of Oceanography and Experimental Geophysics (OGS) and IFP Energies Nouvelles (IFPEN), which reconstructed the geology and the lithostratigraphy of the studied area and performed the fluid-dynamic simulations, respectively. Additional details can be found in Castelletto et al. . The geomechanical issues listed above have been investigated based on the time-dependent evolution and local distribution of the fluid pore pressure along with the initial three-dimensional (3D) stress regime. Despite a pluri-decadal effort, modeling the generation and/or reactivation of earth fractures, together with the prediction of its opening and/or slippage, still remains a challenge. Usually, the continuity constraints limit the geomechanical modeling to reasonably indicate how the stress field changes as a result of injection and in which areas the potential for fault reactivation will increase with the injection [e.g., Chiaramonte et al., 2008; Vidal-Gilbert et al., 2010; Rutqvist et al., 2011]. Here we use an advanced geomechanical simulator [Ferronato et al., 2008] where a class of zero-thickness linear interface elements (IEs) compatible with linear and bilinear finite elements (FEs) are incorporated. While standard FEs are used to represent the continuum medium, IEs prove especially suited to address the relative displacements of contact surfaces, being specifically designed for the simulation of fault slippage and opening at the regional scale [Ferronato and Gambolati, 2010; Janna and Ferronato, 2010].
2 The Italian CCS-EEPR Project and Geological Characterization of the Injection Formation
 The Italian EEPR (I-EEPR) project is a large scale project aimed at implementing the CCS technology at one of the three 660 × 106 W units of an existing power plant (PP) located in northern Italy (Figure 1a). The objective is to demonstrate the industrial application of CO2 CCS at a full commercial scale, with the indication of the actual costs and the retrofit option for high-efficiency coal fired units to be built or converted in the next 10–15 years. I-EEPR is planned to use a post-combustion capturing technology based on absorption through liquid solvents, such as ammonia, to wash the exhaust gases after coal is burned and remove the carbon dioxide. The CO2 emissions capture rate from the flue stream of PP is expected to be over 90%, with the separation of about 1 × 106 ton/a of CO2 that will be transported offshore and injected into a deep saline aquifer.
 The reservoir selected for CO2 sequestration, thereafter named Reservoir, is located in the northern Adriatic sedimentary basin, at a depth ranging between 1100 and 2500 m below mean sea level (MSL) within the Layer-4 formation (Figure 2). The selection of the Reservoir structure satisfies three basic requirements: (1) it is less than 100 km from the PP; (2) it is located offshore to prevent possible public opposition; and (3) it is far enough from the natural gas fields presently developed in the area.
 The geological structure of the Reservoir formation has been characterized in detail by an available seismic survey reprocessed by OGS [Castelletto et al., 2013]. The OGS analysis has allowed for the definition of several horizons, at both the regional and the local scale, and has revealed the presence of a very complex network of faults and thrusts that partition the injectable porous volume into several compartments, possibly disconnected from the hydraulic point of view (Figure 3).
3 Geomechanical Model
3.1 Numerical Model
 The geomechanical response of the Reservoir formation to CO2 injection is theoretically described by the 3D fully coupled poroelasticity model originally developed by Biot . Gambolati and Teatini  and Settari and Walters  have shown that in the reservoirs and aquifers of the northern Adriatic basin, coupling between the flow and the stress fields is weak, with a one-way coupling approach fully warranted on any timescale of practical interest. One-way coupling basically consists of solving first the fluid 3D flow equations, assuming a simplified mechanical behavior of the reservoir/aquifer (in essence uniaxial vertical consolidation), and then the full 3D geomechanical problem using the updated pore pressure field, with no feedback from geomechanics to fluid dynamics. Thus, in the present analysis, the pore overpressure as provided by IFPEN is used as input into the structural equilibrium equations that are solved by the geomechanical FE-IE simulator GEPS3D (Geomechanical Elasto-Plastic 3D Simulator) developed at the University of Padova based on the infinite pore pressure gradient formulation [Gambolati and Ferronato, 2001].
 A tetrahedral FE mesh was initially generated based on the geology of the horizons and thrusts at the regional scale. As shown in Figure 4, a model domain with an area extent of 50 × 50 km2 centered on the Reservoir structure is constructed, bounded on top by the traction free ground surface/sea bottom provided by a digital elevation model (DEM) and at the bottom by a 10 km deep rigid basement. The limits of the geomechanical model are prescribed far enough from the formation of interest (Figure 1b, red box) so that the boundary conditions of zero displacement do not interfere with the deformation and stress fields induced by fluid injection. An ad hoc code has been developed to transfer the pore pressure change computed by the IFPEN flow-dynamic simulator on a structured finite-difference (FD) grid covering an area of 5 × 5 km2 (Figure 1b, white box) into the source of strength that is the input to the unstructured tetrahedral FE used by GEPS3D. The implemented procedure assigns to each FE the pressure of the FD cell where the centroid of the former is located. It has been verified that this procedure allows for a satisfactory reproduction of the pore pressure field in the geomechanical model, with the pressurized volumes, namely the integral of the pore pressure change computed over the Reservoir structure, in the FD and FE grids differing by less than 5%. The average dimensions of an FD cell are 200 × 200 × 30 m3, while a maximum volume constraint is prescribed to the FEs overlapping the FD grid to obtain an average tetrahedral edge length and an inscribed sphere radius of about 100 and 20 m, respectively.
 A very peculiar feature of the Reservoir is the complex set of faults and thrusts that separate the porous volume planned for the CO2 sequestration into different blocks disconnected from the hydraulic point of view, and must be correctly implemented into the model for a reliable fluid-dynamical and geomechanical prediction. The incorporation of the above discontinuities into the 3D unstructured discretization of the geologic horizons is performed using the interface element approach developed by Ferronato et al.  for addressing the fault opening and slippage. The final grid consists of 463,783 nodes and 2,868,292 elements, and 14 faults/thrusts discretized with 50,789 IEs (Figure 5).
 The TetGen code [Si, 2008] was used to generate the boundary constrained conforming (Delaunay) tetrahedralization, which was post-processed to incorporate the IEs into the FE grid.
3.2 Geomechanical Setting
 Available information from borehole breakouts and leak-off tests shows that normal to strike-slip stresses characterize the present tectonic regime for the upper 3–4 km thick deposits in the study area [Mariucci and Müller, 2003]. The total vertical stress is computed from density logs, with the vertical direction being a principal one. Due to the lack of field data, the total horizontal stresses are assumed to be isotropic. We compute the confinement factor , with ν the Poisson ratio, as suggested by Eaton  and based on solving a problem known in elasticity as the bilateral constraint [Zoback, 2007]. Frac pac tests of in situ hydraulic fracture performed by Eni-E&P, the Italian oil company, provide over the range 0.45–0.55 [Teatini and Baù, 2000], thus 0.24<ν<0.35, with an average ν=0.3 in the simulations that follow. The Biot coefficient b is set to 1.0 as derived from available data [Gambolati and Teatini, 2000].
 A hypoplastic constitutive law based on radioactive marker measurements carried out in the northern Adriatic sedimentary basin is used for the geomechanical characterization of the porous formations. Developed more than 25 years ago [de Loos, 1973], the radioactive marker technique (RMT) is being currently used worldwide and appears to be the most effective methodology for a realistic estimate of the actual vertical rock compressibility cM in producing gas/oil reservoirs. RMT has been extensively used to define the geomechanical properties of the northern Adriatic basin [Baù et al., 2002]. Due to the stress history and mineralogical properties of the basin, i.e., an almost normal consolidation and negligible sediment cementation, the cM values as calculated from compacting marker spacings are assumed to be representative of loading conditions (first loading cycle). Denoting conventionally compressive stresses as positive, the constitutive equation originally advanced by Baù et al.  reads
with cM in [bar−1] and the vertical effective stress σz in [bar].
 Rock expansion during fluid injection is instead controlled by unloading/reloading conditions (II loading cycle). From in situ marker data and odometer tests, Baù et al.  and Ferronato et al.  estimate an average value of the ratio s between loading and unloading/reloading cM over the range 1.8–3.5 for 100 <σz< 600 bars (i.e., for a depth range between 1000 and 6000 m in normally pressurized conditions), with s decreasing as σz rises. This parameter takes on a key role for the geomechanical formation behavior during CO2 sequestration wherever the effective stress resulting from gas injection is less than the preconsolidation stress. The representative value of s is taken to be equal to 3.5 at the depth of interest [Ferronato et al., 2003].
 The results of well logs were used to derive the depth-dependent behavior of the total vertical stress
with in [bar] and z in [m], and the fracturing gradient Gfr
with Gfr in [bar/10 m] and z in [m].
 Young's modulus E is computed as follows:
An initial E value is given to each FE according to its depth and undisturbed pore pressure p0, assumed to be hydrostatic (Table 1). During injection, the actual cM, and consequently E according to equation (4), is updated in each element on the basis of the actual vertical effective stress as related to the pore pressure variation provided by the IFPEN fluid-dynamic simulations and the loading cycle, i.e., the value of s in the injected sequences where the effective stress decreases.
Table 1. Initial Rock Geomechanical Properties Used in the Simulations Within Two Tetrahedral Elements Representative of the Reservoir Formation and the Caprock Close to an Injection Well
Young's modulus, E (bar)
2.54 × 104
1.81 × 104
Poisson's ratio, ν (-)
Total vertical stress, (bar)
Total stress ratio, (-)
Biot's coefficient, b (-)
3.3 Formation and Caprock Failure
 Gas injection into a geological reservoir generally increases the risk of shear and tensile failures. The rock can fail with the generation of local fractures, a sharp increase of the hydraulic conductivity and a significant reduction of the stress bearing capacity, thus potentially impacting on both the CO2 plume evolution and the porous medium deformation. The failure mechanism is better described with the aid of the Mohr-Coulomb representation of the effective stress state in the (σ,τ)-plane (Figure 6). During the reservoir depletion, the pore pressure p decreases (p<p0) while the effective stress increases, with Mohr's circle moving farther from the shear- (τ-) axis. By contrast, during gas injection, the pore pressure raises possibly exceeding the original pressure p0. In this case the effective stress decreases with Mohr's circle approaching the τ-axis. It is worth pointing out that during both production and injection, the maximum (σ1) and minimum (σ3) principal effective stresses may follow different paths, possibly yielding an increase of the diameter of Mohr circle that approaches the failure line bounding the envelope of the allowable stress states (Figure 6).
 Two failure mechanisms can be envisaged with the aid of a Mohr representation of the stress state:
 if the stress state is such that Mohr's circle touches the failure line, a shear failure may occur. The potential for failure is inversely proportional to the safety factor where |τm|=(σ1−σ3)/2 and are the current largest and maximum allowable shear stress, respectively, with c the internal cohesion and φthe angle of internal friction. Whenever χapproaches zero, a shear failure is likely to occur;
 if Mohr's circle crosses the τ-axis, a tensile failure takes place. The failure condition is simply σ3≤0, with the safety factor ψ=σ3/σ3,0 being σ3,0 the undisturbed initial principal stress. Similarly to χ, whenever ψbecomes zero, a tensile failure is likely to occur.
 The failure analysis is performed in two steps: (i) a basic geomechanical analysis to calculate injection-induced changes in the stress field, and (ii) an a posteriori failure analysis using the stress field calculated in Step 1 to estimate χand ψ in each element of the grid.
3.4 Reactivation of Existing Faults/Thrusts and Induced Seismicity
 The Mohr-Coulomb failure criterion may also be used to assess fracture stability [Rutqvist et al., 2007], i.e., to investigate the possible reactivation of preexisting faults and thrusts, with the generation of a possible path for the CO2 to escape. Reactivation along preexisting fractures potentially occurs wherever the shear stress τs and the normal effective stress σn acting on the fracture plane fail to satisfy the following bound:
with cf and φf the cohesion and the friction angle of the fracture, assumed equal to the c and φ values provided above. When |τs| equates or exceeds the previous limit, the fault can slip. Moreover, in case of a tensile normal stress, the fault can open.
 The normal and shear stress acting on the faults depend on the pore pressure variation experienced by the reservoir, the fault dip angle, the initial stress regime and the hydraulic conductivity of the faulted zone [Cappa and Rutqvist, 2011b]. Based on available information [Dainelli and Vignolo, 1993], Reservoirfaults/thrusts are assumed impervious, namely constituting an impermeable barrier to fluid flow.
 To estimate the magnitude of potential seismic events induced by fault reactivation, we follow the methodology advanced by Mazzoldi et al. . Based on seismological theories, Mazzoldi et al.  suggest to correlate the magnitude of a seismic event with fault activated area and slip displacement based on equations below. The seismic moment (M0) of an earthquake for a rupture patch on a fault is defined by
with G=E/2(1+ν) the shear modulus of the host rock, δs the average slip, and A the area of the activated fault. M0 is then used to evaluate the magnitude (M) by the equation [Kanamori and Anderson, 1975]
with M0 expressed in Nm units.
3.5 Land Surface Stability
 The land surface stability has to be addressed in terms of absolute and differential displacements, the latter being the key factor controlling the integrity of the existing ground structures and infrastructures. The release of the vertical effective stress due to the injection overpressure causes the reservoir to expand. The migration of the reservoir expansion to the ground surface induces a land uplift with a mechanism that is the same as the one underlying the anthropogenic land subsidence above compacting gas/oil fields, but with a reverse sign. A strong evidence of such an occurrence in several sites worldwide is provided by satellite interferometric measurements which have become available only in very recent times [Teatini et al., 2011]. This can be regarded as a positive side effect that may contribute to the mitigation of land subsidence especially in low-lying coastal areas, provided that no new concerns arise for the stability and the safety of buildings and infrastructures above or close to the storage reservoir.
 Because the Reservoir structure is located offshore, with the sea bathymetry of 30 m, only very minor effects are to be expected because of the possible upheaval of the sea bottom. Nevertheless, the displacements induced by the pore pressure rise could have some adverse impact on the anthropogenic structures planned to transport and inject the CO2 into the subsurface.
4 Numerical Simulations
4.1 Modeling Scenarios
 A few scenarios are simulated to address the main uncertainty of the petrophysical properties, i.e., permeability (κ) and porosity (n) as derived from wellbore information. The flow-dynamic simulations have been performed by IFPEN making use of the code COORES™ (CO2 Reservoir Environmental Simulator). For the details on the derivation of the petrophysical properties and the flow-dynamic simulations, see Castelletto et al. . Different flow conditions along the southwest-northeast lateral boundaries orthogonal to the fault and thrust system are also investigated. As no direct measurements of c and φ are available, it is conservatively assumed that c=0 (no cohesion) and a friction angle φ=30°, i.e., typical values are suggested in the literature for sandstone [Fjaer et al., 2008].
 Two vertical wells are planned (Figure 7 top) and three main scenarios are studied in relation to various assumptions used in the flow-dynamic simulations:
 scenario 1: it can be thought of as the baseline scenario, i.e., the scenario with the most realistic data set in relation to the petrophysical model as derived from up-scaling the porosity and permeability obtained from two exploratory boreholes. The Reservoir formation is characterized by sand and shale sequences. In both facies, a Gaussian distribution is assumed for porosity, whose average value and a standard deviation are 0.35 and 0.07 for sand and 0.28 and 0.08 for shale. The permeability is given by empirical power-law expressions [Castelletto et al., 2013] with an average κequal to 100 mD and 0.25 mD for sand and shale, respectively. Moreover, the conservative assumption of impermeable lateral boundaries (white box edges AD and BC, Figure 1b) is used;
 scenario 2: it explores the effects of a more conservative assumption on the petrophysical properties relative to scenario 1, i.e., a smaller maximum porosity and a shorter correlation length of the horizontal permeability. In particular, a Sequential Gaussian Simulation method combined with ordinary kriging has been used to compute the spatial distribution of the shale content based on well logs. Porosity ranges between 0.05 and 0.35, while the permeability spans 5 order of magnitude, from 10−3 to 102 mD [Castelletto et al., 2013].
 scenario 3: the same as scenario 1 with pervious lateral boundaries (white box edges AD and BC, Figure 1b).
 The main goal of the modeling study is to evaluate the time length T of safe injection of 1 × 106 ton/a overall, i.e., 0.5 × 106 ton/a per well. In particular, the Tflow value (computed by the flow-dynamic simulator) that produces a pore overpressure equal to the fracturing overpressure obtained with frac pac tests performed by ENI-E&P, is compared with Tgeom as derived from the geomechanical model on the basis of the failure of the ψ and χcriterion.
4.2 Modeling Results
 Figure 8 shows the yearly behavior of the pore overpressure along well w1 as provided by the flow-dynamic simulations for the three scenarios investigated by the study. A significant overpressure, ranging from 50 to 150 bar depending on the scenario, is expected to develop, with a strong influence exerted by the boundary conditions. As it has been observed in previous studies of CO2 sequestration through vertical wells [e.g., Birkholzer et al., 2009; Vilarrasa et al., 2010], in scenario 3, the induced overpressure buildup exhibits a sharp profile at the begin of injection that eventually tends to stabilize. This is not the case for scenarios 1 and 2 because of the impervious lateral boundary assumption, which practically induces the pore pressure to increase continuously. Comparison with the fracturing limit points out that Tflow is equal to 7 and 4 years in scenarios 1 and 2, respectively, while CO2 injection can continue over a 10 year period or more without reaching the safety bound in scenario 3. Consistent with the flow model, in scenarios 1 and 2, tensile failure develops, i.e., ψapproaches zero, in a few FEs around the well intakes after 7 and 4 years, respectively (Figure 7). With permeable boundaries (scenario 3), as the overpressure remains well below the fracturing limit, we have ψ>0.3 in the whole domain over the 10 year simulation period. Note that according to its definition, ψ is equal to 1.0 prior to injection.
 Interesting outcomes are obtained when the possible shear failure is investigated. It can be shown that the safety factor χis initially equal to 0.2 throughout the Reservoir, as it depends only on the ratio σH/σz, and the Mohr coefficients c and φ. In particular, the geomechanical results show that shear failure in a large portion of the reservoir occurs prior to tensile failure, with the safe injection period from the geomechanical point of view shorter than the one predicted by the fluid-dynamic modeling. Tgeom is limited to 4 and 3 years for scenarios 1 and 2, respectively, during which the shear failure is likely to remain almost confined around the injection wells (Figure 9). Again, in scenario 3 CO2 sequestration can be extended over 10 years even though, with this optimistic case as well, shear failure is expected to occur close to the injection well at the end of the simulation interval.
 Note in Figures 7a and 9a the important effect of the faults/thrusts that bound the over-pressurized volumes with a strong partitioning of the regions where ψ and χdecrease significantly. The effect of compartmentalization is even more evident in Figure 10 showing the initial effective vertical and horizontal stresses and their variation at Tflow along with the pore pressure change in a vertical cross section through the two injection wells. In the vertical direction z, decreases in effective stress Δσz correspond to practically equal increases in pore pressure, except for some local effects close to the injection screen, i.e., changes of the total vertical stress are negligibly small. On the other hand, in the horizontal direction, the in situ total stress increases in the pressurized volume as a consequence of the lateral confinement exerted by the surrounding rock mass [Rutqvist et al., 2008], being the effective horizontal stress change much smaller than the vertical one.
 Similarly to the injected formation, the safety factor ψ and χ have also been calculated in the caprock sealing the Reservoir formation. The results in Figure 11 show that the safety factors are practically unchanged with no risk of caprock failure.
 The implementation of IEs allows to investigate the possible activation (opening and/or slippage) of the faults and thrusts partitioning the Reservoir structure. No faults open. However, at Tflow some IEs (colored in red, Figure 12b) slide in scenario 2 and only very few ones do in scenario 1. Indeed, during injection, both |τs| and σn acting on the fractures separating the compartments of the injected formation increase, and sliding occurs in those IEs where the bound provided by equation (5) is not satisfied (Figure 12a). Assuming simultaneous IEs activation, equation (7) provides a potential seismic magnitude M=0.5, being A=10×103 m2, δs=0.5×10−3 m and G=1.5×109 Nm2 for scenario 1. As to scenario 2, we obtain M=1.4, with A=100×103 m2, δs=1×10−3 and G=1.5×109 Nm2. The largest sliding attains a few millimeters in both scenarios.
 The ground surface uplift at Tgeom is provided in Figure 13. For times less than Tgeom, the geomechanical properties of the medium at the global scale can be assumed practically unaffected by the local shear failure in the injection well neighborhood. Due to the large overpressure caused by the very small vertical communication between layers, in scenario 2, an 18 cm uplift is predicted above the reservoir. The actual impact of the land uplift depends to a large extent on the vulnerability of the environment and the structures resting on the sea-bottom surface. The change of sea-bottom elevation is quite negligible on account of the average 30 m depth of the Adriatic Sea in the area above the Reservoirformation. As far as the safety of the injection infrastructures is concerned, it should be emphasized that possible damages are most likely caused by differential displacements, independently of the absolute value of the vertical ground motion. The maximum displacement gradient (ζMAX) predicted by the present simulations is 9×10−5, i.e., approximately 50 times smaller than the limiting bound acceptable for steel structures.
 Finally, a sensitivity analysis carried out on the parameters c and φ shows that only a relatively large cohesion (c>5 bar) may lead to a significant reduction of the porous volume likely to fail by shear stress. A more important role is played by φ: a reduction of 5° is responsible for a shear failure to occur in almost the whole injected formation, while a similar increase would yield a reduction of the failed elements, not enough, however, to completely guarantee the safety of the sequestration. More details are provided in Teatini et al. (3D geomechanical modeling for CO2 geological storage in faulted formations. A case study in an offshore northern Adriatic reservoir, Italy, submitted to International Journal of Greenhouse Gas Control, 2012).
 The most important results are summarized in Table 2.
Table 2. Summary of the Main Results Obtained for the Three Scenarios Addressed by the Geomechanical Simulations. “Yes” Means that the Process is Likely to Occur, “No” Unlikely
3.5 × 10−5
9.2 × 10−5
4.0 × 10−5
 The safety of large-scale geologic storage of CO2 in faulted reservoirs is an issue of major interest. Recently, Zoback and Gorelick  have risen some concern about the possibility that small to moderate-sized earthquakes may be triggered by the pore pressure increase in the vicinity of preexisting potentially active faults. According to Zoback and Gorelick, the induced seismicity often occurs in response to very small increase in pore pressure. As a major consequence, few centimeters of fault slip might create a permeable hydraulic pathway that jeopardizes the integrity of the impermeable units sealing the injected aquifers. Faults can also exert an important effect on the uplift of the ground surface, as was recently observed above the In Salah CO2 project, Algeria [Vasco and Rucci, 2010]. Here, the presence of a double-lobe uplift pattern has been detected in the ground-deformation data and interpreted as the result of the activation of a ≃1800 m deep fracture/fault zone intersecting the reservoir [Rinaldi and Rutqvist, 2013]. The overpressure responsible for the activation is quite substantial, up to 100 bar, corresponding to about 160% of initial hydrostatic pressure [Rutqvist, 2012]. Unfortunately, no microseismic monitoring network was installed at the time. No evident preexisting faults have been observed in connection with other well-known CO2 sequestration projects, including the Weyburn project, Canada [Verdon and Kendall, 2011], where some microseismicity was recorded above the field in correspondence, however, of the oil producing wells, and the Utsira formation at the Sleipner gas field in the North Sea [Arts et al., 2008].
 The impact of compartmentalization on fluid pressure/stress evolution and fault stability or reactivation has been studied by numerical modeling in a few case studies. Even less are the sites where the actual 3D setting of the injected formation and/or the fault plane is addressed due to the modeling complexity. At the Teapot Dome pilot field, Wyoming [Chiaramonte and Zoback, 2008], the anticline is intersected by a series of reverse faults clearly revealed by the seismic data. Slip potential of a major fault nearly perpendicular to the maximum horizontal stress was investigated using the Coulomb failure criteria. Chiaramonte et al.  found that ≃100 bar of excess pore pressure would be required to reactivate the fault at the 1700–1800 m depth of the injection unit even with the assumption of the most pessimistic risk scenario for the fault mechanics and the stress regime. The geomechanical response to oil production after CO2 injection into a ≃1500 m deep carbonate oil field in the Paris Basin, France was simulated by Vidal-Gilbert and Nauroy . A large normal fault with a throw ranging between 50 and 100 m crosses the formation 2 km east of the field structural axis. Using a one-way coupling hydromechanical 3D model combined with a Mohr-Coulomb analysis, Vidal-Gilbert et al.  evaluated the fault slip tendency. Their results show that a maximum 40–50 bar pore overpressure is expected due to CO2 injection, whereas a pore pressure increase of 130 bar is needed for activating the fault. The effect of the maximum horizontal stress, fault strength, reservoir stress path, and Biot coefficient on the likelihood of fault reactivation has been investigated by Vidal-Gilbert et al.  for the Naylor field, Australia. This is a small depleted 1900–2000 m deep natural gas field that was selected as a demonstration site for the CO2 geological sequestration. The Naylor reservoir is bounded by normal faults forming part of the structural closure which contains the injected CO2 plume. The pore pressure increase required for the fault reactivation ranges from 20 to 170 bar.
 Our modeling outcomes support in part the existing literature. We provide evidence that the fault/thrust system comparting the Reservoir structure strongly bounds the pore pressure and stress evolution within the injected formation. Both σz and σH decrease significantly within the injected compartments, in the range 50–80 bar and 35–55 bar, respectively, according to the various scenarios. Conversely, the principal component of the horizontal stress, i.e., the component normal to the fault planes, increases up to 20 bar in the blocks surrounding the two compartments where the injection wells are located (Figure 10). The expansion of the two overpressured volumes is not expected to migrate separately toward the ground surface as in the case of the In Salah field. At variance with the case in Algeria, here the geomechanical simulations show that a single uplift dome-shaped is predicted irrespective of the addressed scenario. The difference is likely due to the tensile activation of the faults that is negligible for the Reservoir study while a considerable tensile opening is assumed at In Salah [Rinaldi and Rutqvist, 2013]. Consistent with the studies reported above, reactivation of the preexisting faults/thrusts occurs only at a large reservoir overpressure. Only in scenario 2, and subordinately in scenario 1, the faults start activating where the pore overpressure in the proximity of the discontinuity planes rises to 60–70 bar. Concurrently, |τs| increases by 15–20 bar. No fault activation is predicted in scenario 3 where the pore overpressure at the fault location remains below 40 bar. In our view, the induced microseismicity, on the order of M ≃1, and a maximum fault slippage of a few millimeters only should be a good evidence of the safety of the sequestration. In fact the Reservoir structure is sealed by an unfaulted 300-400 m thick clay formation, and such small microseismic events are unlikely to create preferential pathways for the CO2 to migrate upward.
 Finally, the simulated scenarios emphasize the importance of collecting better hydrogeomechanical data during the operation of the field or, whenever possible, in advance, to restrict the interval of uncertainty of the geomechanical response and thus increase the reliability of predictions. Although the geometry of the geologic formations and the regional to local faults/thrusts system comparting the Reservoir structure is well known, with a good geomechanical data set derived from lab tests and log surveys, nevertheless, little information is available on initial stress regime and the hydraulic and mechanical properties of the faults/thrusts. More technological and economical efforts should be implemented to fill this gap.
 We have developed an advanced Finite Element—Interface Element model to represent the complex three-dimensional geologic setting of a multi-compartment reservoir in the offshore northern Italy, and address the most relevant geomechanical issues connected with geological CO2 sequestration in this large-scale formation. The distribution of the petrophysical and geomechanical properties, original in situ stress and pore pressure are available from ad hoc wellbore measurements performed in the basin.
 The study emphasizes that geomechanics can play a very important role in relation to a safe geological disposal of carbon dioxide, and especially so if faults and thrusts intersect the formation selected for the sequestration. Geomechanical constraints, in particular the maximum allowable shear failure within the injected formation, can significantly reduce the time for the safe injection of 1 × 106 ton/a of CO2, up to almost one half the time required for the generation of a fracturing pressure. Also, the hydraulic properties of the injected formation can significantly change because of a large amount of failed volume and must be adequately accounted for in the fluid-dynamic modeling. The actual fault/thrust slope and an unfavorable stress regime play an important role in the possibility of fault reactivation, which is strongly controlled by the in situ stress field tangential to the discontinuity plane.
 The research was supported by ENEL Ingegneria e Innovazione S.p.A. The authors are indebted to Monia Politi (ENEL S.p.A.) for her effort in coordinating the study. OGS and IFPEN are acknowledged for providing the basic input used to perform the geomechanical simulations. We thank the associate editor and two anonymous reviewers for their careful review of the manuscript that contributed to improve significantly the quality of the presentation.