The standardized protocol consisted of four injection cycles, which each consisted of progressively increased pressure or flow rate steps (Fig. 4). In all cases, the steps were kept constant until quasi-steady-state flow conditions were reached. The first two cycles were step-pressure injections and were intended to estimate the pre-stimulation jacking pressure and injectivity of the injection interval. Here, the first cycle serves to break down the injection interval so that the near-wellbore fracture responses during subsequent cycles are largely elastic. The third cycle was a step-rate injection that constituted the actual stimulation. The majority of the fluid volume was injected during this cycle and was intended to propagate the stimulation effects away from the injection well. The last cycle was performed to estimate the post-stimulation jacking pressure and injectivity and began under pressure control but then switched to flow rate control to obtain higher flow rates in the last two injection steps. Each injection cycle was followed by a shut-in phase and a subsequent venting phase. During venting, the pressure lines were opened to the atmosphere in the AU tunnel for 20 to 40 min. However, the lines leading to the pressure monitoring intervals were opened only after the actual stimulation phase and the final injection cycle for intervals that showed a significant pressure change. Thus, the induced fluid pressure disturbances within the fractures of the rock mass were partly, but not entirely, drained at the beginning of each injection cycle. Following each experiment, all intervals were allowed to drain for a minimum of 12 h before the next experiment. The total volume of fluid injected in each experiment was limited to approximately 1000 L to ensure low seismic hazard and few disturbances to nearby experiments (Gischig et al., 2016). The backflows from the injection borehole and all pressure monitoring intervals were measured during the venting phases.

Hydraulic jacking pressure and injectivity were determined from a P–Q cross-plot of the test data, on which P presents the injection pressure and Q is the injection flow rate. Each point denotes the flow and pressure values at the end of each step when quasi-steady-state conditions are reached, typically after 10 min. The injectivity of the test interval is taken as the ratio between the flow rate and injection pressure at low injection pressures when mechanical effects are negligible. The intersection of the low-pressure linear trend with the pressure axis defines the initial formation pressure. Dahlø et al. (2003) noted that there is no consensus as to which feature in the P–Q plot provides the best estimation of the jacking pressure (i.e., the normal stress across the fractures that supports the lowest normal stress) because it is unclear at which point along the steepening P–Q curve the compliant fracture response turns into liftoff of the fracture surfaces. We take our best estimate of the jacking pressures before and after the stimulation as the intersection of the low- and high-pressure trends of the second and fourth cycles, respectively. In both these cycles it is assumed that the response of the fracture network to the step increases in pressure is purely elastic and repeatable. Therefore, the low-pressure injectivity was also derived from the second and fourth injection cycle (Figs. 6 and A3 in the Appendix). In addition, we picked the injection pressure limit during the actual stimulation (injection cycle 3) for all experiments (Fig. A3).

The induced slip dislocation within the injection intervals were estimated from acoustic televiewer (ATV) logs that were run before and after each HS experiment. The ATV probe has a travel time precision of 0.1 µs, yielding a radius precision of 0.07 mm for borehole fluids with a P-wave velocity of 1483 m s-1. The travel time precision of the ATV decreases as the measured amplitude of the received P-wave decreases. Thus, the precision strongly decreases as the borehole wall becomes very rough or the borehole radius becomes strongly elliptical (Moor and Valley, 2018). Since the S3 shear zones are located in weak meta-basic dykes, which appear rougher at the borehole walls than the S1 shear zones, the radius resolution is lower for S3 shear zones than for S1 shear zones. By comparing the pre- and post-stimulation geometry of the borehole cross section across fractures it is possible to determine whether dislocation has occurred and estimate the relative displacement vector (Cornet et al., 1998; Evans et al., 2005b). To enable comparability between the images, all logs were recentralized using an ellipse fit function. Afterwards, a difference log was produced for each test interval by subtracting the pre-stimulation log from the post-stimulation log. In this difference log, a positive caliper change at a location along the borehole wall indicates that the location has moved away from the borehole axis during stimulation (Fig. 5b). The resolved radius changes can be due to (a) stimulation-induced fracture reactivation (i.e., sinusoidal traces along the borehole wall; see Fig. 5c – HS2) or (b) damage along the borehole wall (i.e., diffuse traces; see Fig. 5c – HS4). To validate the orientations and locations of reactivated fractures, the ATV logs were compared with the brittle fractures mapped in the optical televiewer images.

To estimate the magnitude of slip dislocation across a reactivated fracture, the areas of radius increase and decrease are mapped along the fracture trace (Fig. 5b). The sum of the absolute maximum radius changes on both sides of the fracture (i.e., ΔrX1 and ΔrX2) revealed the apparent amount of slip dislocation (Δrapparant). Since the radius changes are measured normal to the borehole axis, the true slip dislocation Δrtotal is calculated by correcting the apparent dislocation Δrapparant with respect to the intersection angle between the borehole axis and the fracture plane (α).

Along the sinusoidal trace of the reactivated fracture within the difference log, the radius change varies from positive to negative and back. The location at which these radius change variations occur (Δr=0) is normal to the direction of induced permanent dislocation (Fig. 5b).

The three PRP boreholes were equipped with customized grouted packer systems to continuously monitor fluid pressure in a total of seven intervals within the test volume (Fig. 1e). The pressure monitoring intervals were named according to the borehole name and the interval number counted from the borehole bottom upwards (e.g., PRP2-1 is the lowermost interval in PRP2). The intervals were chosen to cover the different target shear zones in the volume (Table 2). The distances between the monitoring intervals and the injection locations are listed in Table A2 in the Appendix. The packer system consists of a grouted section (uppermost part), the open pressure monitoring intervals (two to three per borehole), resin sections in intervals without pressure monitoring, and inflatable packers to seal off the monitoring intervals. The packers have a length of 0.2 m and were inflated with pressures between 2 and 3 MPa. The seven intervals were connected to pressure sensors in the AU tunnel through saturated polyamide lines of 2 mm o.d. The sensors used were Keller PAA33-X transmitters that had an accuracy of 0.025 MPa. A detailed description of this packer system can be found in Doetsch et al. (2018a).

In addition to the fixed pressure monitoring intervals in the PRP holes, a double-packer system was installed in one of the two INJ boreholes that was not used for injection. The system allowed pressure to be monitored between the two packers, and between the lower packer and borehole bottom. Similarly, the pressure was also monitored between the lower packer and the borehole bottom in the INJ borehole that was used for injection. The packer systems in the INJ holes were moved for each experiment to allow injection into and monitoring of the target shear zone (Table A1).

A total of 18 piezoelectric acoustic emission (AE) receivers (type Ma-Bls-7-70m by GMuG) were installed along the tunnel walls around the test volume. Additionally, eight sensors of the same type were deployed in four dedicated boreholes (i.e., referred to as GEO boreholes; Fig. 1f). The eight borehole sensors are closest to the injection locations (3–25 m distance) for all six experiments. The sensors have a bandwidth of 1 to 100 kHz. Additionally, five calibrated one-component accelerometers (type 736T by Wilcoxon) were collocated with the five AE sensors at the tunnel wall for magnitude calibration purposes.

Seismic data were recorded continuously throughout the experiments at a sampling rate of 200 kHz using a 32-channel acquisition system with 31 active channels. Induced seismic events were located using an anisotropic velocity model based on manually picked P-wave onsets. For more details on the seismic monitoring and event localization, see Doetsch et al. (2018a), Gischig et al. (2018) and Villiger et al. (2020).

The initial injectivity and near-well transmissivity for the S1 intervals were systematically lower than those for the S3 intervals by 1–3 orders of magnitude (Table 1). Despite this difference, the post-stimulation transmissivities were remarkably similar, all lying between 5.5×10-8 and 2.3×10-7 m2 s-1. Thus, substantial transmissivity increases of up to 3 orders of magnitude were realized for the S1 shear zones, whereas the increases for the S3 shear zones were limited to less than an order of magnitude (Fig. 7). The final low-pressure injectivities, measured during the last injection cycle, ranged 0.4–1.7 L min-1 MPa-1 (Table 1).

The initial jacking pressures in the two injection intervals covering S3 shear zones are systematically larger than those for the S1 shear zones, in most cases by ∼1.5 MPa. Following the stimulations, four of the intervals showed the same or slightly reduced jacking pressure, with one showing a significant decrease (S1 stimulation - HS1) and one a significant increase (S3 stimulation - HS5) (see Table 1). The final jacking pressures for the S1 intervals varied between 3.9 and 5.5 MPa, whereas for the S3 intervals the variation was 6.8 and 7.4 MPa. As opposed to S3 intervals, the maximum recorded interval pressure during cycle 1 in S1 intervals exceeded the jacking pressure.

An upper limit of injection pressure despite increasing flow rates (referred to as pressure-limiting behavior) was observed during the main stimulation injection cycle in all experiments, with some slight systematic differences between the S1 and S3 intervals. For the S3 stimulations, a slight increase in pressure with increased flow rate was evident, as the P–Q curves become progressively steeper with increased flow rate when a pressure limit was approached on the final step (Fig. 6). In contrast, the P–Q curves for the S1 stimulations showed pressure monotonically declining with higher flow rate in some cases (i.e., HS2, HS3) and declining before recovering in others (e.g., HS1, HS8). As for the jacking pressures, the maximum injection pressures attained in the stimulation injections were consistently higher for the S3 shear zones than for S1 shear zones (Table 3).

The resolved slip was localized on a single fracture, as in HS2, or distributed over various fractures as in HS4 (Fig. 5). The maximum value of ∼1.4 mm was found for an S1 stimulation (HS3 in Fig. 7d). Dislocations slightly less than a millimeter were also identified for other stimulated S1 shear zones (HS1 and HS2, and perhaps also HS8), although the uncertainty is large. A value of ∼1 mm was obtained for an S3 stimulation (HS4), but the uncertainty in this estimate was large because of the greater borehole wall roughness at the S3 shear zones (Fig. A4). The direction of the slip vector could only be determined for two zones: for HS2 it was 261/02 (i.e., dip direction/dip), and for the two reactivated fractures in HS3 it was 264/01 and 286/04. All three fractures were reactivated in a right-lateral strike-slip dislocation in an east–west direction.

No systematic differences in the recorded pressure magnitude responses to injection into S1 and S3 shear zones were evident away from the injection well. The highest fluid pressure perturbations were detected in monitoring intervals that cross the target shear zones. Transient fluid pressure perturbations were observed on almost all PRP pressure monitoring intervals during all six stimulation experiments. In four experiments, the pressure increases rarely exceeded 1 MPa, even though peak injection pressures ranged 5–9 MPa (Figs. 8 and A5). However, fluid pressures up to 6.7 and 4.2 MPa in magnitude were observed during an S3 stimulation (HS5) and an S1 stimulation (HS8), respectively (Fig. A5). These perturbations were only seen in a few pressure monitoring intervals during these two experiments. Although one of the monitoring intervals that detected the strong pressure signals covered the target shear zone that was being injected (i.e., PRP1-2 during HS5), the remainder of the strong responses were from intervals that covered other zones, indicating that the shear zones are interconnected.

A tendency for the pressure in the PRP intervals to react more immediately to shut-in after injections into S1 intervals compared to S3 intervals can be discerned, particularly at the end of the stimulation injection cycle (Fig. 8 and Table A3). The immediate pressure response of S1 shear zones (e.g., PRP1-1) to shut-in after injection into another S1 shear zone (e.g., HS2) contrasts with the delayed reaction of pressure intervals sampling S3 structures (PRP1-3 and PRP2-2) that were target during HS4 stimulation (Fig. 8). Indeed, for the S3 stimulations, almost all monitoring intervals that included the target shear zone showed a delayed response to the shut-ins.

The fluid pressure in most monitoring intervals at the end of the experiments remained perturbed from their initial values but in all cases had returned to initial values by the start of the subsequent experiment the following morning (Fig. A5). The pressures prevailing in PRP1-1, PRP2-1 and PRP3-1 at the end of the experiments were below the initial values due to the effect of venting the intervals following the stimulation injection cycle and last injection cycle (Figs. 8, A5). The venting responses of intervals covering the S1.3 shear zone (PRP1-1 and PRP2-1) consistently differed from all other intervals in that significant backflow occurred during venting so that the interval pressure declined relatively slowly. In contrast, the pressure in all other intervals declined rapidly to the atmospheric pressure in the tunnel when the valve was opened, although it was clear in some cases that backflow into the interval was occurring as the pressure began to climb once the valve was closed (e.g., S1 covering pressure interval PRP3-1 in HS4) (Fig. 8). Thus, backflow into S1 intervals upon venting tended to be greater than that for intervals cutting S3 fracture zones.

Pressure perturbations were also detected at remote monitoring intervals. The largest distance to injection was 25 m for PRP1-1 during HS8 (Fig. A6e) and typically 15–19 m for the other stimulations (Figs. 8 and A6). These distances are Euclidean distances. Thus, the true distances of pressure diffusion along hydraulically active fluid pathways might be underestimated. No systematic difference in pressure transmission distances for S1 and S3 stimulations was evident. For both shear zone types, the pressure perturbations registered in intervals that cut the target shear zone tended to be greater than at other intervals located at a comparable distance. For HS4, however, a relatively weak response was observed at an interval (PRP3-2) that covered the shear zone into which the fluid was injected (Fig. 8).

During all experiments, the FBG sensors measured compressional and extensional strain perturbations in response to the injections. The largest strain magnitudes were observed during periods of fluid injection, and the magnitudes decreased during shut-in and venting (see Fig. 10a–b for experiments HS2 and HS4). For each strain signature, we define the permanent (i.e., irreversible) strain as the strain remaining at the end of the experiment and the reversible strain as the difference between the peak strain and the permanent strain (Fig. 9). Here, the peak strain corresponds to the largest strain excursion in the coda and may be positive (i.e., compressional) or negative (i.e., extensional). In most cases, the peak strain was observed during the stimulation injection cycle, C3 (Fig. 9a–b), when the largest amount of fluid was injected. Generally, we observed that the reversible strain amplitudes were often larger than the irreversible amplitudes (Table 4). Nonzero permanent strains were detected for each experiment on all operational gauges.

Stress information retrieved from the injection and pressure observations at the injection interval revealed a distinct behavior of S1 and S3 stimulations: during the first injection cycle of S1 stimulations, the peak injection pressure exceeded the jacking pressure and the maximum injection pressure of the stimulation cycle (i.e., limiting pressure). This indicates that S1 shear zones had a tensile strength component that had to be overcome to open the fracture. This was not observed during S3 stimulation. Such tensile strength at S1 shear zones is consistent with the observation that S1 shear zones had much lower initial transmissivity compared to S3 shear zones (Fig. 7). Prior to the stimulations, the jacking pressures of the two S3 intervals were systematically higher (∼ 7–8 MPa) than the values obtained for the S1 intervals (∼5 MPa). For all experiments, an injection pressure limit was observed during the stimulation cycle, which we interpret as liftoff of fractures (Pearson, 1981). Again, the limiting pressure was systematically higher for S3 stimulation (7–9 MPa) than for the S1 stimulation (5–6 MPa). We interpret this as higher normal stress acting across S3 shear zones than across S1 shear zones. This disagrees with the stress tensors established by Krietsch et al. (2018c) (Fig. 3), from which one would expect slightly higher normal stress across S1 than across S3. Further, the expected normal stresses would be higher (>8.5 MPa for the unperturbed stress tensor and >10.5 MPa for the perturbed stress tensor) than observed during the stimulations. We explain this inconsistency with stress heterogeneity in the stimulated rock volume (note that the perturbed stress tensor has been measured at about 40 m from the stimulated rock volume). Indeed, during the stress measurements a jacking pressure of 3 MPa was obtained at the margins of the S3.1 structure, which is lower than across the same structure at the INJ1 borehole.

The induced slip displacements imaged at the injection intervals occurred along single or multiple fractures (Figs. 5 and A4). The number of reactivated fractures was larger for the initially high transmissive S3 intervals compared to the initially low transmissive S1 intervals. In all cases with multiple reactivated fractures, one fracture trace experienced distinct large shear dislocation, similar to that observed by Evans et al. (2005b) for the Soultz-sous Forêt stimulation projects. The slip direction of each reactivated fracture was compared with the direction of the maximum shear stress vector resolved for the individual fractures using the perturbed stress tensors. The angle between the maximum shear stress vectors and the azimuth of induced slip dislocations varied between 28 and 33∘ (Table 6). Thus, the derived slip direction corresponds to right-lateral shear sense, while the predictions of the stress tensor measured nearby point to oblique right-lateral shear sense with a thrust faulting component. The angular misfit might be explained by a transient local stress transfer between adjacent fractures during fluid injection at the injection well (Kakurina et al., 2019). However, stress heterogeneity, as already inferred above, may also explain why slip directions are not well predicted by the measured stress tensor.

Given the different architectures and properties of the S1 and S3 shear zones, stress heterogeneity is expected. High fracture densities and the presence of meta-basic dikes produce elasticity contrasts around the S3 shear zones (Doetsch et al., 2020). Enhanced foliation and associated elastic anisotropy have been measured for the S1 shear zones (Krietsch et al., 2018b; Doetsch et al., 2020). Additionally, these material properties not only vary between shear zone types, but also laterally along individual shear zones (see seismic velocity distributions along S3 shear zones; Doetsch et al. 2020). Thus, we argue that stress variations related to material contrasts (changes in both magnitude and stress rotation) give rise to larger changes in normal stress along the different shear zones types than their orientation in a constant stress field.

We found that the near-wellbore transmissivity enhancement was most efficient for structures that had low initial transmissivities. After stimulation, the near-wellbore transmissivities were similar in magnitude for all experiments. This may reflect an upper limit on shear-induced irreversible transmissivity enhancement. Lee and Cho (2002) found similar effects in laboratory experiments, which suggests that achievable shear-induced transmissivity enhancement depends on the height of asperities along the fracture surface. Nonetheless, in our case, the magnitude of transmissivity enhancement depends on the architecture of the stimulated structures. Stimulation of the intensely fractured rock mass around S3 shear zones was associated with only limited transmissivity enhancement, while the less intensely fractured S1 shear zones contributed larger transmissivity enhancement. We argue that stimulation of long open-hole sections would have led to less advantageous stimulation outcomes. Prior to stimulation, the fractures in all intervals differed in near-wellbore transmissivities but had similar slip tendencies (Fig. 3). Thus, a combination of low-transmissivity structures (i.e., S1 shear zones) in the same packer interval with initially high-transmissivity structures (i.e., S3 shear zones) would not have led to a transmissivity enhancement of the low-transmissivity structures because the highly transmissive structures would have taken most of the injected fluid. This highlights the advantage of short injection intervals over long open-hole injections.

The monitored pressure signals indicated that the pressure diffused predominantly along the target shear zones, similar to observations of stimulations at other EGS sites (e.g., Evans et al., 2005a). Because individual fractures associated with the target shear zone often intersected the same pressure monitoring intervals as the target shear zones, it was impossible to resolve which portion of the pressure signal propagated along which individual fracture. We observed rapid increases in high fluid pressures (i.e., on the order magnitude of the injection pressure) only during one S1 and one S3 stimulation (HS8 and HS5, respectively; Fig. A5). During most experiments and the majority of monitoring intervals, the pressure signals are far below the injection pressure and rarely exceed 1 MPa (Fig. A5). These findings are somewhat unexpected: according to Murphy et al. (2004), among others, fracture dilation during fluid injection leads to nonlinear pressure diffusion and promotes higher fluid pressures and pressure increases further away from the borehole than linear pressure diffusion would produce. Most pressure signals in our case resemble a linear diffusive pressure field. Nonetheless, FBG strain measurements and P–Q curves (Fig. 7) confirm that fracture dilation occurred during stimulation. Thus, our observations suggest that the flow field in the fault planes is heterogeneous and high-pressure signals away from the injection point may be limited to flow channels. As shown by Krietsch et al. (2020) flow along channels may change during ongoing injections. This observation challenges common conceptual models of stimulation treatments based on oversimplified fault geometries (i.e., single penny-shaped fracture) and pressure diffusion models (radially or spherically symmetric diffusion models) (e.g., Cappa et al., 2019).

Based on the observed pressure response upon shut-in, we divided our pressure signals into two different types. (1) The pressure monitoring intervals that cover the target shear zone often observed a delayed response to shut-in, indicating a diffusion-controlled pressure signal. (2) In contrast to this, pressure intervals outside the target shear zone often responded immediately to shut-in (Fig. A5 and Table A3). In some cases, this behavior was detected further away from the injection location than the diffusion-controlled signals. These signals are most likely associated with a poro-elastic far-field response (Segall, 1989)

Similarly, the observed deformation signals can loosely be divided into a near- and a far-field response. The deformations in the near field to the injection reflect stress field changes predominantly arising from effective normal stress reductions across fractures, which can produce both normal opening and also relaxation of shear stress through slip (Stein, 1999). The observed magnitude and sign (tension or compression) of these hydromechanical deformations strongly depend on the position and orientation of the strain sensor with respect to the stimulated zone. This interpretation is consistent with McClure and Horne (2014) and Rutledge et al. (2004), who note that deformations arising from mode I, II and III dislocations can occur simultaneously. It is noteworthy that the observed peak strain often far exceeds the permanent strain remaining after stimulation. This implies that the reversible component of fracture dislocation (a combination of normal and shear compliance) may be larger than the irreversible component (a combination of slip and shear dilatancy). In the near field, we also observed the formation of new fractures that propagated away from the stimulated shear zone (Fig. 12). These fractures are interpreted as tensile fractures that formed due to stress concentrations induced by shear dislocation along irregularities (asperities) of the main shear zone (McClure and Horne, 2014) or by gradients of the slip magnitude.

In the far field, i.e., outside this complex strain field, the vast majority of strain measurements show compression. We interpret these compressive signals in the far field to be produced by volumetric compression as a consequence of the volumetric expansion in the near field to the injection (Segall and Fitzgerald, 1998). The tilt signals also belong to this category of far-field responses, as they do not directly measure fluid-pressure-induced effective normal stress changes and the corresponding elastic and inelastic consequences (e.g., fracture opening and closure, slip, etc.). Similarly,the rapid pressure increases some distance away from injection may also be related to far-field volumetric compression (Segall, 1989).

Based on hydraulic and mechanical observations, we suggest two distinct zones around the injection point: (1) a complex near-field zone dominated by pressure diffusion, fracture opening, closure, shear slip and the formation of new fractures; and (2) a far-field zone dominated by stress transfer and the associated poro-elastic response (Fig. 14). Thus, we subdivide the overall effected rock volume into a primary stimulation zone, which is close to the injection point, and a secondary effected zone that captures the far-field responses (Fig. 13).

The pressure monitoring observations indicate that the radial extent of the diffusion-controlled pressure changes extended up to 15 m from the injection point (Figs. 14 and A6). Beyond this distance, between 15 and ∼22 m, the poro-elastic response was dominant. Due to the sparse monitoring, we cannot exclude the possibility that poro-elastic responses reach much further into the rock mass. The transition between the two different responses was taken as 15 m from the injection point along the shear zone. This also corresponds to the transition between the “complex” strain field, which appeared to be directly affected by active fracture slip and normal opening, and the compressional strain field, which decayed in magnitude with distance and appeared to be a far-field effect. As for the pressure, we cannot determine the outer limit of the compressional strain field because strains larger than the detection limit were observed on the most remote FBG sensors during all stimulations.

The extent of the primary stimulation zone depends on the target shear zone properties, such as initial transmissivity and the number of reactivated fractures.

Doetsch et al. (2018b) and Schopper et al. (2020) used active seismic methods to analyze P-wave velocity changes that were observed during the stimulation experiments. They found that 4D seismic tomograms allowed for the tracking of fluid pressure and strain evolution. Close to the injection location, a zone with decreasing P-wave velocities was detected that was surrounded by a zone of increased P-wave velocities. These distinct zones of P-wave velocity changes correspond to the primary stimulated and secondary effected zone. We consider the isoline marking a 0.1 % P-wave velocity decrease to denote the boundary between the two stimulated zones (Fig. 15). The extent of the boundary was measured parallel and normal to the target shear zone and was found to be elongated along the target shear zone, which is consistent with the monitored pressure perturbations. In general, the extent of the zone with decreased P-wave velocities was larger during S1 stimulations than during S3 stimulations, which is in agreement with strain field observations. Based on the active seismic observations, the primary stimulated zone can be characterized as being ellipsoidal, as inferred from the strain data.

Villiger et al. (2020) analyzed the seismicity induced during the stimulation. For details on the network sensitivity and its impact on the estimate of the seismically active zones, we refer to their article published in this journal. The localization accuracy of the seismic events is better than 1.5 m. The radial extent of the clouds was found to be similar in both directions (i.e., parallel and normal) for S1 and S3 stimulations. However, more seismic events were detected along the target shear zone than normal to it. The seismic cloud has a smaller extent than the primary stimulation zone estimates from HM monitoring (Fig. 15). Thus, it seems to underestimate the total volume that has been affected by the stimulation. This agrees with the suggestion of various authors such as Duboeuf et al. (2017) and Guglielmi et al. (2015), who showed that a large portion of the stimulation-induced dislocation is aseismic. However, it disagrees with Cappa et al. (2019), who argued that seismicity is induced ahead of the hydraulically pressurized zone.