D4.6 Important physical processes and modeling schemes of induced seismicity

The aim of the development of a geothermal site by hydraulic and/or chemical stimulation is to irreversibly increase the injectivity/productivity of a well by a local increase of the transmissivity of the fractures and/or faults intersecting the well and an improvement of the connectivity of the well to fracture and/or fault network. The generation of microseismicity is inherent in these methods of development of wells. However, if a large number of microearthquakes is inevitable, major events (magnitude between 2 and 3) must be studied carefully. Understanding these events may lead to propose scenarios for the stimulation reducing or eliminating them while maintaining the effectiveness of the stimulation relative to the injectivity. The WP4 work package is dedicated to understanding the geomechanical causes and processes of induced seismicity in various contexts and at various scales. This report corresponds to the deliverable D4.6 for the GEISER project. It recalls the results obtained by the partners within the three sub-tasks of WP4 and highlights important physical processes and modeling schemes of induced microseismicity. Two different reservoir sandstones (Flechtingen Bausandstein, Bentheim Sandstone) have been selected as sample material for the experiments. Samples have been carefully characterized with regard to mineralogical composition, porosity and microstructure. Samples were first loaded to failure to produce a rough preexisting fracture. Specimens were then subjected to elevated pressures and loaded axially until the fracture was stressed close to frictional failure. The effect of pore fluid migration into dry sandstone samples was monitored using a set of twelve piezoceramic P-wave transducers and periodic elastic wave velocity measurements were taken. The fluid (fresh water), was injected at the sample bottom and significant acoustic emission (AE) activity was recorded in almost the entire specimen during transient pore pressure increase. Subsequently, different pore pressure levels (i.e. different effective stress states) were tested and the evolution of AE activity in the sample was recorded. The experiments show that the re-activation of a fracture through fluid pressure perturbations into saturated specimen is characterized by fewer high amplitude events than into dry specimens. Other laboratory experiments have also been conducted with specimens which had a saw-cut plane oriented at different angles to the direction of the major imposed principal stress Stick-slip test have been conducted, and the results show that slip along the saw-cut plane was accompanied by an increased number of AE events, all located in different patches along the saw-cut plane. New fractures, far from the saw-cut planes were not induced. Mohr-Coulomb envelopes reached sliding lines with different coefficients of static friction, which indicates a change in the roughness of the saw-cut planes (higher coefficients of static frictions, thus, more rough surfaces). Shear cracks were dominant during the sliding, while the number of tensile cracks was not negligible. The latter were related to local pore pressure perturbations and local dilatancy taking place during the sliding. The work of INGV dealt with modeling of time-dependent Coulomb stress changes due to fluid injection and fluid withdrawal in wells by using a thermo-fluid dynamical approach in porous media with a coupling of the two codes THOUGH2 and COMSOL. The approach consists in a two-step procedure: in the first step, injection of withdrawal of water is simulated by TOUGH2, by computing pressure and temperature changes in the volume; in a second step, the incremental stress tensor is computed by COMSOL. According to the models, stress changes can attain considerable values, able to significantly re-orient the mechanisms of most stressed fractures, with respect to the orientation due to the regional stress field alone. Also, the results highlighted that the main causes of induced seismicity during stimulation are due to the Coulomb stress changes generated by water injection, rather than by the effect of the temperature of the injected fluid. Changes of Coulomb stress due to induced seismicity during stimulation have also been analyzed by partner GEOWATT. For this purpose an efficient method has been developed to calculate coseismic stress changes from an elliptical slip distribution on a circular fracture using superposition of rectangular sources. The method has been applied on a dataset of 715 focal mechanisms derived from seismic recordings of the Soutz-sous-Forêts GPK2 2000 stimulation to calculate temporal evolution of change of Coulomb stress. It comes out that the distribution of stress changes at the hypocenters reveals no evidence for events to be triggered from previous induced seismicity. However, about 5% of the events occur in areas where the stress is increased by more than 1 MPa and are thus candidates for triggering by previous events. Partner TNO was involved in the three sub-tasks of WP4. They have shown within subtask 4.1 that for creating an EGS in compartmentalized sedimentary reservoirs, both the effects of pressure changes in the rock matrix and the impact of large scale heterogeneities should be taken into account. They also have shown within sub-task 4.2 related to the role of temperature that for simple systems dominated be few large fractures, poro-elasticity processes were most likely to control the overall thermomechanical process, by increasing the shear strength resistance of the fracture and balancing local traction and fracture opening effects, thus preventing large shear failure to develop. Within sub-task 4.3 related to the role of existing fault segments, partner TNO developed a specific coupled continuous approach with FLAC3D in order to study the role of pore pressure changes causing fault zone reactivation and inducing seismicity during the EGS operations. They performed a sensitivity analysis on key parameters such like the in-situ stress regime, the fracture mechanical properties and compared the results to the Block-Spring model as proposed by Baisch. The models were also applied on the GPK3 stimulation case that was performed in Soultz-sous-Forêts in 2003. The basic features of the observed seismicity were reproduced. The 3D physical approach and the Block-Spring model showed their specific advantages: the Block-Spring model, if calibrated, establishes a fast modeling tool for sensitivity analyses; the FLAC3D implementation allows better understanding as it is based on actual physics. The results of this study were submitted to the special edition of Geothermics. A copy of the submitted article (Article 1) is enclosed in this report. Partner ARMINES studied with a Discrete Fractured Network approach the impact of the thermal effect during a long term flow circulation test based on a forward thermohydro- mechanical modeling of the in-situ observations made at Rosemanoves site (UK). They showed that when injection pressure is maintained at constant level over time, shear can be re-activated at places where a previous rupture occurred, as soon as thermal tractions accumulated with time reach a magnitude similar to the stress drops generated during reservoir stimulation. The delay of a new event to occur is controlled by heat diffusion phenomenon in the rock blocks adjacent to the fracture. Note that calculated seismic moments due to the long term cooling of the rock are one order of magnitude lower, but that a small increase in injection pressure can precipitate new seismic activity as soon as some thermal stress has been stored. It comes out that seismic activity during the life cycle of a reservoir is not independent of pre-stimulation phases during initial reservoir development. This is of prior interest for mitigation strategies. The role of fractures heterogeneities has also been undertaken by partner ARMINES by studying different reservoir configurations: a 3D multi fracture system and a single fault segment with heterogeneous properties under a given stress regime. According to the models, the development of shear is controlled by pressure diffusion, and therefore in both cases hydraulic properties, contrasts in properties of the various structures and boundary conditions are governing the mechanical process. The critical conditions with regard to the occurrence of large induced seismic events are obtained in the situation of some large fracture with unknown hydraulic properties exists at an intermediate distance of the well, and when the ‘bulk’ hydraulic diffusivity of the massif is limited. In the case this structure is also of poor natural hydraulic diffusivity or locked by other fault segments, the risk of inducing there a significant event is higher because it can be pressurized over a large region. Therefore tracking by geophysical methods any of these structural surfaces of potential failure at distances up to some hundreds of meters from a well is crucial. Testing them separately, if cross cut by a well, even with short term hydraulic tests would produce most valuable data, that would be helpful to design the most reasonable stimulation scenario, with the constraint of a pre-identified admissible pressure in the fault. This recommendation should form part of the mitigation strategy. Drilling capabilities allows now the possibility to develop deviated wells, or wells with multiple branches in smaller diameters. Targeting such geological structures should be evaluated in case VSP had identified them for instance. Although not affordable at first glance, these investments would anyhow participate to the longterm survey of any EGS production field at industrial scale and have the potential of allowing re-entry in the future for further injection or production developments. It is therefore suggested that the deepest section of a well in a new site should be first used at survey and not be automatically designed to be included in the final system. Partner BRGM focused its study on the coupled hydro-mechanical response of a fault network intersecting a deep geothermal well to a hydraulic stimulation. The stimulation zone around Soultz-sous-Forêts deep wells was modelled using a Distinct Element Method. According to the stimulation phase and post-stimulation phase carried out here, the models reproduce the evolution of the Overpressure-Flowrate curves observed in-situ, even if the total increase of injectivity has not been reached yet. Contrary to what is traditionally done, looking at the orientation of the most favourably oriented fault zone in respect to the principal stress does not allow determining the most efficient fault network. On the basis on the study performed, the three kinds of behaviour can be met: - Direct stimulation due to shearing in fault segment favourably oriented with respect to σH. But, all the favourably oriented with respect to σH are not stimulated; - Little/no stimulation, even if favourably oriented with respect to σH; - “Indirect stimulation” of fault segment, not favourably oriented with respect to σH, due to the connectivity in 3D of the fault network. Those three kinds of behaviour are met either with the Mohr-Coulomb model or with the Continuously Yielding Joint Model. The initial calculations of the seismic moments performed on the basis of the Mohr-Coulomb model yield to overestimate the magnitude, due to the impossibility to discriminate the coseismic part of the shear from the aseismic one. The sheared seismic areas are strongly influenced by the intersections between the faults. The introduction of the CYJ Model allows discriminating the coseismic part of the shear from the aseismic one and thus has a better estimation of the magnitude. Otherwise, numerical development enables to identify the distinct seismic areas on a single fault segment from an analysis of the neighbour gridpoints that rupture at the same time. The implementation of these calculations show that: - In the directly stimulated fault segments, the maximum seismic moments reach a plateau (M < 2) due to the repartition of the shear in several distinct seismic areas, clearly lower than the maximum seismic moments calculated with the Mohr-Coulomb model (M around 3); - In the indirectly stimulated fault segments, there is barely any shear displacement at the origin of major seismic events. Although the CYJ Model brings a more realist physical behaviour, its use is trickier since it can lead to numerical instabilities in specific configurations. Finally, with the developments undertaken by BRGM, tools are available to study the occurrence of potential major induced seismic events on the same mechanical basis as the evaluation of the efficiency of the hydraulic stimulation of a deep well. Moreover, from the point of view of the knowledge of EGS well, the results obtained show the importance of taking into account the 3D reality of the fault network around the well, for the efficiency of the stimulation as well as for the estimation of the magnitude of the major seismic events. In order to lead stimulation scenarios, limiting the maximum magnitude of seismic events, an improvement is still needed in choosing the parameters of the hydro-mechanical behaviour to reach the final value of the well injectivity. The implementation of the effect of fluid injection into a pre-existing Boundary Integral Element Method (BIEM) code modeling dynamic rupture has also been done by partner BRGM in order to build, based on physical parameters, catalogues of synthetic events (location, time, magnitude) that are necessary to evaluate the seismic hazard (WP5). With this approach, it is assumed that the evolution of fluid migration and induced seismicity takes place along a single planar fault, described as a finite permeable zone of variable width. A sensitivity analysis has been performed in order to study the parameter interdependencies. It turns out that seismicity and fluid migration are strongly influenced by the injection rate and the fault rheology heterogeneity. The simulated seismicity generally tends to rapidly evolve after triggering, independently of the injection history. This self-induced seismicity takes place in the case where shear rupturing on the planar fault becomes dominant over the fluid migration process. On the contrary, if healing processes take place in the fault, so that the fluid mass becomes trapped within the fault, rupturing occurs continuously during the injection period. In comparison to the other approaches adopted by ARMINES and TNO which can also build catalogues of synthetic events, this approach is able to provide them fast, based on the variability of various physical parameters. The results of this study were submitted to Geophysical Journal International. A copy of the revision article submitted in June 2013 is enclosed in this report (Article 2). Laboratory experiments and several modeling activities with complementary approaches have been addressed throughout WP4. Different numerical tools have been developed, or adapted, in order to study the coupled mechanisms that induce seismicity during EGS operations. The efforts put in WP4 also include the development of tools (ARMINES, BRGM, TNO) to simulate catalogs of synthetic events, based on rock mass physical features and on injection rates, that are useful in order to assess the seismic hazard associated to EGS operations. A considerable progress has been therefore made in GEISER.FP