Grant Agreement no. 241321-2

Geothermal Engineering Integrating Mitigation of Induced Seismicity in Reservoirs

Project Acronym: GEISER

D3.2 – Source characterisation of Large Magnitude Events (LME)

and their occurrence in space and time Due date of deliverable: 31.03.2012 Actual submission date: 25.04.2012 Start date of project: 1.1.2010 Duration: 42

Participant short name: ETHZ, EOST, GFZ, NORSAR NAMES INVOLVED: A. Zang, B. Goertz-Allmann, N. Deichmann, V. Oye, M. Bohnhoff, Ph. Jousset, G. Kwiatek, P. Zhao, F. Catalli, V. Gischig Revision: 1

Dissemination Level

PU Public PP Restricted to other programme participants (including the Commission

Services)

RE Restricted to a group specified by the consortium (including the Commission Services)

x

CO Confidential only for members of the consortium (including the Commission Services)

Table of Contents

1  Executive Summary ......................................................................................................................... 3 2  Source characterization of LME .................................................................................................... 10 

2.1  Double-couple and Moment Tensor Solutions of the Induced Seismicity by the Simulation of Enhanced Geothermal Site in Basel: Draft manuscript to be submitted to Geothermics, Zhao et al. (2012) ................................................................................................................................................ 10 

2.1.1  Introduction ..................................................................................................................... 10 2.1.2  Methodology.................................................................................................................... 11 2.1.3  Tectonic setting and seismic data .................................................................................... 12 2.1.4  Results ............................................................................................................................. 13 2.1.5  Discussion........................................................................................................................ 14 2.1.6  Acknowledgements ......................................................................................................... 16 2.1.7  References ....................................................................................................................... 16 

2.2  Identification of faults activated during the stimulation of the Basel geothermal project from cluster analysis and focal mechanisms of the larger magnitude events............................................. 26 

2.2.1  Introduction ..................................................................................................................... 26 2.2.2  Tectonic setting................................................................................................................ 27 2.2.3  Seismic networks............................................................................................................. 27 2.2.4  Cluster definition ............................................................................................................. 27 2.2.5  Relative locations ............................................................................................................ 28 2.2.6  Focal mechanisms............................................................................................................ 29 2.2.7  Results ............................................................................................................................. 30 2.2.8  Conclusions ..................................................................................................................... 32 2.2.9  Acknowledgements ......................................................................................................... 33 2.2.10  References.................................................................................................................... 34 

3  Modeling the probability of LME.................................................................................................. 43 3.1  Probability in space and time: parts from Goertz-Allmann and Wiemer (2012).................... 43 3.2  Dependence on depth and crustal strength: parts from Goertz-Allmann and Wiemer (2012) 44 3.3  Dependence on pumping parameters: Gischig et al. (2012) ................................................... 50 3.4  References............................................................................................................................... 52 

4  Coulomb stress changes at the Basel geothermal site: can the Coulomb model explain the induced seismicity in an EGS? ........................................................................................................................... 54 

4.1  Abstract................................................................................................................................... 54 4.2  Introduction............................................................................................................................. 54 4.3  Dataset and methodology........................................................................................................ 55 4.4  Coulomb stress changes and Coulomb indexes: discussion of results.................................... 56 4.5  Conclusions and outlook......................................................................................................... 58 

5  High-resolution analysis of seismicity induced at Berlín geothermal field, El Salvador: Kwiatek et al. (2012) ............................................................................................................................................... 66 

5.1  Abstract................................................................................................................................... 66 5.2  Summary................................................................................................................................. 66 

6  A survey of the induced seismic responses to fluid injection in geothermal and CO2 reservoirs in Europe: Evans et al. (2012) ................................................................................................................... 69 

6.1  Abstract................................................................................................................................... 69 7  Appendix........................................................................................................................................ 70 

7.1  Draft manuscript by Kwiatek et al. (2012) ............................................................................. 70 7.2  Manuscript by Goertz-Allmann and Wiemer (2012).............................................................. 70 7.3  Article by Evans et al. (2012) ................................................................................................. 70 

2

1 Executive Summary One major topic addressed by a number of papers is the characterization of Large Magnitude Events (LME), coupled with the investigation of strategies to describe, and ideally mitigate the seismic risk of stimulation operations. Microearthquakes are induced during high-pressure stimulation and they enhance and/or create the required permeability. At the same time, it is crucial not to induce or trigger LME, which may not only cause damage at the surface, but also lower the efficiency of the geothermal system by the creation of “hyper-permeable” pathways. If a LME occurs, it may create a master pathway for fluids that can shortcut the reservoir, preventing heat exchange to be efficient. A better physical understanding of the reservoir- and therefore seismicity generation process is needed in order to develop techniques to reduce the probability of large magnitude earthquakes.

It was first proposed by McGarr (1976) that the total moment of an induced seismicity cloud is proportional to the volume of injected fluid. However, the actual amount of seismicity (or the proportionality factor) can vary quite drastically between different locations. Using a completely different approach of pressure diffusion theory, Shapiro et al. (2007, 2010) arrive at a similar result in which the total number of induced events is proportional to the injected fluid volume. In addition, they describe the proportionality factor, called the seismogenic index, as a function of measurable seismological quantities and rock properties. Combining these considerations with the assumption that seismicity always follows a Gutenberg-Richter-type magnitude distribution leads to a probabilistic description of the maximum expectable magnitude.

The consequence is that the seismic hazard and hence the probability of LME generally increases with injected volume, even though there are significant regional differences. For this reason, we shall differentiate between long-term injection operations (e.g., wastewater or CO2 disposal) where large volumes may accumulate over time, and the rather short term stimulation operations at the beginning of an EGS project. Most of the LME reported in the literature have occurred either after long-term fluid injection (e.g., Ake et al., 2005, Frohlich et al., 2011), or as a result of reservoir impoundment (see Gupta, 2002 for an overview), which can have a similar effect as water injection on reducing criticality in the shallow crust. Of the short-term stimulation activities, EGS stimulations have generally shown a much higher propensity to produce LME, compared, e.g., to hydraulic fracturing in the oil- and gas industry. Shapiro et al., 2010 find significant higher seismogenic indices for geothermal stimulations in crystalline rocks than for comparable operations in sedimentary formations.

The above considerations and results provide the background for the research results presented in this special issue. Most of this work was conducted as part of the GEISER program, funded by the European Commission within the 7th Framework Programme. The compilation of papers range from the description of large-magnitude seismicity in past geothermal operations to numerical forward-modeling of seismicity clouds, all with the aim to obtain some general conclusions about the creation of LME in geothermal operations.

Evans et al (2012) summarize the results from over 40 European case histories, describing the seismogenic response of crystalline and sedimentary rocks to fluid injection. The data suggest that injection into sedimentary rocks tends to be less seismogenic than in crystalline rocks. Large or damaging earthquakes tend to occur on developed or active fault systems. In other words, large earthquakes are unlikely to occur where there is not a fault large enough to release sufficient energy. Therefore, the risk of producing LME is increased if faults near the well are present. The authors compare the maximum magnitude of each project with the long-term probability of exceeding a threshold peak ground acceleration (PGA) level, and

3

speculate that fluid injection in areas with lower natural seismicity may have a lower risk of producing LME. However, a causal relationship, if any, between the two properties remains unclear and requires a more detailed geomechanical examination of each project.

Table 1.1 lists the cases with the largest LME in order of decreasing event magnitude. Some of these case studies, mainly Basel (Switzerland), Soultz-sous-Forets (France), and Berlin (El Salvador), were studied in detail as part of the on-going research programme. Apart from accurate magnitudes (moment magnitudes wherever available), table 1 also lists the time and location of recorded LME at these sites. It has been frequently observed that LME occur after shut-in and at larger distances to the injection point (i.e. at the edges of the seismic cloud).

Kwiatek et al. (2012, this issue) contribute a case study on the geothermal project in Berlin, El Salvador, where a Mw=3.7 LME occurred 2 weeks after shut-in of the injection. Kwiatek et al. (2012) find that the event is located on some part of an active fault that did not rupture before. They also find a dependence of stress drop to distance from the injection point, similar to what was found by Goertz-Allmann et al. (2011) for the Basel dataset. However, their limited dataset does not allow a spatial mapping of stress drop nor a correlation to the LME.

One of the most intensively studied EGS stimulation that triggered several LME causing damage at the surface is the project in Basel, Switzerland. In Basel the largest magnitude event (ML=3.4) occurred a few hours after shut-in and three additional events with ML>3 occurred one to two month later. These events occurred despite the implementation of a traffic-light system that reduced the wellhead pressure after the first occurrence of a M2.5 event. The largest earthquake occurred at the edge of the seismicity cloud. Comparison of the results of high-precision relative locations of hypocenters pertaining to clusters of similar earthquakes and fault plane solutions with the strike of the strongly aligned seismicity cloud by Deichmann (2012, this report) suggests that the strike or dip of the identified faults deviates substantially from this overall orientation. The concept of a single fault zone with a more or less constant orientation is obviously too simplistic. This has important consequences for models of fluid migration during stimulation and thus of the process of permeability enhancement as well as for seismic hazard assessments.

Zhao et al. (2012, this issue) contribute a paper in which they analyse the 19 largest events of the sequence using a full-waveform MT inversion. Special emphasis is on the non-double-couple (non-DC) components of the moment tensor, where they find an increase of the isotropic component for earlier events near the injection point. Their result implies sizeable volume changes caused by large pore pressures at the early times close to the injection.

Spatial variations of source parameters such as stress drop and b-value of the Basel induced seismicity have been estimated by Goertz-Allmann et al. (2011) and Bachmann et al. (2012), respectively. It has been observed that stress drops are greatly reduced and b-values increased in the vicinity of the injection point where pore pressures are highest. LME in Basel are located in areas of higher stress drop and lower b-value. A comparison with estimated pore pressure variations due to fluid injections suggest that both stress drop and b-value are mainly driven by pore pressure variations. This has important implications on the distribution of seismic hazard.

Catalli et al. (2012) analyse the role of Coulomb failure stress variations (ΔCFS) of the Basel seismicity and find that 70% of event locations are consistent with positive ΔCFS. They find that three out of the four largest events in Basel are located in areas of positive ΔCFS. While pore pressure changes are certainly a first-order factor driving the seismicity at close distances

4

to the injection, their results suggest that Coulomb stress variations ΔCFS due to the already induced seismic events increases in importance at later times and at larger distances from the injection point where the pore pressure change is strongly reduced. They conclude that event interaction is a key parameter in assessing seismic hazard of induced seismicity.

Apart from observational papers, some modeling papers provide more insight into the physical mechanisms driving induced seismicity and the creation of LME. Goertz-Allmann and Wiemer (2012) developed a simple stochastic seed model to simulate induced seismicity in the Basel reservoir using linear pressure diffusion and seismicity triggering based on Coulomb friction. They link Gutenberg’s b-value to differential stress in an inverse linear relationship. By randomly assigning magnitudes based on varying frequency-magnitude distributions, they can estimate the spatio-temporal variation of the probability of an event of a certain magnitude. Goertz-Allmann and Wiemer (2012) find a substantial increase of the mean probability for the time period after shut-in and also at further distances from the injection point, compared to a constant b-value model. Their results not only explain the observations at Basel very well (Goertz-Allmann et al., 2011; Bachmann et al., 2012; Zhao et al., 2012, this issue), but are also consistent with observations at other geothermal projects (Kwiatek et al., 2012, this issue; Charlety et al., 2007; Evans et al., 2012). The presented modeling and magnitude probability estimation technique could be used in a refined traffic light system, where the future magnitude probability is estimated based on the previously recorded seismicity of an on-going operation.

Investigating the influence of some modelling parameters on the resulting seismicity suggests that the probability of LME is reduced if we either drill less deep, or drill into softer formations such as sediments instead of granite. This thought is taken further by Gischig et al. (2012) who investigate the influence of the injection parameters on the probability of occurrence of LME. They show that varying the injected volume has a strong effect on the probability of LME, independently confirming the initial McGarr-Shapiro model of induced seismicity. However, their modelling results also suggest that the injection rate has a strong influence on the magnitude distribution: according to their model, while maintaining the same volume, reducing the wellhead pressure and increasing the injection time reduces the probability of inducing a LME.

One of the consequences of the spatial b-value dependence (Bachmann et al., 2012) and its connection to differential stress (Goertz-Allmann and Wiemer, 2012) is that, due to decreasing b-values, the seismic hazard increases with time and distance from the injection. This could result in a symmetry break of frequency-magnitude distributions with effectively lower b-values (closer to the tectonic average of 1.0) at higher magnitudes. This result is in contrast to calculations by Shapiro et al., 2011 who find a non-linear frequency-magnitude distribution of induced events based on geometrical considerations. The probability of LME is reduced in this model due to the fact that the stimulation volume is finite. The latter model does not take into account the possibility of pre-existing faults that extend beyond the seismogenic volume and might be reactivated over a surface extending beyond the stimulated volume. On the other hand, assuming a power-law frequency-magnitude distribution (constant b-value) implies a power-law distribution of fault sizes in the volume surrounding the injection point. An initial hazard assessment, especially the probability estimation of LME, may require an analysis of the size and distribution of pre-existing faults in the prospect area.

In summary, we have gained the following insight into the creation of large-magnitude earthquakes at geothermal stimulations.

5

The total seismic moment, and therefore also the probability of a LME is mainly driven by the injected fluid volume.

Injecting less fluid typically leads to less seismicity, but also to smaller permeability changes in the reservoir. The latter can be achieved for example by subdividing the stimulated rock volume into several segments along a horizontal well, as is typically done in hydraulic fracturing for hydrocarbon reservoirs.

The LME probability is also reduced for shallower injections, and for injections into sediments instead of granite.

The magnitude distribution of the seismicity and hence the LME probability can be driven by the injection parameters, most notably the injection pressure. At the same injected volume, lower injection pressure leads to lower LME probability.

Many LME have occurred on reactivated pre-existing fault planes. It is therefore necessary to identify pre-existing, seismogenic faults beforehand in order to better assess the seismic risk of a stimulation operation. The spatial distribution of faults and fractures can have an impact on the assumed frequency-magnitude distribution. Different frequency-magnitude distributions have a significant impact on assessments of the probability of occurrence of LME

6

Sites analyzed in GEISER 

LME,year 

LME, time and location  

Geology, rock type, stress 

Pmax, MPa 

Reservoir depth (km), fracture mechanism 

Previous References 

The Geysers , California USA 

4.6*, 1982  

on the edges of seismic cloud 

meta‐graywacke 

7  3 km, cooling‐induced shear slippage, since 1975 

Oppenheimer 1986, Rutqvist et al. 2010 

Berlin, El Salvador 

4.4*, 3.7#, 2003 

2 weeks after shut‐in; on part of fault that did not rupture before 

young volcanic weak rock 

13  2 km, opening and closing of flowing fractures,  since 1991 

Bommer et al. 2006, Kwiatek et al. (2012) 

Cooper Basin, Australia 

3.7*, 2003 

  granite with 3.6 km sediment cover, TF SH ‐ EW 

68  4.1 to 4.4 km, slip on pre‐existing sub‐horizontal fractures, since 2003 

Asanuma et al. 2005, Baisch et al. 2006 

Alkmaar, NL  

3.5*, 2001 

  sandstones, 2.6 to 3.1 km depth 

18  2 km, reactivation Roer Valley Rift faults, gas production since 1963 

van Eck et al. 2006, Dost + Haak (2007) 

Basel , Switzer‐land 

3.4*, 2006 

few hours after shut‐in but before bleed‐off; at the edge of the seismic cloud 

granite, Sh= 0.7SV, SH – N144°E±14° 

30  4.4 to 4.8 km, pre‐existing,en‐echelon‐type shear zone, since 2006 

Häring et al. 2008, Evans et al. 2012, Deichmann and Giardini 2009. 

Soultz‐sous‐Forets, France 

2.9*, 2003 

in 2000, 2003, 2004 after shut‐in 

granite, NF + SS SH – N170°E 

16  4.5 to 5.0 km (GPK3), single large tectonic fracture zone, since 1987 

Cuenot et al. 2008, Dorbath et al. 2009 

Paralana, Australia 

2.5#, 2011 

at the end of second stimulation, at the base of seismic cloud 

Sedimentary basin, with basement below 4km, TF 

62  4 km, reverse fault events 

Hasting et al., 2011, Albaric et al. (this issue) 

Rosman‐owes, Cornwall, UK 

2.0*, 1987 

  Carnmenellis granite batholite 

16  2 km, system of natural fractures, since 1977 

Pine + Batchelor 1984, Turbitt et al. 1987 

KTB, Germany 

1.2*, 1994 

  gneiss, metagabbro SS (1‐8km),  SH – N160°E 

53  9.1 km, scientific wells, dilatant shear cracks, since 1987 

Zoback + Harjes 1997, Baisch + Harjes 2003 

7

Groß‐Schöne‐beck,  Germany 

‐1.0#, 2007 

  Rotliegend sandstone, volcanic rock NF, SH‐N18°E 

60  4.1 km, only a total of 80 seismic events detected, doublet in 2007 

Huenges et al. 2006, Kwiatek et al. 2010 

Seismic: *= local magnitude, #= moment magnitude Hydraulic: Pmax= maximum well head pressure Stress: SH= maximum-, Sh= minimum horizontal-, SV= vertical in-situ stress, SH-Azi (N°E) Faulting type: NF= normal-, TF= thrust, SS= strike-slip faulting Table 1.1: Large magnitude events (LME) in deep crustal injection experiments listed in decreasing order of local magnitude.

References Ake, J., K. Mahrer, D. O. Connell, and L. Block (2005), Deep-injection and closely monitored induced seismicity at Paradox Valley, Colorado, Bulletin of the Seismological Society of America, 95, 664–683. Bachmann, C., S. Wiemer, B. Goertz-Allmann, and J. Woessner (2012), Influence of pore pressure on the size distribution of induced earthquakes: in press Geophys. Res. Lett. Catalli, F., M.-A. Meier , and S. Wiemer (2012) Coulomb stress changes at the Basel geothermal site: can the Coulomb model explain the induced seismicity in an EGS?, in preparation for Geophys. Res. Lett. Charlety, J., N. Cuenot, L. Dorbath, C. Dorbath, H. Haessler, and M. Frogneux (2007), Large earthquakes during hydraulic stimulations at the geothermal site of Soultz-sous-Forets: Int. J. Rock Mech. Min. Sciences, 44, 1091–1105. Deichmann N, Gardini D (2009) Earthquakes induced by the stimulation of enhanced geothermal system below Basel (Switzerland). Seismological Research Letters 80(5), 784-798

Evans, K., A. Zappone, T. Kraft, N. Deichmann, and F. Moia (2012), A survey of the induced seismic response to fluid injection in geothermal and CO2 reservoirs in Europe: Geothermics, 41, 30 – 54. Frohlich, C., C. Hayward, B. Stump, and E. Potter (2011), The Dallas–Fort Worth Earthquake Sequence: October 2008 through May 2009, Bulletin of the Seismological Society of America, 101(1), 327–340, doi: 10.1785/0120100131 Gischig, V., B.P. Goertz-Allmann, C.E. Bachmann, and S. Wiemer (2012), Effect of non-linear fluid pressure diffusion on modeling induced seismicity during reservoir stimulation, EGU 2012 Conference contribution Goertz-Allmann, B.P., A. Goertz, and S. Wiemer (2011), Stress drop variations of induced

8

earthquakes at the Basel geothermal site, Geophys. Res. Lett., 38.

Goertz-Allmann, B.P., and S. Wiemer (2012), Geomechanical modeling of induced seismicity source parameters and implications for seismic hazard assessment, submitted to Geophysics Gupta, H.K. (2002), A review of recent studies of triggered earthquakes by artificial water reservoirs with special emphasis on earthquakes in Koyna, India, Earth-Sci Rev, 58(3-4), 279-310

Häring MO, Schanz U, Ladner F, Dyer BC (2008) Characterization of Basel 1 enhanced geothermal system. Geothermics 37, 469-495

Kwiatek, G., F. Bulut, M. Bohnhoff, G. Dresen, and S. Oates (2012), High-resolution analysis of seismicity induced at Berlin geothermal field, El Salvador, to be submitted to Geothermics special issue

McGarr, A. (1976), Seismic Moments and Volume Change, J. Geophys. Res., 81(8) Shapiro, S.A., C. Dinske, and J. Kummerow (2007), Probability of a given-magnitude earthquake induced by a fluid Injection, Geophys. Res. Lett., 34(L22314), doi:10.1029/2007GL031615 Shapiro, S.A., C. Dinske, C. Langenbruch, and F. Wenzel (2010), Seismogenic index and magnitude probability of earthquakes induced during reservoir fluid stimulations, The Leading Edge Zhao, P., V. Oye, D. Kühn, and S. Cesca (2012), Full Waveform Inversion of Moment Tensor Solutions of the Induced Seismicity by the Stimulation of Enhanced Geothermal Site in Basel, to be submitted to Geothermics

9

2 Source characterization of LME

2.1 Double-couple and Moment Tensor Solutions of the Induced Seismicity by the Simulation of Enhanced Geothermal Site in Basel: Draft manuscript to be submitted to Geothermics, Zhao et al. (2012)

Our study presents the results of moment tensor (MT) inversion of 19 microseismic events with ML > 2, associated with the enhanced geothermal reservoir stimulation operation at Basel, Switzerland (starting 2 December, 2006). We use the software package 'Kiwi' and adopt a three-step procedure to retrieve point solution parameters based on a waveform fit. The inversion is performed on displacement spectra and time series seismograms in the firstand second step, respectively. After the first two steps, we obtain focal solutions of these 19 events assuming a double couple (DC) source model. Our results are in agreement with focal mechanisms from a previous study, which used first-motion polarities, whereas our solutions are achieved using full waveform data from less than 10 (mostly 6) stations. In the last step, MT components of each event are solved using the best DC solution as the initial model input. The isotropic components of MT solutions of some events are not negligible, which might be caused by the volume change due to fluid injections. We also investigate the spatio-temporal patterns of isotropic components of the MT solutions to infer the co- and post-stimulation process of the Basel site. We also discuss the effect of velocity model and station selection for the inversion.

2.1.1 Introduction

To generate an efficient hydraulic subsurface heat exchanger system for creating an enhanced geothermal system (EGS), a large amount of fluid needs to be injected at depth to fracture the rocks and increase the permeability of geothermal reservoirs. The process is often associated with occurrences of numerous microearthquakes, and a few of them might be large enough to be felt by local communities and cause considerable damage. It, therefore, is critical to understand the mechanisms of hydraulically induced fractures and their related seismic effects, not only for efficiently generating subsurface fracture networks but also their potential seismic hazard at the surface. The Deep Heat Mining (DHM) Project in the city of Basel is initiated to stimulate the 'hot dry rock' for a local geothermal power plant (Häring et al., 2008). Between 2 December and 8 December, 2006, approximation of 11,579 m3 of river water was injected in a five-km-depth well (Häring et al., 2008; Deichmann and Giardini, 2009). Since the start of injection, the seismic activity in the reservoir began to increase (Figure 5 of Häring et al., 2008) and due to an unacceptably high level of seismicity after six days of injection, it was decided to stop the injection in the morning of 8 December, 2006. Two events with ML 2.7 and 3.4 occurred in the afternoon and evening, respectively, of the same day, which caused temporally suspension of the project. The well head was opened afterwards to allow water to flow back and the seismicity declined rapidly since then, but even after two years, sporadic seismicity are still detected in the stimulated rock volumes (Deichmann and Giardini, 2009).

10

Although the entire project was suspended and eventually abandoned after independent risk assessment, the large amount of data collected from this project provide a great opportunity to study different aspects of hydraulic fracturing during the operation, such as responds of rock to the high pressure and fluid rate injection, seismic hazard and ground shaking from the possible strongest induced event. All of these provide valuable lessons for better design and management of any future EGS project. Some of preliminary results can be found at Häring et al. (2008), Dyer et al. (2008), and Deichmann and Giardini, (2009). Due to large amount of induced seismicity occurred during and after the injection, passive seismic tools are useful for both real-time monitoring the seismicity level and offline analysis of earthquake locations and source mechanisms. Image of seismicity cloud might delineate fluid channels or the fracture network due to the stimulation (e.g., Phillips et al., 2002). On the other hand, physics of seismic sources can contain much insight about the stress distribution and fluid participation than seismic location itself. Deichmann and Ernst (2009) documented the focal mechanisms of 28 strongest induced seismic events during and after the 2006 stimulation operation in Basel. Their result shows that majority of them can be fit to pure shear slips and corresponding P and T axes are in large consistent with the in-situ stress fields from the well and ambient stress field estimated from natural seismicity in the same region. However, they also discovered inconsistent polarities at two nearby stations from three nearly co-located events, which might be interpreted as the existence of non-double-couple components with possible volume changes (Deichmann and Ernst, 2009). Alternative evidence for non-double-couple component could be directly revealed from moment tensor solutions of these seismic sources. In this study, we invert the source parameters of 19 strongest induced seismic events in Basel for both double-couple (DC) and moment tensor (MT) models. We try to correlate the temporal and spatial patterns of isotropic components of MT solutions with the injection parameters. We also investigate the stability of inversion results for different sets of velocity models and data. In the next two sections, we first explain the methodology in details, and followed by a brief introduction of regional networks and dataset. The results are presented in Section 4 and further discussed in Section 5.

2.1.2 Methodology

We use a software package named Kiwi (KInematic Waveform Inversion) from University of Hamburg, Germany, which aims to fast automatically invert source parameters of medium to large earthquakes based on the fit of seismic waveforms (Cesca et al., 2010). Kiwi includes three source models: both DC and full MT components for a point source and kinematic model with rupture parameters. Considering the size of analyzed seismicity (2<ML<3.4) and the applied frequency band (1-4Hz) in present study, we only utilize the DC and MT models. To reduce the computation time, a Green function (GF) database has to be created in advance from a 1D layered model for a range of earthquake depths and epicenter distances. During the inversion procedure, GFs are interpolated to generate synthetic seismograms for any given configuration of source-receiver pair. To build the GF database, we adopt the 1D velocity and density profiles from Table 2 of Ripperger et al. (2009) with slight modification. This 1D model is constructed from seismic reflection and refraction measurements (Campus and Fäh 1997) and has been used as a reference model for ground motion simulation in Basel area (Ripperger et al. 2009).

11

We apply a three-step approach to retrieve the source parameters in sequence, as described by Cesca et al. (2010). The summary of each step is listed in Table 2.2.1. In each step, to ensure a solution will converge to the global minimum of a complicated misfit function, many starting points in the model space have been implemented as described below. In the first step, the inversion is performed in the frequency domain to fit the amplitude spectrum of the data. A DC source type is assumed in this step and inverted five source parameters include strike, dip, rake, depth, and seismic moment. The initial model solutions are composed of a combination of different values of strikes, dips, rakes and depths from a large range. As a result, there are 452 initial model ‘points’ and for each of them we inverted a final solution using Levenberg-Marquardt (LM) algorithm. The solution of Step 1 is chosen among the best of all 452 final solutions in terms of spectrum fit. Since polarity of waveform has not been used in this step (i.e., only amplitude spectrum fit), it is unable to determine the compressional and dilatational quadrants of the DC source. In addition, there is an ambiguity to distinguish the primary and auxiliary fault planes for a pure shear source (Shearer 1999). As a result, we obtain four focal solutions with equally well fit to the amplitude spectrum (Figure 2.1.1). We also estimate the uncertainty of strike, dip, rake and depth using bootstrap method (Cesca et al., 2010).

The second step is also assuming a DC source, but using waveform data in time domain. We fix the values of seismic moment, strike, dip, and rake obtained from Step 1 and invert for relative epicenter location and occurrence time to the original input values. We first compute the waveform fit between synthetic and observed waveforms using the four results of the last step and keep the one with the lowest misfit as the initial input. Then a regular grid search is applied to look for the best solution around this initial input. The fine mesh expands from -2 to 2 km with an increment of 0.2 km in both horizontal directions and -0.2 to 0.4 second with an increment of 0.1 second for the original time, resulting in a total of 3087 grid points. Again, due to the ambiguity of the primary and auxiliary fault planes for a DC model, we will have two equivalent focal solutions (Table 2.2.1, Figure 2.1.3).

The last step is to invert the six independent moment tensor components with time domain waveforms. Hypocenter location, original time, and the seismic moment are fixed and 128 initial solutions are generated with combinations of positive and negative moment tensors components from the DC solutions of previous step. Moreover, seven specific types of sources will be added to initial model solutions: one with Mxx=Myy=Mzz≠0 and Mxy=Mxz=Myz=0 (i.e., an explosion), three dipole sources with only one non-zero diagonal component: Mii≠0 and Mjj≠ii = Mkk≠ii =Mjk=0 (i=x,y, and z), and three compensated linear vector dipole (CLVD) sources. Then each of these initial models is inverted using the LM algorithm and the final solution is the one with the lowest misfit (Table 2.2.1). The MT solution is then decomposed into isotropic parts (ISO), DC, and CLVD parts and calculated the corresponding weighting in percentage (Figure 2.1.3).

2.1.3 Tectonic setting and seismic data

The Basel area is located at the Southeastern margin of the Upper Rhine Graben (Häring et al., 2008), intersecting with Jura Mountains in its south. Due to the long term geological deformation, it is one of the most active seismic regions in central Europe, well known for its destructive event in 1356 (Mayer-Rosa and Cadiot, 1979). The orientation of principal stresses measured at the borehole Basel 1 at granite part is about 144o for Shmax and 54o for Shmin, respectively (Häring et al., 2008; Dyer et al., 2008). It is also accord with the orientation

12

of maximum horizontal stress derived from natural seismicity in this region (Deichmann et al., 2000). Roughly speaking, the 5-km-deep injection well penetrates throughout a 2.4 km of sedimentary rocks on the top and 2.6 km of granitic basement (Häring et al., 2008). Since the stimulation, more than 10500 events have been observed near the injection well at granitic basement (Deichmann and Giardini, 2009), forming a lens-shaped seismicity cloud elongated at SSE direction. In this study, we choose events with ML≥2 from the catalog of Deichmann and Giardini (2009), resulting 19 of them (Table 2.2.1).

The seismic networks in this area are operated by three institutions/companies: Schweizerischer Erdbebendienst (SED), Landeserdbebendienst Baden-Württemberg (LED), and Geothermal Explorers Ltd. (GEL). The networks are composed of the different instruments, including both broad-band seismometers and accelerometers, with various sampling rate (from 62.5 Hz up to 1000 Hz). More detailed information about the station and network near the site can be found at Deichmann and Giardini (2009). We first select seismic stations within 15km to the injection well, which helps to reduce the affect of lateral heterogeneity of velocity structure for stations at distance. There are a total of 40 seismic stations within this range (Figure 2.1.1). Among them, six are borehole sensors from GEL installed near the well, which provide great details about source locations and mechanisms of the induced seismicity (Häring et al., 2008; Dyer et al., 2008; Deichmann and Ernst, 2009; Goertz-Allmann et al., 2011). Current version of Kiwi can only handle data recorded at the same depth. Since there are more surface stations and borehole sensors are installed at different depths (ranging from 317 m to 2740 m), the seismic data from borehole sensors are excluded in the inversion. In addition, another six stations have timing problem (WL2, WL7, WL9, WL10, CHBPF and LOER) and hence are disregarded. As a result, we have a total of 28 stations, 3 of which are 1-second Lennartz seismometers and the rests are temporal/permanent accelerometers.

To prepare the dataset, the original acceleration/velocity seismograms are converted into displacement after removing the instrument response using SAC (Goldstein and Snoke, 2005; Goldstein et al., 2003) and resampled into 100 Hz. Next, we remove the seismograms with signal to noise ratio (SNR) less than 5 and visually inspect all remaining waveforms. During the Kiwi inversion, the data are tapered in the time domain with a 6-second trapezoidal window containing both P and S arrivals and then bandpass filtered at 1 to 4 Hz. This relative low frequency range helps to remove the influence of unmodeled structures in our 1D layered model and our test shows its results match the observation much better than higher frequency one. The initial locations and occurrence time of each event are obtained from Deichmann and Ernst (2009) and are also listed in Table 2.2.1.

2.1.4 Results

Inversion result often depends on assumed mathematical model and the statistic distribution of noises within the data. To search the optimal dataset, we simply take all data within an epicenter distance for inversion and calculate the misfit values between synthetic waveforms and real observations for all 19 events. We repeat this process for a range of epicenter distances from 1 to 10 km with an increment of 1 km. The results are plotted in Figure 2.1.2 for both DC (Figure 2.1.2a) and MT (Figure 2.1.2b) solutions. As shown in Figure 2.1.2, the mean value of misfit for both DC and MT solutions remains more or less constant until 5 km, where it clearly increases and keeps flat again. It is reasonable to use the one with relative lower misfit value and meanwhile keep as much data as possible. Thus, we use 4 km as a

13

threshold for station selection, which includes nine 3-component accelerometers SBAF, WEIL, OTTER, SBAP, SRHB, SBAT, SBAE, SBAJ, and SBAM (Figure 2.1.1). However, 15 out of 19 events have data recorded at the first six listed stations.

Figure 2.1.3 shows results after steps 2 and 3 for event id 39 (Table 2.2.1) and its synthetic waveforms of MT solution are plotted with bp filtered data in Figure 2.1.4. In general, both waveforms match quite well. Since our dataset is also analyzed in Deichmann and Ernst (2009) with different method and velocity models, it is intuitive to compare our DC results to theirs (Figure 2.1.5). In general, the two agree with each other very well, except few events, such as event id 113. It is also interesting to check whether the projected first-motion polarities from Deichmann and Ernst (2009) match our solutions. As illustrated in Figure 2.1.5, majority of projection points fall into the right quadrants. This is certainly improved if we look at the results of MT solutions (Figure 2.1.5), which is also in consistent with the fact that misfit values of MT solutions are lower than that of DC solutions (Figure 2.1.2). This is not surprising, since MT solutions are inverted from the best DC solution as mentioned in the Section 2. Improvement of misfit for MT solutions might indicate the existence of non-double-couple type source mechanism. Note that most of our solutions are obtained with data from only six stations, whereas the polarities of nearly 40 stations are used in Deichmann and Ernst (2009). It is also the first time that Kiwi is applied to acceleration data with a local network.

2.1.5 Discussion

Figure 2.1.4 demonstrates both synthetic and real S waves of event 39 are generally in phase except the North component of station WEIL, which has relative weak amplitude. However, synthetic P arrivals are systematically later than real P waves at all stations, which has also been observed for other events. Since S waves have the strongest and consistent energy among different stations, it is expected that inversion program will search the best solution aligning at S phases. Thus, delayed synthetic P arrivals imply that our 1D model does not represent the true velocity structure. In addition, many surface accelerometers sit at sediment layers and their local site effects could certainly not be taken into account in the model, which was developed for molasse bedrock in Basel (Ripperger et al., 2009). All these pose an severe issue for this study: how the inaccuracy from the velocity model will be mapped into our final solutions? Filtering data at low frequency and selecting stations within small epicenter distance make the result less sensitive to the velocity structure, which is in fact confirmed by the great similarity between our DC solutions and that of Deichmann and Ernst (2009) and that most estimated polarities fall into the right quadrants of MT solutions (Figure 2.1.5). Note that Deichmann and Ernst (2009) apply several 1D velocity models to different stations to imitate the 3D velocity structure in the region.

Moreover, we performed the inversion on another velocity model with a constant Vp/Vs ratio of 2.1 and keep the same Vp profile. The reason we test this new model is that the median value of Vp/Vs estimated from these 19 events is about 2.1 assuming Vp/Vs≈∆Ts/∆Tp. Here ∆Tp and ∆Ts are travel time of P and S waves, respectively. This value (i.e., 2.1) is clearly deviated from our 1D velocity model, whose Vp/Vs is about 1.73 at all layers (see Table 2.1.2 of Ripperger et al., 2009). DC solutions from this new model are identical to the ones before and the differences mostly appear on MT solutions. Nevertheless, MT result from this new model is actually worse in terms of waveform fit, so we decide to keep the previous ones. This small test suggests the stability of our results with regards to certain variation of velocity

14

models, especially for DC ones. In the future, proper velocity analysis would be necessary in order to remove discrepancy between data and model, which is next plan for this study. Further development of inversion algorithm capable of incorporating 3D velocity model is useful for complicated areas (e.g., Gharti et al., 2011).

Due to the problem of velocity model, we prefer to use the hypocenter location and original time from Deichmann and Ernst (2009), which is sensitive to the assumed velocity model, but keep MT solutions for the interpretation of results. ISO components of MT solutions links to the volume change in the source region, with positive and negative values presents the crack opening and closure, respectively. Plot of temporal evolutions of ISO components (Figure 2.1.6) show that positive ISO dominates the injection period until a few hours after the shut-in of the well, and followed by two events with negative values. For later events occurred on 2007, their ISO become negligible except one event (id 184, Table 2.1.2). Figure 2.1.7 illustrates that events with similar ISO pattern (open, close, and neutral) are generally clustering in space. For example, events with considerable ISO values (i.e., yellow, red and blue dots) are mostly located near the wellbore, whereas ones with smaller ISO values extend roughly from 400m to 700m away from the wellbore. In addition, majority negative ISO events are shallower than the positive ones (Figure 2.1.7b).

Goertz-Allmann et al. (2011) estimated stress drop from about 1000 microseismic events in Basel and they observed a trend of increasing stress drop as distance to the injection point. This trend correlates negatively with modeled diffusive pore pressure perturbation up to 300m. They explained reduction of stress drop near the injection point as increase of pore pressure and vice versa. Similarly, spatio-temporal distribution of ISO from our study (Figures 2.1.6 and 2.1.7) can be interpreted as variation of pore pressure due to the stimulation. Higher injection rate in the beginning likely raised pore pressure near the injection point, which could be large enough to partially open cracks. Following the sudden stop of injection on 8 December 2006, some opened cracks were force to close as the drop of pore pressure near the well within few days. Due to the irregular occurrence time of our events, it is difficult to define a clear boundary in time between the ‘open’ and ‘closure’ phenomena. For events occurred in 2007, pressure has fallen back to ambient level and the propagation of diffusive fluid front in the large distance can only reduce the normal stress and cause shear slips. One difference between ours and Goertz-Allmann et al. (2011) is that the transition distance from strongly to weakly pressure affected regions is ~400m along-strike distance to the well (Figure 2.1.7b), but it is 300m to the injection point from theirs. However, this could arise from different event catalogs used in two studies and assumed fluid dynamic model and model parameters. Härling et al. (2008) mentioned that the hydrostatic pore pressure does not exceed minimum horizontal stress, which implies no or few tensile cracks should be generated during the injection. In reality, this is strongly affected by stress fields and geological structures in the local scale. However, the fact that most of events have ISO less than 20% suggests shear slip still dominates source mechanisms.

During hydraulic stimulation, pore pressure can effectively reduce the normal stresses of rocks and make it more susceptible for shear fracturing (Scholz, 1990). Moreover, if pore pressure is large enough to exceed the minimum horizontal stress, it will open the crack, similar to Model I type fracture. This type of seismic source is often referred to as tensile earthquakes (Walter and Brune, 1993; Vavryčuk 2001). Tensile earthquakes might not generate strong shear wave amplitude as pure shear slip, but it potentially create more conductive fluid path. Thus, it is important to image the locations and temporal evolution of these induced fluid paths. Our study show MT inversion can provide unique information

15

about the volumetric changes in the stimulated rock, which is more related to extension of these conductive fluid paths than from seismic location only. In contrast to Deichmann and Giardini (2009), our results show the two 2007 events they discussed in a repeat cluster actually have ~0% ISO (Table 2.1.2), which does not support the hypothesis of non-double-couple for both earthquakes. Additional evidence of existence and distribution of tensile cracks during the 2006 Basel geothermal stimulation can be verified by systematic analysis of amplitude ratio of P and S waves (Walter and Brune, 1993) in the future.

In summary, we have performed waveform inversion to retrieve both DC and MT source parameters for 19 strongest induced microseismic events in city of Basel. The DC solutions agree well with the result from a previous study based on the projection of first-motion polarities at ~40 stations. In comparison, data from less than ten (mostly six) stations has been used in our study. The MT solutions indicated the existence of significant ISO including both crack opening and closing, for events occurred during and immediately after the injection but not the ones later. The spatio-temporal pattern of ISO of MT solutions can be explained by the perturbation of pore pressure. Additionally, better velocity model need to be constructed in order to obtain accurate MT solutions.

2.1.6 Acknowledgements

This work would not be possible without the contribution from Nicholas Deichmann, who provides us the waveform data, instrument response and the first-motion polarities for all 19 events in Deichmann and Ernst (2009). We are also thankful to Geopower Basel AG and the Landeserdbebendienst Baden-Würtemberg for access to waveform data. P.Z. would like to thank Myrto Pirli for helping on the instrument correction for the dataset. Parts of this work are funded by the EU Project GEISER Grant Agreement no. 241321-2.

2.1.7 References

Cesca S, Heimann S, Stammler K, Dahm T (2010) Automated procedure for point and kinematic source inversion at regional distances. JGR 115, B06304, doi: 10.1029/2009JB006450. Deichmann N, Ernst J (2009) Earthquake focal mechanisms of the induced seismicity in the 2006 and 2007 below Basel (Switzerland). Swiss. J. Geosci. 102, 457-466 Deichmann N, Gardini D (2009) Earthquakes induced by the stimulation of enhanced geothermal system below Basel (Switzerland). Seismological Research Letters 80(5),784-798

Dyer BC, Schanz U, Ladner F, Häring MO, Spillmann T (2008) Microseismic imaging of a geothermal reservoir stimulation. The Leading Edge 27, 856-869.

Gharti HN, Oye V, Kühn D, Zhao P (2011), Simultaneous microearthquake location and moment-tensor estimation using time-reversal imaging, SEG Expanded Abstracts 30, 1632; doi:10.1190/1.3627516.

16

Goertz-Allmann BP, Goertz A, Wiemer S (2011) Stress drop variations of induced earthquakes at the Basel geothermal site. Geophys. Res. Let. 38, L09308, doi:10.1029/ 2011GL047498.

Goldstein P, Snoke A, (2005) “SAC Availability for the IRIS Community”, Incorporated Institutions for Seismology Data Management Center Electronic Newsletter.

Goldstein P, Dodge D, Firpo M, Minner L (2003) “SAC2000: Signal processing and analysis tools for seismologists and engineers, Invited contribution to “The IASPEI International Handbook of Earthquake and Engineering Seismology”, Edited by WHK Lee, H. Kanamori, P.C. Jennings, and C. Kisslinger, Academic Press, London.

Häring MO, Schanz U, Ladner F, Dyer BC (2008) Characterization of Basel 1 enhanced geothermal system. Geothermics 37, 469-495

Mayer-Rosa D. and Cadiot B (1979): A review of the 1356 Basel earthquake, Tectonophysics, 53, 325-333.

Phillips WS, Rutledge JT, House LS, Fehler MC (2002) Induced microearthquake patterns in hydrocarbon and geothermal reservoirs: six case studies. Pure Appl. Geophys. 159, 345-369.

Walter, W. R., and J. N. Brune (1993), Spectra of Seismic Radiation From a Tensile Crack, J. Geophys. Res., 98(B3), 4449–4459, doi:10.1029/92JB02414.

input data source model Inversion algorithm Num of initial solutions inverted parameters Num of final solutions

step1 amplitude spectrum double‐couple source Levenberg‐Marquardt 452 strike, dip, rake, depth and seismic moment 4

step2 displacement waveform double‐couple source grid search 3087 latitude, longtitude and original time  2

step3 displacement waveform full moment tensors Levenberg‐Marquardt 135 6 MT components 1

Table 2.1.1: Summary of key parameters at each step for our inversion method. A point source is assumed for all 3 steps.

17

Event ID Time Longitude Latitude Depth (km) Magnitude ISO (%)

39 2006/12/06 22:27 7.593 47.587 4.2 2.2 25

86 2006/12/08 03:06 7.595 47.585 4.1 2.6 4

87 2006/12/08 03:24 7.594 47.583 4.8 2.3 11

94 2006/12/08 09:04 7.593 47.585 4.8 2.2 22

98 2006/12/08 11:36 7.596 47.584 4.6 2.2 13

102 2006/12/08 15:13 7.595 47.583 4.7 2 16

104 2006/12/08 15:31 7.595 47.585 4 2.1 9

105 2006/12/08 15:47 7.593 47.588 4.1 2.7 9

108 2006/12/08 16:49 7.593 47.584 4.7 3.4 13

112 2006/12/08 19:27 7.596 47.582 4.7 2.3 0

113 2006/12/08 20:20 7.594 47.583 4.9 2.6 36

147 2006/12/10 06:11 7.595 47.584 4 2 15

159 2006/12/14 22:39 7.595 47.584 4 2.5 7

168 2007/01/06 07:20 7.596 47.582 4.2 3.1 1

170 2007/01/12 03:35 7.597 47.581 4.2 2.2 1

174 2007/01/16 00:09 7.596 47.582 4.1 3.2 0

176 2007/02/02 03:54 7.596 47.582 4 3.2 2

184 2007/03/21 16:45 7.596 47.581 4 2.8 8

185 2007/05/06 00:34 7.596 47.581 4 2.3 2

Table 2.1.2: Catalog of 19 events analyzed in this study. The event id is the same as that of Deichmann and Ernst (2009) and the last column lists the weight of isotropic components for moment tensor solutions in percentage.

18

Figure 2.1.1: Map view of Basel area. Seismic stations around the injection well and 19 analyzed events are marked by triangle and gray circles, respectively. The solid black triangles present the nine stations whose data are used for inversion with the corresponding names labeled (Reference to the text for more details). The inset shows a zoom-in map around the injection well, which is plotted as an open circle with a black dot.

19

Figure 2.1.2: Misfit values of DC (a) and MT (b) solutions vs. the maximum epicenter distance for stations selections. For every epicenter distance, gray asterisks present the misfit for each event, whose mean value is marked by a black diamond and the length of error bar is proportional to its standard deviation.

20

Figure 2.1.3: Example of final solutions after Step 2 (top) and Step 3 (bottom) for event id 39. (Top) Inversion results include relative epicenter location and relative original time estimations. The size of gray dots increases with fit quality and the red circle mark the best fit. Note that due to the ambiguity of primary and auxiliary fault planes, there are two sets of solutions for strike, dip, and rake. (Bottom) The full MT solution is decomposed into isotropic (ISO), double-couple (DC), and CLVD parts and their corresponding weighting is marked in percentage as well. For all source mechanisms, black and white colors present the compressional and dilatational quadrants, respectively.

21

Figure 2.1.4: Example of three-component synthetic waveforms (black lines) from MT solutions of event id 39 (Figure 2.1.3 bottom) are plotted on top of corresponding true observations (red lines). The waveforms are bp filtered at 1-4 Hz and trace normalized for better comparison. The station names, epicenter distance, and azimuth are listed on the left. The gray trapezoid in each window represents the applied time taper.

22

Figure 2.1.5: Comparison of fault-plane solutions from Deichmann and Ernst (2009) (top panel), focal mechanisms from DC (middle panel) and MT (bottom panel) inversion in this study. The first-motion polarities from Deichmann and Ernst (2009) are plotted on top of each solution: the solid circles correspond to compressive first motion (up) and open circles to dilatational first motion (down).

23

Figure 2.1.6: Plot of percentage of isostropic components of MT solutions vs. occurrence time of each event in the logrithmic scale since 6 December, 2006. The size of circle is proportional to the magnitude of corresponding event.

24

Figure 2.1.7: ISO component of each event in map view (a) and cross-section view along the main strike of seismic cloud (i.e., 155o) (b). In both figures, event locations are obtained from Deichmann and Ernst (2009) and marked by a circle scaled to its magnitude with filled color presents of ISO value in percentage. In the map view, the injection well is indicated by an open circle with a black dot, whereas in the cross-section view the vertical line marks the cased (thick) and open (thin) section of the borehole.

25

2.2 Identification of faults activated during the stimulation of the Basel geothermal project from cluster analysis and focal mechanisms of the larger magnitude events.

Report for deliverable 3.2 of EU project GEISER, March 2012 Nicholas Deichmann, Schweizerischer Erdbebendienst, ETH Zürich.

2.2.1 Introduction

To stimulate the reservoir for an enhanced geothermal system (EGS) initiated by a private/public consortium in the city of Basel, Switzerland, approximately 11,500 m3 of water were injected at high pressures between December 2nd and December 8th 2006 into a 5-km-deep well below Kleinhüningen (Häring et al. 2008). A six-sensor borehole array, installed by the operators of the project at depths between 317 and 2,740 meters around the well to monitor the induced seismicity recorded more than 10,500 seismic events during the injection phase. Hypocentral locations could be calculated for more than 3,000 of these events. The gradual increase in flow rate and wellhead pressure was accompanied by a steady increase in seismicity, both in terms of event rates and magnitudes. In the early morning hours of December 8th, after water had been injected at maximum rates in excess of 50 l/s and at wellhead pressures of up to 29.6 MPa for about 16 hours (Häring et al. 2008), a magnitude ML 2.6 event occurred within the reservoir. This exceeded the safety threshold for continued stimulation, so that injection was stopped prematurely. In the afternoon and evening of the same day, two additional events of magnitude ML 2.7 and 3.4 occurred within the same source volume. As a consequence, the well was opened and in the following days about one third of the injected water volume flowed back out of the well (Häring et al. 2008). Though the seismic activity declined rapidly thereafter, three more events with ML > 3 occurred in January and February 2007. In this report we present results of an ongoing analysis of the larger magnitude events that were induced by the stimulation of the Basel enhanced geothermal system. In this context, by larger magnitude events we mean all those seismic events that were recorded not only by the local borehole network installed by the project operators, but also by the regional seismometer and local surface accelerometer networks of the Swiss Seismological Service and the Landeserdbebendienst of Baden-Württemberg. The goal is to examine the role that these larger events play in the stimulation process, by mapping the faults on which they occurred. As discussed in greater detail by Deichmann & Giardini (2009), the temporal evolution of the seismic activity induced by the Basel geothermal project can be subdivided into three periods. The first from December 2nd to December 8th corresponds to the period of active stimulation and ends when the well was vented; 108 larger magnitude events, according to the definition given above, occurred in this first period. The second, which lasted until the end of December 2006, is characterized by a steady decrease of activity both in terms of magnitude and of number of events, as part of the water flowed back out of the well. Another 57 events occurred during this second period. The third began in January 2007 with a renewed increase in seismic activity that was followed by a gradual decline in spring and summer. The last of the 195 seismic events recorded by the regional networks occurred at the end of November 2007. Here we restrict our analysis to the 165 events recorded by the regional networks during

26

the stimulation and the immediate post-stimulation period between December 2nd and December 31st 2006. As already noted previously by several different authors, a substantial part of the seismicity induced by the stimulation of the Basel geothermal project occurred in clusters of events with similar waveforms, or so-called families of similar events. This implies that the hypocenters of the events in each cluster must be located very close to each other and that their focal mechanisms must be nearly identical. Very commonly, the hypocenters of events in such clusters are located on a plane that coincides with one of the nodal planes of their focal mechanism. In the present analysis, we take advantage of the high signal similarity and apply a cross-correlation procedure between the signals of the different events to obtain precise relative arrival times. These arrival times serve as input for a master-event location technique to compute high-precision relative hypocenter locations that can be compared to focal mechanisms, which are based on first-motion polarities observed at both the local borehole seismometers and the regional surface networks.

2.2.2 Tectonic setting

Basel is located at the southern end of the Rhine Graben, where it intersects the fold and thrust belt of the Jura Mountains of Switzerland (Figure 2.2.1 and Figure 1 of Valley & Evans 2009). As such it is an area that in the geologic past has seen both extension (rifting phase of the Rhine Graben) and thrusting (folding of the Jura Mountains). A recent comprehensive summary of the evolution of the Upper Rhine Graben and Jura Mountains through geologic time, together with an exhaustive reference list, can be found in Ustaszewski & Schmid (2007). The borehole itself is situated at the southern end of the Rhine Graben and reaches a depth of 5 km below the Earth’s surface. As shown in the lithological section reproduced in Häring et al. (2008) and in Valley & Evans (2009), it penetrates a 2426 m thick sedimentary sequence before entering the crystalline basement.

2.2.3 Seismic networks

The seismic data available for the Basel geothermal project and analyzed in this article were recorded by several different seismometer and accelerometer networks operated by three separate institutions, Schweizerischer Erdbebendienst (SED), Landeserdbebendienst Baden-Württemberg (LED) and Geothermal Explorers Ltd. (GEL). The locations are shown in Figure 2.2.1. Detailed documentations of the instruments and digital data acquisition systems can be found in the articles by Deichmann and Ernst (2009) and by Deichmann and Giardini (2009). It should be noted that accelerometers at epicentral distances of a few km installed at the Earth’s surface in the middle of a noisy city such as Basel can provide good-quality data even for events with magnitudes ML < 1.

2.2.4 Cluster definition

27

Figure 2.2.2 shows a typical example of similar seismograms of a set of events that belong to the same cluster, recorded at one of the borehole seismometers. A common procedure to identify seismic events that belong to a common cluster is based on an analysis of the cross-correlation coefficients between signals of different events recorded at a given station. However, both the threshold of the cross-correlation coefficient for including or excluding an event from a given cluster and the filters used to limit the frequency bandwidth for the cross-correlation are somewhat arbitrary. Higher signal frequencies and higher correlation coefficients are more restrictive than lower frequencies and a lower threshold. A restrictive choice finds only clusters with highly similar signals and thus with source locations that are so close to each other that within the obtainable location precision it can be impossible to identify a planar distribution of hypocenters. A less restrictive choice, on the other hand, tends to increase the spatial extent of the clusters, but entails the risk of associating events that actually do not share a common source. Consequently, the choice of the appropriate parameters depends on the goal of the analysis and needs to be made by an iterative trial and error procedure, based on visual inspection of the signals, on a comparison with the available focal mechanisms and on the results obtained from the subsequent relative locations. For our purpose, we found that the correlation of the signals recorded by the A-component of the borehole station OTER1, filtered with a causal 2nd order Butterworth band-pass filter between 1 and 20 Hz, and a correlation coefficient threshold of 0.9 produced a useful initial cluster selection. In some cases, visual inspection of the signals and comparisons with the corresponding focal mechanisms showed that the clusters obtained in this way needed to be split into two or more sub-clusters. In other cases, two families that were identified as being separate based on the chosen correlation threshold, in the end could be grouped under the same cluster. The results presented in this report concern 10 clusters with a minimum of 5 and a maximum of 13 events.

2.2.5 Relative locations

Starting point for the present analysis is the set of hypocenter locations documented in the article by Deichmann and Giardini (2009). These locations are the result of a master-event technique based on visually determined arrival times, determined mainly from the seismograms recorded by the six borehole seismometers. In this procedure, the travel-time residuals of a chosen master-event are used as station corrections for locating the hypocenters of all other events. As master-event we chose an event (2006/12/03 19:51 UTC, ML 1.7) that was recorded also by a seismometer temporarily deployed close to the bottom of the injection borehole, and its location was fixed at the location obtained by GEL using a 1-D velocity model with station corrections (e.g. Häring et al. 2008). The mean standard deviation of the locations obtained in this way are on the order of 50 m horizontally and 70 m vertically (Deichmann and Giardini, 2009). The next step consists of choosing a master-event for each of the ten identified clusters and of performing signal cross-correlations of all other events pertaining to the given cluster with the master-event as well as among each other. As shown in Figure 2.2.3, these correlations are performed separately for the P- and S-phases recorded at each station. The signal lengths used for the correlations are chosen long enough for a stable result and short enough to include only the direct P- or S-waves. For all the clusters documented in this report, we used signals recorded by the borehole sensors as well as by surface seismometers and accelerometers at epicentral distances out to 40 km. As explained in the Appendix to the article by Deichmann

28

and Garcia-Fernandez (1992), the multiple and thus redundant cross-correlations constitute a quality check of the correlations and allow one to compute least-squares adjusted travel-time differences between the master-event and each other event in the given cluster. The relative timing precision obtained by means of this procedure is on the order of 1-2 ms. By nature of the algorithm, cross-correlations align the maxima and minima of two signals. For events with large magnitude differences and consequently with different dominant signal frequencies, the resulting alignment of the signals does not always correspond to the phase onsets. For the same reason, to guarantee consistency over all stations, acceleration traces must always be integrated to velocity before correlation. Locations of the individual events within a given cluster relative to the corresponding master event are calculated with an algorithm proposed by Console and DiGiovambattista (1987). It is based on the fact that, for hypocenters which scatter over a volume that is very small relative to the distances to the recording stations, the angles at which the rays leave the source to each station are essentially the same for all events. As a consequence, the otherwise non-linear earthquake location problem becomes linear. Given a set of take-off angles for the master-event, the only seismic velocities that affect the results are the P- and S-velocities in the immediate source volume, and their uncertainties contribute only little to the final location error. With more than 20 travel-time differences for each event and a precision of 1-2 ms, the computed standard relative location errors are on the order of 2-6 m. However, it should be noted that for some of the stronger events, the actual error might be larger than this. As explained in the previous paragraph, the reason is that correlations, rather than representing differences in onset times, match the arrival times of the most energetic part of the P- or S-phase. Whereas the hypocenter is defined as the single point on the fault where the rupture initiates, the most energetic waves radiated during the rupture process emanate from a broader area, which for larger events is not necessarily the same for stations that see the fault from different directions. Such effects might lead to additional errors that are not accounted for in the standard errors computed by the location algorithm.

2.2.6 Focal mechanisms

All focal mechanisms in this study are determined from first-motion polarities observed both at the borehole stations and at the stations of the surface networks. The adopted procedure to calculate the take-off angles in the presence of the very heterogeneous seismic velocities below Basel is documented in detail in the publication by Deichman and Ernst (2009). This earlier publication presents the focal mechanisms of the 28 strongest events. In contrast, the data set underlying the present study comprises more than 150 events and includes also events with ML < 1. Of course, for these weaker events, the number of reliable first-motion data points is smaller than for the stronger events, and in principle the range of possible nodal-plane orientations and consequently also the uncertainty of the mechanism are larger. However, in cases where these smaller events have essentially identical signals as a larger event with a well-constrained fault-plane solution, and are thus part of the same event family, the focal-mechanism parameters of the smaller event are set equal to those of the corresponding larger event.

29

2.2.7 Results

The list of the events that are part of the ten clusters identified in this study and the corresponding focal mechanism parameters are listed in Table 2.2.1. The clusters are named either after their strongest event or after the corresponding master event. As an example of the steps leading to the final result, we present details of the analysis of Cluster 14. The initial cross-correlation of all seismograms recorded at station OTER1 actually identified the nine events of this cluster as pertaining to two separate event families. In fact, as can be seen in Figures 2.2.2 and 2.2.4, at some stations the polarities of the P-onsets are reversed, and thus for events 3, 8 and 15 one of the nodal planes of the focal mechanisms has a slightly different dip. However, as shown in the fault-plane solutions in Figure 2.2.4, the other nodal plane is identical for all events. Event number 1 was chosen as master event for the cross-correlations and the computation of the relative locations of the hypocenters. The correlations provide precise travel-time differences for 29-33 P- or S-arrivals for each event, and the mean travel-time residuals calculated by the relocation algorithm are on the order of 1-2 ms. As a consequence the computed uncertainties of the hypocenters relative to the master event are on the order of a few m. The results of the relocations are shown on an epicenter map and two vertical cross-sections in Figure 2.2.4. The hypocenters group into two sub-clusters (named 14 and 15, after their respective largest event) whose members match the observed differences in focal mechanism parameters very nicely. Within each sub-cluster the hypocentral locations scatter only over a very small volume: for sub-cluster 14, +/-1 m in x-direction, +/-5 m in y-direction and +/-9 m in z-direction; for sub-cluster 15, +/-1 m in x-direction, +/-2 m in y-direction and +/-4 m in z-direction. So, taken individually, it is not possible to resolve the active fault plane from the relative locations within the two sub-clusters. However, the two sub-clusters are displaced from each other by about 50 m along a line that matches the strike of the NW-SE striking nodal plane almost perfectly. If one assumes that the signal similarity of these events is evidence for their occurrence on a single common fault, then these events occurred as dextral strike-slip motion on an approximately NW-SE striking near-vertical fault. An alternative way of visualizing these results is shown in the two perspective plots in Figure 2.2.5. There each earthquake source is represented by a circle with strike and dip of one of the nodal planes of the focal mechanism and a size equal to an estimate of its source radius. The latter is proportional to the cube-root of the seismic moment divided by the cube-root of the static stress drop. In our case, we derive the seismic moment from the local magnitude ML with the empirical relation derived for the Basel induced seismicity by Bethmann et al (2010). For the stress drop we arbitrarily assume a constant value of 3 MPa for all events. Since the stress drop enters into the computation of the source radius as the cube root, actual deviations from these assumptions have only a relatively small effect on the source radius estimate (1 MPa would imply a source radius that is 1.4 times larger, and 10 MPa, 1.5 times smaller). So, despite the arbitrariness of the assumptions and the uncertainties of the actual values of seismic moment and stress drop, plots such as those in Figure 2.2.5 help to outline the geometry of the potentially active faults. In the case of Cluster 14 shown in Figure 2.2.5, the rupture areas of the NW-SE striking nodal planes with almost identical strike and dip of the nine events coalesce within one or two standard errors of the locations onto a single plane, while the alternative nodal planes would correspond to two separate faults about 50 m apart. Given the two alternatives and in view of the fact that the seismograms recorded at most stations are almost identical for all nine events, it is more likely that their sources lie on the same fault (even though their rake is slightly different) rather than on two completely separate faults that happen to have exactly the right orientation to produce nearly identical signals.

30

Thus in this case, the NW-SE striking nodal planes represent the fault that was activated during these events. The fault planes chosen for each cluster are shown in the perspective plot in Figure 2.2.6. The dimensions correspond either to the extent of the cluster as a whole or to the size of the largest event in the cluster. Their strike and dip are listed in Table 2.2.1. The underlying criteria for choosing one nodal plane rather than the other as the active fault plane are, firstly, close to identical strike and dip for all events in the cluster and, secondly, the tightness of the cluster when viewed along the strike of the nodal planes. For Cluster 14, discussed in the previous paragraph, as well as for Clusters 43 and 87 the chosen planes scatter over less than 10m rather than over about 50 m for the other planes. For Clusters 39, 88 and 102, the chosen planes lie within less than 20 m versus more than 80 m for the alternatives. Thus for these six clusters, these criteria clearly favor the chosen nodal plane rather than the other. In the case of Cluster 5, the hypocenters extend over a volume of 5 x 8 m horizontally and 20 m vertically (including location uncertainties). This is so small that either of the two sets of nodal planes, when viewed in a depth cross-section, could collapse within location errors onto a single plane. However, the hypocenters align more closely with the vertical E-W striking nodal plane than with the inclined N-S striking plane. So it is the former that is chosen as the active fault plane. Viewed along the N-S striking nodal plane, Cluster 135 extends down-dip for more 200 m and across dip over about 40 m. The latter is considerably more than the expected location error; so it is possible that this cluster consists of more than one fault. Nevertheless, because the focal mechanisms of all events are almost identical, and for the sake of simplicity of the illustration, this cluster is represented by a single plane in Figure 2.2.6. Viewed along strike of the alternative fault planes, the hypocenters scatter over more than 200 m, and this would imply that each of the eight events occurred on a separate fault, which is highly unlikely considering the similarity of the signals. Cluster 82, named after its master event, includes the ML 3.4 mainshock of December 8th 2006. According to the relative locations of Deichmann and Giardini (2009), which are based only on cross-correlations of the six borehole stations, the active fault strikes WNW-ESE and dips with about 75 degrees towards the SSW. In the present study, this cluster was reanalyzed using also the data of the surface stations and including an additional event (number 113 in Table 2.2.1). Deichmann and Giardini (2009) already noted that the focal mechanisms of the events in this cluster are not identical and that this could be symptomatic of the fact that these events do not all lie on a single planar structure. Indeed the results of the reanalysis suggest that this cluster consists of at least three more or less parallel faults that span a distance of about 75 m. Nevertheless, the fault plane identified by Deichmann and Giardini (2009) is still judged to be the more likely fault, since, viewed along the strike of the alternative nodal plane, the scatter of the events is twice as large (150 m), so that even the events with the most similar signals do not come to lie on a common fault. The results for Cluster 86 are the most difficult to interpret unequivocally. This cluster comprises seven events, including three events with ML between 2.0 and 2.6. Their focal mechanisms offer the choice between a N-S or an E-W striking fault plane; the inferred dip of the former varies between 75 and 78 degrees, while of the latter it varies between 61 and 80

31

degrees. Viewed along strike of either of these planes, the hypocenters scatter over a width of about 50 m, so that we cannot invoke the tightness criterion to favor one over the other. Either choice implies the existence of two or possibly three more or less parallel faults within this cluster. Nevertheless in Figure 2.2.6, this cluster is included as two E-W striking planes, on the basis of the observation that the dip of the corresponding nodal planes is practically identical for all events, whereas the dip varies by almost 20 degrees for the N-S striking nodal planes.

2.2.8 Conclusions

With the premise that larger magnitude events as defined here are all the events detected by the regional surface networks around Basel and that they include events with ML as small as 0.7, the results of the analysis presented in this report can be summarized as follows.

The ten earthquake clusters with at least five events each, analyzed so far, include more than half of the larger magnitude events induced in December 2006 by the stimulation of the Basel enhanced geothermal system. The induced seismicity in this same time period includes two or three additional clusters with three or four events each as well as a dozen doublets, which might be amenable to the same analysis.

In most of the cases presented in this report, the high precision relocations of the

events within each cluster allow one to identify the active fault plane with good confidence. The locations together with rough estimates of source size for each event define faults with dimensions between less than 100 m and several hundred meters. Accepting that the microseismic cloud as a whole is near-vertical and has a predominant NNW-SSE orientation, we note that, except for Clusters 88 and 102, the strike or dip of the identified faults deviates more or less strongly from this overall orientation. The concept of a single fault zone with a more or less constant orientation is obviously too simplistic. This has important consequences for models of fluid migration during stimulation and thus of the process of permeability enhancement as well as for seismic hazard assessments.

Except in the case of Cluster 88, which was active for only 24 hours, activity in each

individual cluster spans several days. The two sub-clusters 14 and 15 as well as cluster 5, which are the closest to the borehole, were active only in the first three days of the stimulation and included only smaller events (ML between 0.8 and 1.7). Moreover the events in each of these three clusters are so closely co-located and their signals are so similar that they must represent repeated slip on exactly the same fault patch. This is only possible if each event releases merely a fraction of the total stress drop and each fault patch is repeatedly reactivated, as it is increasingly weakened due to the fluid pressures rising with time. Activity of other clusters is due to a combination of repeated slip on the same fault patches and activation of neighboring patches on the same fault. It is important to note that, of the ten multi-event clusters analyzed in this study, only cluster 135 occurred entirely during the post-stimulation period. Moreover, among the 30 earthquakes recorded by the surface networks in 2007, the same selection criteria as applied to the events of December 2006 found only two event-doublets and not a single larger cluster. Evidently, the propensity of seismic activity to

32

The still ongoing detailed analysis of the seismicity induced by the stimulation of the Basel enhanced geothermal system will be expanded to include also the weaker events, recorded only by the borehole network, as well as the events that occurred in the post-stimulation period during 2007.

2.2.9 Acknowledgements

A large part of this study was undertaken in the context of project GEOTHERM, financed by the Competence Center for Environment and Sustainability (CCES) of ETH, and thus constitutes an in-kind contribution to the GEISER project. Access to the borehole data of Geopower Basel, acquired by Geothermal Explorers Ltd., is gratefully acknowledged. Thanks are due also to Keith Evans and Toni Kraft for helpful discussions.

33

2.2.10 References

Bethmann, F., N. Deichmann, and P. Mai (2011): Scaling relations of local magnitude vs. moment magnitude for sequences of similar earthquakes in Switzerland. Bull. Seismol. Soc. Am., 101, 2, 515–534, 2011, doi:10.1785/0120100179 Console, R., and R. DiGiovambattista (1987). Local earthquake relative location by digital records. Physics of the Earth and Planetary Interiors 47, 43–49. Deichmann, N., Ernst, J. (2009): Earthquake focal mechanisms of the induced seismicity in 2006 and 2007 below Basel (Switzerland). Swiss J. Geoscience, 102/3, 457-466, DOI: 10.1007/s00015-009-1336-y Deichmann, N., Giardini, D. (2009): Earthquakes induced by the stimulation of an enhanced geothermal system below Basel (Switzerland). Seismological Research Letters 80/5, 784-798, doi:10.1785/gssrl.80.5.784. Deichmann, N., Garcia-Fernandez. M.: Rupture geometry from high-precision relative hypocenter locations of microearthquake clusters. Geophys. J. Int., 110, 501-517, 1992. Häring, M. O., Schanz, U., Ladner, F., Dyer, B. C. 2008: Characterization of the Basel 1 enhanced geothermal system. Geothermics 37, 469-495, doi:10.1016/j.geothermics.2008.06.002. Ustaszewski, K. & Schmid, S.M. 2007: Latest Pliocene to recent thick-skinned tectonics at the Upper Rhine Graben - Jura Mountains junction. Swiss Journal of Geosciences 100, 293-312. Valley, B. & Evans, K.F. 2009: Stress orientation to 5 km depth in the basement below Basel (Switzerland) from borehole failure analysis. Swiss Journal of Geosciences 102/3, DOI:10.1007/s00015-009-1335-z.

34

Table 2.2.1. List of the clusters analyzed in this report, with focal mechanism parameters for each event. The nodal planes are given in terms of strike, dip and rake, and the P- and T-axes with azimuth and plunge, in degrees (strike and azimuth, clockwise from North). In all cases, the fault plane identified as active is plane number 1.

EVN DATE TIME ML PLANE 1 PLANE 2 P-AXIS T-AXIS ---------------------------------------------------------------- cluster 14 1 2006/12/03 00:59 0.8 124/88/-178 034/88/-002 349/03 079/00 2 2006/12/03 04:08 1.1 124/88/-178 034/88/-002 349/03 079/00 3 2006/12/03 06:41 0.9 125/88/ 167 215/77/ 002 171/08 079/11 4 2006/12/03 17:15 1.1 124/88/-178 034/88/-002 349/03 079/00 6 2006/12/03 23:45 1.0 124/88/-178 034/88/-002 349/03 079/00 8 2006/12/04 02:04 1.0 126/88/ 174 216/84/ 002 171/03 081/06 14 2006/12/05 01:22 1.2 124/88/-178 034/88/-002 349/03 079/00 15 2006/12/05 01:54 1.1 125/88/ 167 215/77/ 002 171/08 079/11 20 2006/12/05 19:30 1.0 124/88/-178 034/88/-002 349/03 079/00 cluster 5 5 2006/12/03 19:51 1.7 085/86/ 154 177/64/ 004 134/15 038/21 9 2006/12/04 05:17 1.5 085/86/ 154 177/64/ 004 134/15 038/21 10 2006/12/04 12:35 1.4 085/86/ 154 177/64/ 004 134/15 038/21 13 2006/12/04 21:20 1.3 085/86/ 154 177/64/ 004 134/15 038/21 16 2006/12/05 03:55 1.1 085/86/ 154 177/64/ 004 134/15 038/21 18 2006/12/05 11:25 1.4 085/86/ 154 177/64/ 004 134/15 038/21 cluster 39 19 2006/12/05 18:56 1.6 149/43/-064 296/52/-112 146/72 041/05 27 2006/12/06 05:35 1.7 151/47/-062 293/50/-117 136/70 042/01 32 2006/12/06 14:34 1.1 152/45/-063 296/51/-114 142/71 043/03 36 2006/12/06 19:50 1.4 152/45/-062 295/51/-115 142/70 043/03 39 2006/12/06 22:27 2.2 154/42/-062 298/54/-113 152/71 044/06 40 2006/12/06 23:19 1.6 152/45/-063 297/51/-114 143/71 044/03 49 2006/12/07 06:04 1.7 152/45/-063 297/51/-114 143/71 044/03 55 2006/12/07 11:50 1.4 152/45/-063 296/51/-114 142/71 043/03 72 2006/12/08 00:41 1.0 153/47/-067 301/48/-113 138/73 047/00 78 2006/12/08 01:29 1.2 152/45/-063 296/51/-114 142/71 043/03 cluster 43 12 2006/12/04 19:54 1.1 165/46/-056 301/53/-120 151/66 052/04 21 2006/12/05 21:50 1.0 165/46/-056 301/53/-120 151/66 052/04 22 2006/12/05 23:06 1.2 165/46/-056 301/53/-120 151/66 052/04 25 2006/12/06 04:46 1.0 167/52/-066 311/44/-118 138/71 240/04 28 2006/12/06 11:09 1.0 167/49/-056 301/51/-123 146/65 054/01 31 2006/12/06 13:04 1.5 167/46/-070 319/47/-110 156/76 063/01 37 2006/12/06 21:12 1.2 167/47/-054 300/54/-122 150/64 052/04 42 2006/12/07 01:43 0.8 167/49/-056 301/51/-123 146/65 054/01 43 2006/12/07 01:44 1.9 167/52/-066 311/44/-118 138/71 240/04 52 2006/12/07 09:40 1.5 167/46/-069 318/48/-110 156/75 062/01 54 2006/12/07 11:38 1.1 165/46/-064 310/50/-114 153/72 057/02 63 2006/12/07 19:02 1.4 167/47/-071 320/46/-109 152/76 244/00 66 2006/12/07 21:16 1.6 167/46/-071 321/47/-109 156/76 064/01 cluster 82 44 2006/12/07 02:30 0.9 116/72/-152 017/63/-020 339/32 245/06 46 2006/12/07 03:55 1.3 116/72/-152 017/63/-020 338/32 245/06 58 2006/12/07 16:45 1.1 116/72/-152 017/63/-020 338/32 245/06

35

59 2006/12/07 17:27 1.7 115/72/-156 017/67/-020 337/30 245/03 69 2006/12/07 22:41 1.2 115/72/-152 016/63/-020 337/32 244/06 82 2006/12/08 01:50 1.9 116/72/-152 017/63/-020 338/32 245/06 90 2006/12/08 05:32 1.5 105/79/-167 012/77/-011 329/17 239/01 94 2006/12/08 09:04 2.2 111/74/-165 017/76/-017 334/22 064/01 108 2006/12/08 16:49 3.4 105/77/-165 012/75/-013 329/20 238/02 113 2006/12/08 20:20 2.5 107/75/-153 009/64/-017 331/30 236/07 cluster 86 60 2006/12/07 17:48 1.3 088/78/ 151 185/62/ 014 139/11 043/29 67 2006/12/07 21:28 1.5 088/78/ 151 185/62/ 014 139/11 043/29 71 2006/12/07 23:07 1.5 088/78/ 150 185/61/ 014 139/11 043/30 86 2006/12/08 03:06 2.6 088/75/ 151 186/62/ 017 139/09 044/31 100 2006/12/08 13:20 1.3 088/75/ 170 181/80/ 015 314/04 045/18 104 2006/12/08 15:30 2.1 088/78/ 150 185/61/ 014 139/11 043/30 147 2006/12/10 06:10 2.0 088/77/ 158 183/69/ 014 137/06 044/25 cluster 87 77 2006/12/08 01:22 1.3 161/88/ 010 071/80/ 178 296/06 027/08 85 2006/12/08 02:37 1.6 161/87/ 009 071/81/ 177 296/04 027/08 87 2006/12/08 03:24 2.3 161/87/ 009 071/81/ 177 296/04 027/08 111 2006/12/08 19:25 1.3 162/88/ 010 072/80/ 178 297/06 028/08 120 2006/12/09 01:58 0.8 162/87/ 009 072/81/ 177 297/04 028/08 cluster 88 88 2006/12/08 03:44 1.6 359/85/ 014 268/76/ 175 133/06 224/13 97 2006/12/08 11:03 1.6 359/85/ 014 268/76/ 175 133/06 224/13 118 2006/12/08 23:12 1.4 359/85/ 014 268/76/ 175 133/06 224/13 134 2006/12/09 10:52 1.2 359/85/ 014 268/76/ 175 133/06 224/13 140 2006/12/09 23:03 1.1 359/85/ 014 268/76/ 175 133/06 224/13 146 2006/12/10 05:40 1.3 359/85/ 014 268/76/ 175 133/06 224/13 151 2006/12/10 10:45 1.6 359/85/ 014 268/76/ 175 133/06 224/13 163 2006/12/21 07:40 1.4 359/85/ 014 268/76/ 175 133/06 224/13 cluster 102 92 2006/12/08 06:37 1.1 333/70/-041 080/52/-154 290/43 030/11 95 2006/12/08 09:16 1.3 333/70/-041 080/52/-154 290/43 030/11 101 2006/12/08 13:30 1.7 333/70/-041 080/52/-154 290/43 030/11 102 2006/12/08 15:12 2.0 333/70/-041 080/52/-154 290/43 030/11 107 2006/12/08 16:29 1.2 325/70/-050 077/44/-150 278/49 027/15 125 2006/12/09 04:52 1.0 333/70/-041 080/52/-154 290/43 030/11 131 2006/12/09 09:39 0.8 325/70/-050 077/44/-150 278/49 027/15 cluster 135 135 2006/12/09 14:18 1.6 177/45/-020 282/76/-133 151/42 043/19 143 2006/12/10 04:23 0.9 178/46/-013 277/81/-135 148/37 040/22 149 2006/12/10 09:28 1.5 178/47/-019 282/76/-135 150/41 044/18 150 2006/12/10 09:48 1.3 178/46/-020 282/76/-134 151/42 044/18 152 2006/12/10 13:53 1.3 178/46/-020 282/76/-134 151/42 043/19 153 2006/12/10 15:23 1.6 182/49/-013 281/80/-138 150/36 045/20 154 2006/12/10 17:31 1.4 178/46/-020 282/76/-134 151/42 043/19 158 2006/12/13 11:49 1.0 178/47/-022 283/74/-135 151/43 045/17 -------------------------------------------------------------------------------------------------

36

Figure 2.2.1. Top: Seismic stations in Switzerland and southern Germany that supplied data used for the fault-plane solutions of the induced seismicity in Basel. Bottom: Seismic stations in Basel and surroundings, during the stimulation in December 2006 and for about six months thereafter. The darker shaded areas correspond to the city of Basel and surrounding towns, while wood- and farmland are the light grey and white patches. The epicenters of the induced seismicity and the Basel injection well are located immediately east of station SBAF and in between stations WEIL and OTTER.

37

Figure 2.2.2. Example of seismograms (borehole station MATTE-EHB) of a family of similar events (Cluster 14), aligned at the P-wave onset. The ordinate is numbered with the event number. Note the polarity reversal of events 8 and 15 (right-hand figure), in spite of the high similarity of the S-phase – this occurs at stations that lie close to a nodal line on the focal sphere and is symptomatic of small differences in the focal mechanism.

38

Figure 2.2.3. Example of a cross-correlation (Cluster 14, Station HALTI) for the P-phase (top) and the S-phase (bottom). The red trace is the master-event and the vertical dotted lines in the left panels mark the signal window used in the correlation.

39

Figure 2.2.4. Relative hypocenter locations and focal mechanisms for Cluster 14: top left, epicenter map; bottom left, depth cross-section trending NW-SE; bottom right, depth cross-section trending SW-NE; top right, fault-plane solutions for events 14 and 15. The focal mechanism of events 3 and 8 are similar to the one of event 15 and the other are identical to the mechanism of event 14. The size of the crosses corresponds to the relative location error (one standard deviation), and the numbers next to each cross is the event number. Note the polarity reversals at stations END, WL11 and MATTE, which constrain a different dip of the SW-NE striking nodal plane. The seismograms recorded at station MATTE are shown in Figure 2.2.2. The colored lines correspond to the traces of the corresponding nodal planes in the fault-plane solutions. Only the (red) NW-SE striking plane is common to both sub-clusters and thus, most likely, constitutes the fault that ruptured in these nine similar events.

40

Figure 2.2.5. Perspective plot of the hypocenter locations relative to the casing shoe for Cluster 14, viewed at 30 degrees from above and towards the NNW (azimuth 336 degrees). Each event is represented as a circular rupture patch with a size equal to the rupture dimension, estimated from the respective seismic moment and an assumed common stress drop of 3 MPa. The left panel shows the fault planes that form the common NW-SE striking plane, while the right panel shows the two SW-NE striking fault-planes of each sub-cluster. The vertical line denotes the location of the borehole (black, the cased section, and blue, the open hole); in the right panel it is located about 10 m behind the fault, while in the right panel it is located between the two sets of fault planes.

41

Figure 2.2.6. Perspective plot of the identified faults, viewed at 30 degrees from above and towards the NNW (azimuth 336 degrees). The black dots are the relative locations of Deichmann & Giardini (2009) for the 165 events recorded in December 2006 by both the borehole and the surface networks. The different colors differentiate between the different faults as follows: normal faults (red), oblique strike-slip/normal fault (magenta), N-S strike-slip (green), E-W strike-slip (blue), approximately NW-SE strike-slip (cyan), ML 3.4 mainshock cluster (black). The numbers next to the planes identify each cluster (Cluster 86 is represented by two planes, and the mainshock cluster by three planes corresponding to events 82, 108 and 113). The vertical line denotes the location of the borehole (black, the cased section, and blue, the open hole).

42

3 Modeling the probability of LME

3.1 Probability in space and time: parts from Goertz-Allmann and Wiemer (2012)

Several conditions must be met for significant (damaging) earthquakes to occur. There must be a fault large enough to allow significant slip, there must be stress present to cause the slip along the fault and this stress must exceed the strength of the fault. It is not clear today what is the upper possible magnitude limit (Mmax) for induced events in a comparatively shallow reservoir setting (< 5 km depth). However this is an important input parameter to probabilistic seismic hazard assessment.

Using a simple geomechanical model based on stochastic seeds as described by Goertz-Allmann and Wiemer (2012) we investigate the probability of an event exceeding a certain magnitude with respect to injection time and distance from the injection point (Figure 3.1). We evaluate the events for a- and b-values within specific time and distance bins. For each bin the probability p of a certain magnitude M event is defined as (Wiemer, 2000),

We choose time bins of 105 s, moving at 104 s intervals, and distance bins of 100 m, moving across distance in 10 m increments. In order to obtain more stable results we analyze the induced seismicity from 100 model runs and stack individual probabilities. This allows us to compute a standard deviation to each mean probability estimate. The probability of a large magnitude event is determined by two factors, i) the overall number of events per bin (i.e. the a-value), and ii) the b-value in this bin.

We find that the mean probability of a large magnitude event increases with time up to the shut-in time (upper row Figure 3.1). This can partially be explained by the steadily increasing number of induced events up to shut-in and therefore increasing a-values, which is reflected by increasing probabilities up to the shut-in for a constant input b-value model. A varying b-value causes a substantial increase of the mean probability for the time period right after shut-in. The probability of a large magnitude event also increases to much further distances from the injection point for a varying b-value compared to a constant b-value (bottom row Figure 3.1). The increased probability at later times and larger distances is especially prominent for larger magnitudes (Figure 3.1c). Note that the large difference in absolute probability before the shut-in time between the constant and the varying b-value model mostly depends on the choice of a constant b-value of 1.48. A larger constant b-value would bring the probability levels of the two models closer at early times. The result of an increased probability of larger magnitude events at later times just after the shut-in and at larger distances to the injection point is consistent with observations of large magnitude events during hydraulic stimulations in geothermal systems not only in Basel (Deichmann and Giardini, 2009), but also elsewhere (Charlety et al., 2007; Evans et al., 2012).

43

Figure 3.1: Probability of an event exceeding a magnitude a) M3, b) M4, and c) M5 to occur at a certain time (top row) and distance from the injection point (bottom row). Error bars show the standard deviation computed from 100 model runs. The dashed line marks the shut-in time in time, and the location of the largest observed Basel event in distance. The different colors denote the model were the b-value is linked to differential stress (white), and the model with a constant input b-value (grey).

3.2 Dependence on depth and crustal strength: parts from Goertz-Allmann and Wiemer (2012)

To test a more general applicability of the simple geomechanical model for the purpose of forward-modeling expected seismic responses to fluid injections, we now investigate the influence of some of the modeling parameters onto the resulting seismicity cloud and source parameters. First, we calculate one model with an injection at a shallower depth of only 2.5 km, with correspondingly smaller overall stress magnitudes (Haering et al., 2008) and with only half the pore pressure. In a second step, we vary the strength of the crust by adjusting σ1 and the coefficient of friction µ. Input parameters of the three additional models are shown in Table 3.1. We keep the scaling relation between differential stress and b-value, respectively stress drop, exactly the same as before, even though this dependence was obtained heuristically on the basis of the Basel observations. Varying differential stresses between the three different scenarios therefore lead to varying absolute values of average stress drop and b-value. Our main interest is the relative variation of the source parameters and probabilities with time and distance. These relative variations can be extracted reliably from the modeling runs, and some more general postulates about induced seismicity can be derived despite variations in absolute values. It is unclear at this point whether any scaling between differential stress and b-value or stress drop would be varying in different geologic situations. However, for the time being we consider it a realistic assumption since it explains the observation at Basel quite well. At shallow depth and for a weak crust, we observe overall much higher b-values above two (stars and diamonds, respectively, Figure 3.2) compared to the original model at 4.5 km depth (circles), for which we obtain a good fit to the observed Basel seismicity (squares). The strong crust model results in a much lower b-value (inverted triangles). The stronger the distance dependence of the b-value the worse the fit of the data to a constant average b-value. This is particularly evident for the strong crust model (inverted triangles). The distance dependence of the b-value is greatly reduced for the shallow and weak crust models due to the reduced range of differential stresses (Figure 3.3). Interestingly,

44

the stress drop does not show a strong difference in the distance dependence between the different models.

Figure 3.2: Gutenberg-Richter relations of one model run for different model set-ups compared to the observed Basel data (squares). The circles show the average crust model at 4.5 km depth. The diamonds show the weak crust model. The inverted triangles show the strong crust model, and the stars show the shallow crust model. See Table 3.1 for input parameters. The magnitude of completeness Mc is marked.

In a next step, we impose a depth-dependent gradient onto the background stress field for the average crust case and an injection source at 4.5 km depth. Following Haering et al. (2008) the background stress field changes as follows:

σh (z) = 10 MPa/km z (3)

σ1 (z) = 42 MPa/km z ± 10% (4)

σ3 (z) = 17 MPa/km z ± 10% (5)

Since we leave the relation between differential stress and b-value the same we would expect to observe a depth dependence of the b-value on top of the radial dependence from the injection point. Figure 3.4 shows that this is indeed observed. However, the depth dependence is a secondary effect to the radial dependence and therefore difficult to observe in practice. For the synthetic data, we need to correct for the radial dependence first before we see a variation of b-values from 1.4 near the top of the seismicity cloud to 1.2 for the deepest events (diamonds in Figure 3.4a). A similar, weak depth dependence is also imprinted onto the stress drops (Figure 3.4b). Figure 3.4c and d show the depth variation of b-value and stress drop of the observed Basel data. As expected, the large scatter in the data does not allow us to observe such rather subtle depth variations in reality. While at least the mean b-values per depth bin (white squares in Figure 3.4c) are not incompatible with a depth variation, the scatter is too large to reliably resolve a depth dependence with statistical significance. The observed stress drops show no indication of a depth dependence.

45

Figure 3.3: a) b-value, and b) stress drop versus distance from the injection point for various model set-ups. The symbols show the mean over constant log distance bins of the mean values from 100 model runs with respective standard deviations for the average crust model at 4.5 km depth (circle), the shallow crust model at 2.5 km depth (stars), the weak crust model (diamonds), and the strong crust model (inverted triangles).

46

Figure 3.4: a) b-value and b) smoothed stress drop versus focal depth for the modeled seismicity (average crust, injection at 4.5 km). Circles show the data (mean from 100 model runs with standard deviation) without correction for the distance to the injection point. Diamonds show the same data after correction for the radial distance dependence. c) Observed b-value versus focal depth for the Basel seismicity estimated by Bachmann et al. (2012). d) Observed stress drop versus focal depth for the Basel seismicity estimated by Goertz-Allmann et al. (2011). The bold squares in c) and d) show the mean value within 0.1~km depth bins with respective standard error from bootstrap resampling using 1000 realizations.

In summary, our results suggest that the probability of large events (M ≥4) is reduced if we either drill less deep, or drill into softer formations such as sediments, instead of granite (Figure 3.5a and b). In sedimentary hydrofrac stimulations for enhanced oil- or gas extraction, b-values above 2 have been reported (Maxwell et al., 2009; Wessels et al., 2011). For the strong crust model, the probability of a M ≥4 event is overall higher compared to the average crust model at 4.5 km depth (Figure 3.5c). Nevertheless, our heuristic finding of a relation between b-value and differential stress prevents us from drawing a general conclusion about the behavior of b-values and magnitude probabilities anywhere else than Basel on which we calibrated the modeling.

47

Figure 3.5: Probability of an event exceeding a magnitude M 4 for a) a shallow crust model at 2.5 km depth b) a weak crust model, and for c) a strong crust model to occur at a certain time (top row) and distance from the injection point (bottom row). Error bars show the standard deviation computed from 100 model runs. The dashed line marks the shut-in time in time, and the location of the largest observed Basel event in distance.

However, the common feature of most modeled scenarios is the observation of the highest probability of a large magnitude event after well shut-in and at larger distances to the injection point. To show this more quantitatively, we calculate normalized cumulative probabilities in time and radial distance to the injection point (Figure 3.6) for the different model set-ups. For each case we compare the cumulative probability ratios before and after the shut-in time as well as closer and further away from a 300 m distance to the injection point. Individual values are listed in Table 3.2. Note that the 300 m distance level is arbitrarily chosen to separate the induced seismicity into two equal groups. For example we find that our model predicts a 61.5% higher probability of a M ≥4 event after the shut-in time compared to before the shut-in time and a 53.2% higher probability of an event further away than 300 m if b-value is coupled to differential stress. A constant b-value always predicts higher probabilities before the shut-in time and closer than 300 m to the injection point compared to the model where b-value is linearly linked to differential stress. Whereas the distance dependence is reduced for the shallow crust model, we also find that the weak and strong crust models predict higher probabilities of a M ≥4 event after shut-in and at larger distances. Two important consequences arise from this observation: (i) it is insufficient to shut-in an injection operation upon observation of a threshold magnitude event (so-called traffic-light system), since the largest magnitude probability is yet to come after the shut-in, irrespective of the applied threshold magnitude. (ii) increased probabilities at larger distances from the well means that a much larger area around the injection well is potentially affected by increased seismic hazard.

48

Figure 3.6: Normalized cumulative probability of an event exceeding a certain magnitude versus time (a and c) and distance to the injection point (b and d). a) and b) Comparison of an average crust model at 4.5 km depth for varying b-value (black curves) and constant b-value (grey curves). c) and d) Comparison of an average crust at 4.5 km depth (solid black), shallow crust (dashed grey), weak crust (dotted grey), and strong crust model (solid grey). The vertical lines marks the shut-in time (a and c) or the 300 m distance (b and d).

Our study suggests that a-priori information about the stresses in the area of interest are important for some initial estimation of the expected seismic hazard. Careful pre- operation site characterization of the underground to image possible fault zones capable of larger-magnitude seismicity can impose additional constraints on the maximum expectable magnitude in or near the stimulation volume. Furthermore, the implementation of a near real-time probabilistic seismic hazard prediction model can be used to replace previously used traffic light systems. Our model can also be used to investigate the effect on the induced seismicity and the respective seismic hazard using varying stimulation approaches such as a different injection flow rate or a different wellhead pressure.

49

Table 3.1: Mean medium parameters.

Table 3.2: Comparison of normalized cumulative probabilities pcum of exceeding a certain magnitude event in time and distance to the injection point for various model set-ups.

3.3 Dependence on pumping parameters: Gischig et al. (2012)

50

We use the stochastic seed model by Goertz-Allmann and Wiemer (2012) to explore different injection strategies, i.e. different injection rates and volumes. To be able to control the flow rate and volume, we use a non-linear 2D flow model done in COMSOL, in which permeability is irreversibly increased by a factor of about 150 wherever pressure reaches ~8.5 MPa. The model thus mimics shear induced permeability enhancement during stimulation, and is able to produce a propagating sharp pressure front. Figure 3.7 shows the flow rate time series from the Basel well together with the measured well head pressure and modelled pressure at the injection point. Considering the fit to be sufficient for our purposes, we use the modelled pressure distribution to run 100 realizations of the stochastic seed model. Parameters of the stochastic model were adjusted such that the temporal evolution of cumulative seismicity is reproduced for magnitudes M > 1, 2, 3. The observed cumulative seismicity lies well within the 95% confidence interval derived from the model realizations

(Figure 3.7). We further use the flow and stochastic model calibrated for Basel to explore different injection strategies. Figure 3.8 shows 5 different injection strategies using only 60% of the injected volume in Basel. Scenario 1 uses the flow rate time series from Basel halted as soon as 60 % of the cumulative volume is reached. The resulting seismicity is less than observed at Basel (grey line), which points to scaling of seismicity with volume. Scenarios 2 – 5 are generic injection time series ranging from very ‘violent’ injection (~90 l/s during 1 day) to very ‘soft’ injection (about 10 l/s during 9 days). We observe that seismicity depends on the applied injection rate. At high injection rates the resulting pressure can also induced slip at less critically-stress fractures while at low injection rates only highly critically-stressed fractures start slipping. Note that the overall seismicity increases (a-value), while the b-value does not change dramatically.

Figure 3.7: Fit of well-head pressure and temporal behaviour of cum.

51

Figure 3.8: Example of different injection scenarios using 60% of the maximum volume injected at the Basel reservoir. Also shown is the real data recorded at Basel.

3.4 References

Bachmann, C., S. Wiemer, B. Goertz-Allmann, and J. Woessner (2012), Influence of pore pressure on the size distribution of induced earthquakes, submitted to Geophys. Res. Lett. Charlety, J., N. Cuenot, L. Dorbath, C. Dorbath, H. Haessler, and M. Frogneux (2007), Large earthquakes during hydraulic stimulations at the geothermal site of Soultz-sous-Forets, Int. J. Rock Mech. Min. Sciences, 1091–1105. Deichmann, N., and D. Giardini (2009), Earthquakes induced by the stimulation of an en- hanced geothermal system below Basel (Switzerland), Seismol. Res. Lett., 80. Evans, K., A. Zappone, T. Kraft, N. Deichmann, and F. Moia (2012), A survey of the induced seismic response to fluid injection in geothermal and CO2 reservoirs in Europe, Geothermics, 41, 30 – 54. Gischig, V., B.P. Goertz-Allmann, C.E. Bachmann, and S. Wiemer (2012), Effect of non-linear fluid pressure diffusion on modeling induced seismicity during reservoir stimulation, EGU 2012 Conference contribution Goertz-Allmann, B.P., A. Goertz, and S. Wiemer (2011), Stress drop variations of induced earthquakes at the Basel geothermal site, Geophys. Res. Lett., 38. Goertz-Allmann, B.P., and S. Wiemer (2012), Geomechanical modeling of induced seismicity

52

source parameters and implications for seismic hazard assessment, submitted to Geophysics Haering, M., U. Schanz, F. Ladner, and B. Dyer (2008), Characterisation of the Basel 1 enhanced geothermal system, Geothermics, 37, 469 – 495. Maxwell, S., M. Jones, R. Parker, S. Miong, S. Leaney, D. Dorval, D. D’Amico, J. Logel, E. Anderson, and K. Hammermaster (2009), Fault activation during hydraulic fracturing, SEG Expanded Abstracts, 1552 – 1556. Wessels, S., M. Kratz, and A. D. L. Pena (2011), Identifying fault activation during hydraulic stimulation in the barnett shale: source mechanisms, b values, and energy release analysis of microseismicity. SEG Expanded Abstracts, 1463 – 1467. Wiemer, S., (2000), Introducing probabilistic aftershock hazard mapping, Geophys. Res. Lett., 27, 3405–3408.

53

4 Coulomb stress changes at the Basel geothermal site: can the Coulomb model explain the induced seismicity in an EGS?

Flaminia Catalli1, Men-Andrin Meier1, Stefan Wiemer1.

1Swiss Seismological Service, Institute of Geophysics, ETH Zürich, Zürich, Switzerland

4.1 Abstract

We estimate Coulomb failure stress variations (ΔCFS) by considering interactions among 118 earthquakes occurred during the hydraulic stimulation of the enhanced geothermal system (EGS) in Basel, Switzerland. In spite of the unquestioned role of the pore pressure perturbation as primary source of triggering, we observe that the role of stress changes to forecast subsequent seismicity is important: over 70% of event locations are consistent with positive ΔCFS. On the other hand, we show that the performance of the model, expressed in terms of Coulomb index, CI (fraction of events occurred on positive ΔCFS), is dependent on some parameters and their uncertainties. We analyze the model sensitivity to geometry (strike, dip and rake angles) and location of fault planes, and to some other critical parameters. We conclude that the Coulomb model can contribute for seismic hazard assessment during EGS stimulation.

4.2 Introduction

Between December 2 and 8, 2006 approximately 11,500 m3 of water were injected at high pressure into a 5 km-deep well in Basel, Switzerland [Häring et al., 2008]. The maximum wellhead pressure was 30 MPa, at a maximum flow rate of 50 l/s. This represents a typical “hot-fractured-rock” experiment where the crystalline rock is hydraulically stimulated for producing heat and power [Smith, 1983; Tenzer, 2001]. Such a stimulation produces fractures in the rock, which define the paths where the fluid flows through and heats up. During this process, more than 10,500 earthquakes were induced close to the injection point in the area of Basel, of which more than 3500 were located [Dyer et al., 2010]. Most of the seismicity occurred during the water injection that was reduced and then stopped after the occurrence of the ML 2.7 event of December 8 [Häring et al., 2008]. However, an event ML 3.4 occurred five hours later the same day. The overall time behaviour of the seismicity followed the flow-rate and wellhead pressure trend, with a gradual increase in seismicity during the injection (both in rate and magnitudes) and a rapidly decrease over the three weeks after the bleed-off [Bachmann et al., 2011]. However, three additional events with ML>3 occurred one to two months later (Table 4.1) and sporadic microseismicity was being detected even more than two years later. The bulk of seismicity is located on a near-vertical elliptic-shaped structure that strikes NNW-SSE with a maximum radial distance from the casing shoe of about 900 m at a depth around 4.5 km. The triggering effect of the pore pressure change due to the fluid injection is clear: an increase of the pore pressure reduces the effective normal stress and yields to sliding along pre-existing, favourably oriented subcritical ruptures [Rutledge et al.,

54

2004]. The higher the pore pressure, the less critical stress state can trigger an event. Some studies [Shapiro et al., 2007; 2009; 2010] show the link between the rate, the total number and the magnitudes of expected induced events to the rate of fluid injection and the volume of the stimulated area. Goertzt-Allmann and Wiemer [in press] and Bachmann et al., [in press] present and apply a simple geomechanical model of induced seismicity that is able to reproduce the source parameter variations (b-value and stress-drop) to the first order and to predict the behaviour of the induced seismicity in terms of magnitude and time. So far it seems to be left any capability in forecasting volumes of higher probability of occurrence in an EGS. The Coulomb stress model generally gives such a kind of spatial information. Coulomb stress changes associated to earthquake interactions have been commonly seen as a powerful tool for forecasting subsequent seismicity in a tectonic regime [Steacy et al., 2005 and references therein]. The concept at the basis of the Coulomb model is simple but effective: positive stress changes favour the generation process of next events and vice versa. The study of stress redistribution is also important to quantify the interdependence of earthquakes and to figure out how largely the stress diffusion mechanism can influence next events, also in an EGS. In a case like Basel, for example, stress variations can explain the presence of seismicity further in time and space from the injection, where the amplitudes of pore pressure changes are comparable or even lower than those of ΔCFS. The Coulomb model was used to study the generation process of mining induced seismicity [Orlecka-Sikora, 2010, and references therein] finding positive correlations between locations of events and positive stress variations (62% of the studied events occurred at locations of positive changes of stress, 47% in areas where ΔCFS>0.05 bar). They concluded that the static stress transfer can accelerate the mining induced seismicity generation process because even a small ΔCFS can have a significant effect on faults already loaded by mining stress. Any similar study has been conducted yet for an EGS. In this study, we estimate cumulative ΔCFS due to all previous events at each event location of the studied dataset and calculate the relative Coulomb Index (CI), i.e. the ratio between the number of positive ΔCFS and the total number of events. We perform a sensitivity analysis of the model to the nodal plane (NP) ambiguity, focal mechanism (FM) solutions and hypocentral locations and to some other parameters.

4.3 Dataset and methodology

We use fault geometry information (strike, dip and rake angles), MW magnitudes and hypocenter locations of 118 events that occurred from December 3 2006 to November 30 2007 in the EGS close to the city of Basel, Switzerland, to calculate ΔCFS through the Coulomb model [Steacy et al., 2005, and references therein]. These events have well-constrained FM (105 strike-slip, 5 oblique strike-slip and 12 normal faulting mechanisms) [Deichmann and Ernst, 2009] with relocated hypocenters [Deichmann and Giardini, 2009] and magnitudes ranging from ML 0.7 to 3.4 [Bethmann et al., 2011 and references therein]. We suppose all the faults to be extended and we estimate dimensions and mean slip of extended rectangular slip patches by using the relations of Hanks and Kanamori [1979] and Keylis-Borok [1959]. We use a fixed stress-drop of 2.3 MPa on the faults.

55

The stress field is mathematically represented by a tensor; it means that its values in the three directions x, y and z should refer to a specific plane before being visualized in a map. While computing ΔCFS is then important to identify the receiver fault planes (in terms of strike, dip and rake angles together with their hypocentral locations) for projecting results correctly [Catalli and Chan, 2012; and references therein]. This identification process is not always straightforward. In Basel, the naturally occurred events (i.e. the background events) show a preferred FM with mean strike of 155 (and its auxiliary plane) [Figure 10 in Deichmann and Giardini, 2009]. On the other hand, the induced seismicity shows a preferred strike

orientation of the FM nodal planes that is mainly NS-EW. The ambiguity between the two possible NP is then particularly marked in Basel: in principle, both directions NS and EW could fit with the regional FM direction. It makes especially difficult to speculate on the actual receiver fault planes for ΔCFS calculations. In this study, we assume the strike direction NS as preferential reference direction for the FM in the study volume and we call “preferential” the catalogue where we selected each NP as the closest to that direction. Viceversa, the “auxiliary” catalogue is the catalogue where the NP are the closest to the FM with strike oriented towards EW. We calculate ΔCFS using both catalogues (“preferential” and “auxiliary”, see Figure 4.1) and we find a substantial increase of CI by using the “auxiliary” catalogue. However, while studying the sensitivity of the model through CI distribution analysis (Figure 4.3), we use a random selection of the NP over each distribution to take the uncertainty of the NP into account. In Table 4.2 we report a summary of the parameters used in our computations. We have estimated cumulative ΔCFS at each hypocenter location due to all previous events by using a modified version of the code developed by Wang et al., [2006]. A cross-validation of results was made by using the code of Nostro et al., [1997]. Note that maps of ΔCFS showed in Figure 4.1 contain also information on the depth and have not to be intended as simple planar maps. Each point represents a specific triplet of x-, y- and z- coordinates for each event location. We evaluate the significance of our results by considering uncertainties related to our knowledge of NP, FM and hypocentral locations of the events. For this purpose, we perturb the original catalogue by i) randomly selecting one of the two NP; ii) adding a von Mises distributed error within 10 to strike, dip and rake for each original FM; iii) adding a normal distributed error around respectively 50 and 70 m for x-,y- and z- directions to the hypocenter locations. Each distribution refers to 10,000 perturbed catalogues. The CI distributions are reported in Figure 4.3. We finally analyze the model sensitivity to the friction coefficient parameter, on our computations we used =0.8, which is a reasonable value considering the type or rock in the Basel volume, where fractures are filled of clay [Sikaneta, personal communication]. However, we take into account for the observation of King and Cocco [2000] about the apparent friction coefficient used for ΔCFS computations, ’=(1-B) (where B is the Skempton coefficient, here fixed at 0.5); they affirm that ’ can vary in the range 0.0-0.8. For this reason we let vary within the range 0-2. In Figure 4.4 we show the trend of the CI considering ranging within 0 and 2. Each point in Figure 4.4 represents a mean value of the CI calculated on 3000 catalogues where the NP were selected randomly.

4.4 Coulomb stress changes and Coulomb indexes: discussion of results

Figure 4.1 shows maps of cumulative ΔCFS calculated at each hypocentral location of the 118 studied events under different assumptions: panel a, by using the “preferential” catalogue; panel b, by using the “auxiliary” catalogue (in both panels a and b all the possible event interactions have been considered, i.e. 6903); panel c and d follow the same criteria but for these two computations we limit the number of event interactions introducing a minimum inter-event distance (minD). By introducing a minD we pay the tribute to our limited knowledge of the source slip models: by excluding as possible receivers all the faults too close to a source, we limit the probability to produce a false alarm (in fact, the largest modelling errors are expected in the area around a source). We fix the minD at 1 source length, reducing the number of interactions to 6813. In all the four cases presented in Figure 4.1 the CI values show a positive and in some cases high correlation between event locations

56

and ΔCFS. CI are substantially higher when we calculate ΔCFS on the “auxiliary” catalogue, moving toward over 70% of agreement between hypocentral locations and positive ΔCFS (panels b and d). By introducing the minD (panels c and d) the values of CI increase of about 10%, respectively. On the other hand, one can observe an overall smoothing effect on the ΔCFS amplitudes. However, the effective role of the Coulomb model as indicator of the occurrence of induced seismicity is unquestionable when looking at panels c and d. In the best case, the CI reaches a value close to 80% (panel d). ΔCFS values range between -1 and 1 MPa (when considering the minD). Differently, the pore pressure change reaches values of an order of magnitude larger close to the injection point (within about 100 m from the injection) and during the time of maximum flow rate. Under this premise, we can conclude that, close to the injection (in space and time), ΔCFS can facilitate the pore pressure-induced seismicity generation by accelerating the rupture of faults already loaded and optimally oriented. However, with a closer look to panels b and d, one can observe a SW cloud of positive ΔCFS at locations of events distant from the casing shoe more than 500 m. It is reasonable to think that at this distance from the injection point the pore pressure change is strongly reduced and even lower than 1 MPa. We can then imagine that the generation process of this group of events is dominated by event interactions. We can observe in Figure 4.2a that three of the four largest events (ML≥3.0, the stars) belong to this cloud of positive ΔCFS further from the casing shoe. Moreover, all the largest events (ML≥3.0) are located on highest ΔCFS. It is then straightforward concluding that the CFS model effectively contributes to the forecasting of induced events in an EGS; the lower the pore pressure changes, ΔP, the more effective the role of event interactions in terms of ΔCFS. We show the temporal behaviour of the cumulative ΔCFS observed in Figure 4.1d and 4.2a in Figure 4.2b. In Figure 4.2b is particularly clear the positive performance of the Coulomb model, especially after the 6-days of fluid injection and the shut down of the system. In the first 6-days of injection we observe a dense generation of induced seismicity, both on positive and negative ΔCFS; in this phase we can explain the presence of seismicity on stress shadows by considering that i) we have larger uncertainties for the larger number of smaller events, which characterize volumes of higher ΔP [Goertz-Allmann and Wiemer, in press Bachmann et al., in press]; and ii) under the highest ΔP (which reaches 30 MPa at the maximum injection rate) even faults not optimally oriented can easily slip. We then conclude that during the pumping process the pore pressure and the Coulomb stress act together for the generation of next induced seismicity, the first predominating on the second. But we observe that after the shut down, when the pore pressure decreases, the stress redistribution regains a primary role in forecasting next seismicity, confirming what already concluded in the previous paragraph. For all computations showed in Figures 4.1 and 4.2 we assumed rectangular source models and a fix stress drop (see Table 4.1). However, we also tried to use circular fault models [Keylis-Borok, 1959] with the indirect effect of increase the estimated source dimensions of a factor about 1.3 and thus decrease the mean expected slip on the sources. In terms of CI this can be translated as a decrease of values of about 10-20 %. Moreover, we tried to use spatially variable stress drops calculated by Goertz-Allmann et al. [2011] for Basel, finding that ΔCFS are not sensitive to stress drop variations in this area. To study the sensitivity of the model to the uncertainties of FM and hypocentral locations of the events, we analyze distributions of CI, Figure 4.3. In Figure 4.3a we show the effect of the FM strike, dip and rake angles uncertainties: when we perturb the strike, dip and rake angles of each FM of the catalogue by adding a von Mises distributed error within 10 (ceylon distribution) we observe that the perturbed distribution of CI does not depart from the original one (green distribution). We consider 10 a conservative range of errors for the FM estimated for the Basel sequence [Deichmann, personal communication]. In the extreme case of random FM (magenta distribution) the distribution shifts exactly around CI=50%. We conclude that

57

results of ΔCFS in Basel are robust within the error of 10 for FM. Positions of the red and blue lines (CI calculated respectively using the “preferential” and the “auxiliary” catalogues) confirm what previously observed in Figure 4.1: using the “auxiliary” catalogue makes CI higher. In Figure 4.3b we show the effect of the hypocentral location (x,y,z) uncertainties. For the high-precision relative

4.5 Conclusions and outlook

We observe in all our computations a positive correlation between locations of events and positive ΔCFS during the Basel sequence. The degree of correlation depends on some parameters: receiver fault geometry and locations; NP ambiguity; relative locations of sources/receivers; and physical source characteristics (�, B, source geometry model and slip); however, we can reasonably constrain most of the modelling parameters and our results appear robust to FM uncertainties. Values of CI, which represent the degree of correlation, come up to about 80% in the best case and are always well over 50%. Our main conclusion is that ΔCFS is an important key parameter for a comprehensive hazard assessment in the domain of man-induced seismicity. The CFS model can effectively contribute in identifying more optimally oriented ruptures and structures closest to failure. We distinguish its contribution in two ways: 1) closer in space and time to the injection, where ΔP is sensibly larger than ΔCFS, we conclude that the CFS model can indicate preferential areas of failure; 2) further, where ΔCFS and ΔP are comparable or where ΔCFS is even larger than ΔP, the CFS model regains its indisputable role as forecasting tool. The latter considerations lead also to the conclusion that during the stimulation of an EGS the generation process of induced seismicity is controlled by two main phenomena: the variation of the pore pressure, ΔP, and the event-interactions, ΔCFS. These two processes act in synergy, ΔP predominating close in space and time to stimulation and ΔCFS further away. Our results show the capability of the CFS model to forecast all the largest events occurred during the sequence. Moreover, three out of the four largest events occurred on positive ΔCFS areas at distances from the injection where we can suppose that the main triggering source is represented by event interactions (ΔP < ΔCFS). We observe that the interaction between events is stronger if we assume the orientation of FM closest to the EW-direction (the so called “auxiliary” catalogue). However, we cannot easily conclude that it indicates the actual planes of fracture; we have to consider that under the stress of remarkably high pore pressures, close to the injection, it is reasonable that even unfavourable planes can rupture. On the one hand results seem to be robust to FM and depth uncertainties, but on the other hand the model performance is sensitive to the epicentral errors: when perturbing the horizontal coordinates of locations CI decrease of a factor of 20%. We then conclude that, for using the CFS model in such a small stimulated volume, filled of a dense cloud of events, good knowledge of hypocentral locations is fundamental. We think that the CFS model can be an important ingredient in developing a new geomechanical model for hazard assessment in EGS. It is able to give information on the higher volumes of risk by considering the physical properties of the faults.

58

References

Bachmann C.E., S. Wiemer, J. Woessner, and S. Hainzl (2011a), Statistical analysis of the induced Basel 2006 earthquake sequence: Introducing a probability-based monitoring approach for Enhanced Geothermal Systems, Geophys. J. Int., doi: 10.1111/j.1365-246X.2011.05068.x. Bethmann, F., N. Deichmann, and P. Mai (2011), Scaling relations of local magnitude vs. moment magnitude for sequences of similar earthquakes in Switzerland, Bull. Seismol. Soc. Am., 101, 2, 515–534, 2011, doi: 10.1785/0120100179. Catalli, F., and C.-H. Chan (2012), New insights into the application of the Coulomb model in real-time, Geophys. J. Int., 188, 2, 583-599, doi: 10.1111/j.1365-246X.2011.05276.x. Deichmann, N., and J. Ernst (2009), Earthquake focal mechanisms of the induced seismicity in 2006 and 2007 below Basel (Switzerland), Swiss J. Geoscience, 102/3, 457-466, doi: 10.1007/s00015-009-1336-y. Deichmann, N., and D. Giardini (2009), Earthquakes induced by the stimulation of an enhanced geothermal system below Basel (Switzerland), Seismol. Res. Lett. 80/5, 784-798, doi:10.1785/gssrl.80.5.784. Dyer, B. C., U. Schanz, T. Spillmann, F. Ladner, and M. O. Häring (2010), Application of microseismic multiplet analysis to the basel geothermal reservoir stimulation events, Geophys. Prospect., 58, 791–807. Goertz-Allmann, B.P., and S. Wiemer (2012), Geomechanical modeling of induced seismicity source parameters and implications for seismic hazard assessment, in prep. for Geophysics. Hanks T.C. and H. Kanamori (1979), A moment-magnitude scale, J. Geophys. Res., 84, 2348-2350. Häring, M., U. Schanz, F. Ladner, and B. Dyer (2008), Characterisation of the Basel 1 enhanced geothermal system, Geothermics, 37, 469–495. Keilis-Borok, B. V. (1959), On estimation of the displacement in an earthquake source and of source dimensions, Annali Geofis., Rome 12, 205-214. King, G. C., and M. Cocco, (2000), Fault interaction by elastic stress changes: new clues from earthquake sequences, Adv. Geophys., 44, 1-38. Nostro, C., M. Cocco, and M. E. Belardinelli (1997), Static Stress Changes in Extensional Regimes: An Application to Southern Apennines (Italy), Bull. Seismol. Soc. Am., 87, 1, 234-248.Bulletin of the Seismological Society of America, Vol. 87, No. 1, pp. 234-248, February 1997 Orlecka-Sikora B. (2010), The role of static stress transfer in mining induced seismic events occurrence, a case study of the Rudna mine in the Legnica-Glogow Copper District in Poland, Geophys. J. Int., 182, 1087–1095, doi: 10.1111/j.1365-246X.2010.04672.x.

59

Rutledge, J.T., Phillips, W.S, and M.J. Mayherhofer (2004), Faulting induced by forced fluid injection and fluid flow forced by faulting: An interpretation of hydraulic-fracture microseismicity, Carthage Cotton Valley Gas Field, Texas, Bull. Seismol. Soc. Am. 94, 5, 1817–1830. Shapiro, S., Dinske, C., Langenbruch, C., and F. Wenzel (2010), Seismogenic index and magnitude probability of earthquakes induced during reservoir fluid stimulations, Leading Edge, 29, 3, 304–9. Shapiro, S. A. and C. Dinske (2009), Fluid-induced seismicity: Pressure diffusion and hydraulic fracturing, Geo- physical Prospecting, 57, 2, 301–310. Shapiro, S. A., Dinske, C., and J. Kummerow (2007), Probability of a given-magnitude earthquake induced by a fluid injection, Geophys. Res. Let., 34, 22. Smith, M. C. (1983), A history of hot dry rock geothermal energy systems, Journal of Volcanology and Geothermal Research, 15(1-3), 1–20, Geothermal Energy of Hot Dry Rock. Steacy, S., J. Gomberg, and M. Cocco (2005), Introduction to special section: stress transfer, earthquake triggering, and time-dependent seismic hazard, J. Geophys. Res., 110 (B05S01), doi:10.1029/2005JB003692. Tenzer, H. (2001), Development of hot dry rock technology, Geo-Heat Center Quarterly Bulletin, 22, 4. Rongjiang, W., F. L. Martin, and F. Roth (2006), PSGRN/PSCMP—a new code for calculating co- and post-seismic deformation, geoid and gravity changes based on the viscoelastic-gravitational dislocation theory, Computers & Geosciences, 32, (4), 527-541, doi:10.1016/j.cageo.2005.08.006.

60

ID  Date  lon 

(degrees) 

lat 

(degrees) 

depth

(km) 

ML  MW 

1  2006/12/08 16:49  7.5932  47.5840  4.7  3.4  3.0 

2  2007/01/06 07:20  7.5958  47.5822  4.2  3.1  2.7 

3  2007/01/16 00:09  7.5946  47.5819  4.1  3.2  2.65 

4  2007/02/02 03:54  7.5960  47.582  4.0  3.2  3.05 

Table 4.1: The largest events of the sequence. Hypocentral locations from Deichmann and Giardini [2009]; ML from the SED catalogue (http://hitseddb.ethz.ch:8080/ecos09/index.html); MW from the catalogue courtesy of Geopower Basel AG.

Description Value Dimensions

Fixed stress drop, 2.3 MPa

Friction coefficient, 0.8

Skempton ratio, B 0.5

Rigidity, 2.0e10 N/m2

Reference FM (strike, dip, rake)

0, 90, 0 degrees

FM uncertainty (strike,dip,rake)

10, 10, 10 degrees

Hypocenter location uncertainty (x,y,z)

50, 50, 70 m

Minimum inter-event distance, minD 0/1 source length

Table 4.2: Parameters used in our computations.

61

dc

ba

Figure 4.1: Maps of cumulative ΔCFS calculated at hypocentral locations (x,y,z) of the 118 studied events of the Basel sequence using parameters reported in Table 4.2 (uncertainties are not considered at this stage). Panel 1: ΔCFS are calculated using the “preferential” catalogue (see text) and minD=0 source length; Panel 2: ΔCFS are calculated using the “auxiliary” catalogue and minD=0 s.l.; Panel 3: ΔCFS are calculated using the “auxiliary” catalogue and minD=1 s.l.; Panel d: ΔCFS are calculated using the “auxiliary” catalogue (see text) and minD=1 s.l.. Red crosses indicate positive stress changes and blue circles negative stress changes. The location of the casing shoe is represented with a yellow triangle. CI shows the percentage of locations on positive ΔCFS.

62

63

Figure 4.2: Spatial and temporal behaviour of ΔCFS showed in Figure 1d. Panel a: reproduces only positive ΔCFS observed in Figure 1d with more details on the ΔCFS range of values (color coded, see legend). The largest events (ML≥3.0, see Table 4.1 for ID numbers) are represented with stars and a reference distance of 500 m from the injection point is represented with a blue line. Panel b: time behaviour; on the x-axis the time origin is set at the occurrence time of the first recorded event (12/03/2006). Magnitudes of ΔCFS are showed in the y-axis and colour coded.

7.59 7.592 7.594 7.596 7.598

47.58

47.584

47.588

47.592

CI = 76.5100 m

0<dCFS<0.010.01<dCFS<0.10.1<dCFS<1

casing shoe

ML>3.0

2

3 4

1

500 m

64

Figure 4.3: Distributions of CI. Panel a: the distribution of CI calculated by considering the original catalogue with random selected NP (green bars) is compared with the one calculated on a catalogue where strike, dip and rake are perturbed by adding a von Mises distributed error within 10 (ceylon bars); and the one calculated on a catalogue with random FM (magenta bars). Panel b: the distribution of CI calculated by considering the original catalogue with random selected NP (green bars) is compared respectively with the distribution of normally perturbed depths (z – blue bars), epicentral coordinates (x,y – ceylon bars) and all the components (x,y,z – black bars). In both panels, the red and blue lines represent respectively the CI values calculated by using the “preferential” and the “auxiliary” catalogues.

40 45 50 55 60 65 70 75 80 85 900

500

1000

1500

2000

2500

3000

CI

Nu

mb

er o

f c

ase

s

Perturbed x,y,zPerturbed x,yPerturbed zOriginal locations"Preferential" ctlg"Auxiliary" ctlg

30 40 50 60 70 80

3000

900

500

1000

1500

2000

2500

CI

Nu

mb

er o

f c

ase

s

Perturbed FM

Original FM

Random FM

"Preferential" ctlg

"Auxiliary" ctlg

65

Figure 4.4: Trend of calculated CI against different values of the friction coefficient, �, in the range 0-2 used for ΔCFS estimations. Each red point represents the mean value of the CI calculated on 3000 catalogues where the NP were selected randomly. Vertical bars represent the standard deviation.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

80

76

72

68

64

60

Friction coefficient

CI

5 High-resolution analysis of seismicity induced at Berlín geothermal field, El Salvador: Kwiatek et al. (2012)

5.1 Abstract

We investigate induced microseismic activity monitored at Berlín Geothermal Field (BGF), El Salvador, during a hydraulic stimulation. The site was monitored for a time period of 17 months using 13 3-component seismic stations located in shallow boreholes. Three stimulations were performed in the well TR8A with a maximum injection rate and well head pressure of 140 l/s and 130 bar, respectively. For the entire time period of our analysis, the acquisition system recorded 581 events with moment magnitudes ranging between -0.5 and 3.7. The initial seismic catalog provided by the operator has been substantially improved: 1) We re-picked P- and S-wave onsets and relocated the seismic events using the double-difference relocation algorithm based on cross-correlation derived differential arrival time data. Forward modeling was performed using a local 1D velocity model instead of homogeneous full-space. 2) We recalculated source parameters using the spectral fitting method and refined the results applying the spectral ratio method. We investigated the source parameters and spatial and temporal changes of the seismic activity based on the refined dataset and studied the correlation between seismic activity and production. The achieved hypocentral precision allowed resolving the spatiotemporal changes in seismic activity down to a scale of a few meters. Of the special interest is the largest event (MW3.7) and its nucleation process. This event occurred in the center of the BGF about two weeks after the termination of the second injection in TR8A and is interpreted to be related or even triggered by the shut-in of the wells. This characteristics is in accordance with the occurrence of induced “larger magnitude events” in a number of other geothermal sites.

5.2 Summary

The seismic activity observed during stimulation phases of TR8A well display clear spatial and temporal patterns that could be easily related to the amount of water injected into the well TR8A and other reinjection wells in the investigated area.

The Kaiser effect is observed both for stimulation of shallower part of the reservoir (stages I1 and I2) and for the deeper stimulation (stage I3). Generally, the seismicity emerges only after certain level of the injection rate (or well head pressure) is exceeded (about 70l/s and 90l/s for I1+I2 and I3, respectively) and the following stimulation episodes in the same interval requires increased injection rate level to re-activate the seismic activity.

During the stepwise increase of injection rates in the shallower part of the reservoir (especially for stimulation I1) we observe migration of seismicity outside of the initial point located close by open hole section of the well. The seismicity migrates outside of the initial point towards SE only when an increase in the injection rate is still observed. The migration stops when the well is finally shut-in. This behavior is/was observed in other geothermal reservoirs (e.g. Soultz-sous-Forets, Basel) but here the migration is visible in much smaller scale with total migration length not exceeding 150m.

66

The locations of seismic events observed during injections I1 and I2 display clear spatial alignment trending NW-SE. We attributed this aligned activity to the fault system inferred from the mise-a-la-masse and head-on techniques and already mapped surface fault traces. This is additionally supported by the calculated fault plane solutions of a few seismic events that occurred in this area. The derived focal mechanisms display strike-slip behavior with one fault plane aligned along the fault plane system inferred from the non-seismological studies. Therefore, the events detected during injections I1 and I2 propagate outside of injection point along the pre-existing fault plane until the shut-in. The generation mechanism possibly involves decrease in the effective normal stress due to increase in the fluid pressure on the pre-existing fault plane. The direction of propagation is also consistent with the direction of the maximum horizontal stress in the investigated area.

The last injection I3 into the deeper formations possibly represents the clustered activity in the direct proximity of injection area not attributed to any existing fault plane. As the injection rate is relatively low despite much stronger well head pressure during the third stimulation, it is likely that the fracture network could not develop distinctively and migrate outside of the injection point as in the injection I1+I2.

After the shut-in of the well during injection I1 we observe firstly a few seismic events relatively far away from the fault, likely at the rim of pressure gradient zone induced by the injection that is ellipsoidal in shape with the longest axis trending NW-SE. After that, we observe an outburst of the seismic activity back on the fault plane F1, but in the opposite direction NW from the initial injection point. This area did not display seismic activity during the stimulation. Therefore, we conclude larger magnitude events observed in this area were triggered by imbalance in the shear stress distribution on the fault plane F1 caused by previous fluid injection that induced the seismicity in SE part from injection point. The area where the larger magnitude events occurred may correspond to the previously inactive part of the fault.

We observe increase in the static stress drop values with the distance from injection point for events that migrate along the fault plane F1. This supports observations performed by Goertz-Allmann et al. (2001) in deep heat mining project in Basel, where similar relation was observed and attributed to the pore pressure perturbations. As already suggested by Goertz-Allmann et al. (2001), this suggest static stress drop may be a proxy for monitoring pore pressure distribution if elastic parameters of the stimulated volume of rocks is know reasonably well.

The refined results do not provide any evidence that the multiple stimulations performed in the well TR8A triggered the LME of moment magnitude MW3.6 recorded 2 weeks after the second injection into TR8A. The seismic activity induced by multiple injections into TR8A did not reach distances larger than 500m from the injection point and we do not observe any sequence of events migrating further away towards the nucleation area of MW3.6. However, the injection operations at TR8A supposedly strongly disturbed the distribution of pore pressure at depth as well as influence the fluid migration pattern. As the fluids reinjected into multiple wells including TR8A comes from the same pool, it is still likely the TR8A stimulation could be indirectly responsible for the occurrence of

67

LME. This requires however the detailed analysis of the response of reservoir to fluid (re)injection(s) in the remaining part of the reservoir.

Low overall seismicity level due to highly permeable zones and weak strength of rocks.

68

6 A survey of the induced seismic responses to fluid injection in geothermal and CO2 reservoirs in Europe: Evans et al. (2012)

6.1 Abstract

The paper documents 41 European case histories that describe the seismogenic response of crystalline and sedimentary rocks to fluid injection. It is part of an on-going study to identify factors that have a bearing on the seismic hazard associated with fluid injection. The data generally support the view that injection in sedimentary rocks tends to be less seismogenic than in crystalline rocks. In both cases, the presence of faults near the wells that allow pressures to penetrate significant distances vertically and laterally can be expected to increase the risk of producing felt events. All cases of injection into crystalline rocks produce seismic events, albeit usually of non-damaging magnitudes, and all crystalline rock masses were found to be critically stressed, regardless of the strength of their seismogenic responses to injection. Thus, these data suggest that criticality of stress, whilst a necessary condition for producing earthquakes that would disturb (or be felt by) the local population, is not a sufficient condition. The data considered here are not fully consistent with the concept that injection into deeper crystalline formations tends to produce larger magnitude events. The data are too few to evaluate the combined effect of depth and injected fluid volume on the size of the largest events. Injection at sites with low natural seismicity, defined by the expectation that the local peak ground acceleration has less than a 10% chance of exceeding 0.07 g in 50 years, has not produced felt events. Although the database is limited, this suggests that low natural seismicity, corresponding to hazard levels at or below 0.07 g, may be a useful indicator of a low propensity for fluid injection to produce felt or damaging events. However, higher values do not necessarily imply a high propensity. We document the seismogenic response of crystalline and sedimentary rocks to fluid injection at 41 European sites. The objective is to identify factors that have a bearing on the magnitude of the largest seismic event. All cases of injection into crystalline rocks produce seismic events, albeit usually of non-damaging magnitudes. Injection at sites with low natural seismicity, defined by the expectation that the local peak ground acceleration has less than a 10% chance of exceeding 0.07 g in 50 years, has not produced felt events. The limited data suggest that low natural seismicity, corresponding to hazard levels at or below 0.07 g, may be a useful indicator of a low propensity for fluid injection to produce felt or damaging events. However, higher values do not necessarily imply a high propensity.

69

70

7 Appendix

7.1 Draft manuscript by Kwiatek et al. (2012)

7.2 Manuscript by Goertz-Allmann and Wiemer (2012)

7.3 Article by Evans et al. (2012)

High-resolution analysis of seismicity induced at Berlí n geothermal field, El Salvador

G. Kwiatek1, F. Bulut1, M. Bohnhoff1, G. Dresen1, S. Oates2

1. GFZ German Research Centre for Geosciences, Section 3.2: Geomechanics and Rheology, Telegrafenberg, Potsdam, Germany 2. Shell

Abstract

We investigate induced microseismic activity monitored at Berlín Geothermal Field (BGF), El

Salvador, during a hydraulic stimulation. The site was monitored for a time period of 17 months using

13 3-component seismic stations located in shallow boreholes. Three stimulations were performed in

the well TR8A with a maximum injection rate and well head pressure of 140 l/s and 130 bar,

respectively. For the entire time period of our analysis, the acquisition system recorded 581 events

with moment magnitudes ranging between -0.5 and 3.7. The initial seismic catalog provided by the

operator has been substantially improved: 1) We re-picked P- and S-wave onsets and relocated the

seismic events using the double-difference relocation algorithm based on cross-correlation derived

differential arrival time data. Forward modeling was performed using a local 1D velocity model

instead of homogeneous full-space. 2) We recalculated source parameters using the spectral fitting

method and refined the results applying the spectral ratio method. We investigated the source

parameters and spatial and temporal changes of the seismic activity based on the refined dataset

and studied the correlation between seismic activity and production. The achieved hypocentral

precision allowed resolving the spatiotemporal changes in seismic activity down to a scale of a few

meters. Of the special interest is the largest event (MW3.7) and its nucleation process. This event

occurred in the center of the BGF about two weeks after the termination of the second injection in

TR8A and is interpreted to be related or even triggered by the shut-in of the wells. This

characteristics is in accordance with the occurrence of induced “larger magnitude events” in a

number of other geothermal sites.

Keywords: fluid-induced seismicity; source parameters; enhanced geothermal systems;

1. Introduction Throughout many years the term “induced seismicity” (IS) was typically attributed to the seismicity

related to either impoundments of reservoirs (e.g. Gupta, 1992), human mining activity (Gibowicz

and Kijko, 1994) or fluid extraction in the oil and gas industry (Segall et al., 1994). Many studies

confirmed IS may lead to the occurrence of strong earthquakes reaching sometimes magnitudes

above 5 (e.g. Gupta, 1992, Pechmann et al., 1995). In recent years the fluid-induced seismicity caused

by massive fluid injections performed during Enhanced Geothermal Systems (EGS) projects become

considered as potentially hazardous for engineering structures below and above the ground as well

as the local communities. However EGS projects expands the available heat resources significantly

(e.g. Majer, 2007), at the same time they generate microseismic events. Events with local

magnitudes ML exceeding 2.9 were observed in a number of EGS systems (Baisch et al., 2010):

Soultz-sous-Forêts (Franc), Cooper Basin (Australia), Basel (Switzerland), and Berlín (El Salvador) to

name a few. Some of them (Basel, Soultz-sous-Forêts) are located close by local communities.

Typically, the events related to EGS are too weak to be felt by the local population, or the EGS

systems is simply located far enough from local communities to raise any concern regarding the

seismic hazard. However, in recent years an extreme case of fluid-induced seismicity may be the

ML 3.4 seismic event which occurred on the December 8th, 2008, five days after the beginning of

injection into the borehole located within the city of Basel (e.g. Deichmann and Giardini, 2009). The

ML 3.4 event followed by another three events with ML reaching 3.0 and numerous aftershock

activity resulted in the close-up of the Basel EGS project.

Numerous studies have been performed to understand the response of geothermal

reservoirs to fluid injections, as well as to understand the nucleation process that leads to the

occurrence of Larger Magnitude Events (LME) in EGS. The overall challenge is to mitigate the hazard

related with fluid-induced seismicity and occurrence of LMEs to an acceptable level. The basic

concepts of the generation process of fluid-induced seismicity, occurrence of LME and evolution of

the seismicity and its physical characteristics in time and space are not very different from other

types of induced seismicity as well as natural earthquakes:

1. The seismic activity, as well as the large and damaging earthquakes tends to occurs on the

pre-existing or developed faults systems. This is because less strain energy is required to

trigger slip on pre-existing fault surface than to create a new one. For example Kwiatek at al.

(2010) attributed extremely small fluid-induced seismic events (MW -1.9 to -1.1) recorded at

Gross Schoenebeck geothermal site (Germany) to existing fault plane structure inferred from

active seismic profiling and supported by slip-tendency analysis (Moeck et al., 2009).

Bohnhoff et al. (2004) analysed the spatial and temporal distribution as well as fault plane

solutions of seismicity induced at KTB site. They related clusters of microseismicity to major

faults located in the vicinity of injection area. Fabriol et al., (1998) suggested that seismic

activity observed at Berlin Geothermal Field, El Salvador aligns along the faults located

nearby injection wells. Also, the activity related to ML 3.4 seismic event Basel event

(Deichmann and Giardini, 2009) is suggested to occur on the fault plane. It is worth to note,

however, that in most cases there is no independent (non-seismological) information on fault

planes in the vicinity of injection point and the fault planes are inferred from observations of

hypocenters and focal mechanisms rather than proofed by independent non-seismological

data.

2. Fluid injection is recognized as a triggering mechanism of fluid-induced earthquakes (e.g.

Majer et al., 2007). The dominant process is possibly related to the effective stress reduction

due to increased pore pressure, however other mechanisms including temperature decrease,

volume changes due to fluid withdrawal/injection, chemical alteration of fracture surfaces

are discussed as well (Majer et al., 2007). The understanding of triggering mechanism is

crucial for understanding of the occurrence of seismicity and LMEs in geothermal reservoirs.

However it is a difficult task due to the limited number of data and lack of existing studies

that comprehensively describes the interactions between human operations and the

response of the geothermal reservoir. As suggested by Majer et al. (2007), the importance of

each mechanism must be discussed for each EGS project separately, taking into account

many non-seismological data including stress field, fault structures, rock mechanical

properties and hydrologic factors. Indeed, various EGS sites display quite different responses

to fluid injections and there is a strong demand on documenting the relations between

injection and seismic response to the injection. Recently, Evans et al. (2012) analyzed 41

geothermal projects over the Europe and concluded that stimulation into igneous rocks

seems to be more seismogenic than active stimulation of sedimentary rocks (but it does not

apply for EGS sites with sedimentary rocks and balanced circulation). Davis and Frohlich

(1993) suggested risk of inducing seismic events is likely to increase when injection takes

place near to or within fault zones. Also the high injection pressures seem to increase the risk

of producing larger events, however there is no clear relation between the maximum

magnitude recorded in the reservoir and injection pressure (Evans et al., 2012).

3. It was observed seismicity related to stimulation evolves in time and space in response to

injection(s) performed. This includes clustering of seismicity around injection well or

migration of seismic activity outside of injection point (Sasaki, 1998; Yamashita, 1999; Evans

et al., 2004; Michelet and Toksöz, 2007; Deichmann and Giardini, 2009; Kwiatek et. al., 2010;

Schoenball et al., 2010; Godano et al., 2012). These features are quite well documented,

however the resolution of spatial and temporal patterns may be still improved by using

state-of-the-art (re)location technique that surprisingly still are frequently not used on daily

basis in fluif-induced seismicity. Therefore, there is a strong demand to replace standard

localization methods with more efficient ones such as master-event method or double-

difference relocation.

4. Little is known on the detailed source characteristics of fluid-injection induced seismic

activity. Most frequently the hypocenters and (local) magnitude data is the only available

source characteristics. Whenever it is possible, the focal mechanisms provide and moment

tensor solution provides additional information on faulting geometry and faulting type,

including non-double-couple focal mechanisms (Ross et al., 1999; Rutledge et al., 2002,

Bohnhoff et al., 2004; Cuenot et al., 2005; Charlety et al., 2007; Vavrycuk et al., 2008; Fisher

and Guest, 2011). Recently, static stress drops was found to correlate with pore pressure

variations (Goertz-Allmann et al., 2011) for seismic events recorded in deep heat mining

project in Basel, Switzerland. Surprisingly, little if nothing is known on source characteristics

related to damaging potential of earthquakes regardless of the scale, such as apparent stress

or the rupture velocity.

The purpose of this study is to evaluate the systematic relations between the seismic activity and

fluid injection for the multiple injections experiment performed in Berlín geothermal field, El

Salvador. We focus on an intensive refinement of the original industrial seismic catalogue by state of

the art techniques that allow observing details of spatial migration of the seismicity and revealing

even small changes in physical properties of seismic sources. We aim on correlating the

spatiotemporal and physical characteristics of fluid-induced seismic events to injection operations,

production parameters and non-seismological data including stress measurements and tectonic data.

We discuss scenario leading to the occurrence of larger magnitude events in investigated geothermal

site.

2. Overview Berlín Hot Fractured Rock (HFR) site is one of six geothermal fields in El Salvador (Bommer et al.,

2006) (Fig. FAA). It is located at the northern flank of the Berlín-Tecapa Quaternary volcanic complex

with the nearest Cerro Tecapa volcano (last eruption 1878). The area is characterized by basaltic to

andesitic lava flows, lithic tuffs and scoria, and andesitic to dacitic ignimbrites (Torio-Henríquez,

2007).

Figure FAA. Location of Berlín Hot Fractured Rock (HFR) geothermal field.

The installed power plant produced 54 MWe (as of 2006) from 8 production wells with the

fluid exhausted at 183ºC and disposed via reinjection system comprised of 10 injection wells

(Bommer et all, 2006). Depths of the wells range from 700m to 2500m (Fabriol et al., 1998, Oates et

al., 2004). Fig. FBB presents the location of production and injection wells at Berlín HFR site together

with detected surface traces of faults.

Berlín HFR site is a region of high seismic activity. The main source of deep, large magnitude

events is the subduction of the Cocos plate beneath Caribbean plate with magnitudes exceeding

MW 7.5 and the upper crustal earthquakes are limited to MW 6.6 (Bommer et al., 2001, 2006). As the

high-resolution monitoring coincided with the beginning of extraction at the site it is not clear

whether the local and shallower seismic activity is the manifestation of the hydrothermal activity

around the volcano and/or seismicity triggered by ongoing man-made extraction and injection of

fluids (Bommer et al., 2006).

Figure FBB. The map view of production and injection wells (deep blue points lines). Seismic stations installed for passive seismic monitoring of injections into TR8A are shown using white triangles. The mapped traces of existing faults are shown in red.

3. Stimulations of TR8A Between 2003 and 2004 a sequence of three stimulations (denoted I1-I3) of the low-injectivity well

TR8A were performed (Oates et al., 2004). TR8A is located in the northern part of the main

production zone (Fig. FBB). The performed stimulations aimed on increasing the permeability around

the well. If successful, stimulation of TR8A was seen as a mean to extend the geothermal field further

to the North (Oates et al., 2004). Studies of the tectonic stress regime in the Berlín area suggested

that the fracture network would develop in a NNW–SSE orientation and intersect one of three wells

about 500m away from TR8A. The stimulated well would then become the injector in an HFR doublet

with the well intersected by the fractured region (Bommer et al., 2006). The high-frequency seismic

network was installed in order to monitor the evolution of the microseismicity and therefore

expansion of the reservoir.

Fig. FCC presents the time plot of the well head pressure and injection rate together with the

detected number of seismic events per day. First period of pumping (I1) between June 27th, 2003 and

July 19th, 2003 lasted approximately 21 days and aimed at shallower formations starting at 1750m

TD, about 1250 m below the sea level. The second injection (I2) started on August 17, 2003 and

lasted 18 days until September 6th, 2003. After the work-over of the injection well, the last

stimulation was performed (I3) between December 22nd, 2003 and January 10th, 2004 (18 days) and

focused on the deeper formations at depths of 2240-2248m TD (about 1700 m below the sea level).

Locations of injection intervals are shown in Fig. FBBb. The maximum well head pressure reached

about 13.3 MPa and the maximum injection rate equaled to 159 l/s.

Figure FCC. Variation of the injection rate [l/s] (black lines) and well head pressure [bar] (red lines) with time for three injection periods in the well TR8A. Gray bars display the number of seismic events detected per day during industrial monitoring of induced seismicity.

4. Data

Seismic network The seismic network used in the analysis was deployed in 2002 before planned injections into the

well TR8A. The objectives were to provide locations and magnitudes of microseismicity related to the

stimulation of the reservoir. The real-time analysis of recorded microseismicity aims to provide an

image of the stimulated area (hypocenters of seismic event) and allows controlling and adjusting the

hydraulic stimulation parameters if the activity reaches predefined thresholds (Oates et al., 2004) by

means of the traffic-light system (Bommer et al., 2006).

Six monitoring sites were selected to install 13 seismic stations. The locations of stations were

selected on the basis of location error modeling. The modeling provided an optimal spatial coverage

of expected seismicity with seismic stations (Oates et al., 2004). At 5 of the selected monitoring sites

TR1, TR12, TR14, Camp and Santa Anita new shallow boreholes were drilled. The last site was located

in the existing well TR11b. The spatial distribution of installed sensors is shown in Fig. FBB.

Each site was equipped with two 3-component geophone packages – shallow (few meters

below surface) and deep (54-98m below the surface). Additionally, the very deep borehole sensor at

well TR11b was installed 605m below the surface. The natural frequency of shallow sensors was

either 1 Hz or 4.5 Hz, whereas natural frequency of deeper sensors equaled to either 30Hz or 5.5Hz.

In all cases the sensors were grouted to the well walls. The acquisition system worked in triggering

mode. The data was stored with various sampling frequencies (fs) starting from fs =13Hz up to fs

=3000Hz (56% of waveforms were stored with fs >=1000Hz and above 90% of waveforms with fs

>=500Hz).

Catalog and LME The original industrial catalog contains 581 events recorded between October 2002 and

February 2004. During all three stimulation periods, the seismic acquisition system recorded 134

events. The original processing of the catalog incorporated auto-picking of P and S phases, manual

re-picking when necessary and re-locating of seismicity. Simple velocity model with VP=4200 m/s and

VS=2435 m/s was used (Oates et al., 2004) to locate seismic events. Fig. FDD displays the initial

catalog before and during the injections in the well TR8A.

Figure FDD. Original catalog of seismicity recorded between October 2002 and January 2004. Events that occurred during three stimulation phases of TR8A are shown as black circles (cf. Fig. FBB). The seismicity does not display clear clustering. The location of injection intervals are shown in side view.

On September 16th at 01:20, about 2 weeks after the termination of the second injection the

large magnitude event occurred (MW 3.7, ML 4.4, see Bommer et al., 2006, for an extended discussion

on the discrepancy between moment and local magnitude). The event occurred in the southern part

of Berlín HFR site, about 3 km from the injection point (Oates et al., 2004) (cf. Fig. FDD). The focal

mechanism was determined by independent seismic network and displayed east-west right-lateral

strike slip faulting (Bommer et al., 2006). The event caused no visible damage as the duration of the

peak ground velocity pulse was extraordinary short (Oates et al., 2004) and any structures in the

vicinity were subjected to the high load only for a very short period of time. The question whether

the event was triggered by injection operations in the well TR8A was discussed both in Oates et al.

(2004) and Bommer (2006). They argue LME occurred on the other side of Berlin HFR field, far away

from TR8A, and the pore pressure perturbations due to pumping operations at TR8A could not

efficiently reach the future source area. Moreover, there was no sequence of smaller events linked

TR8A zone with LME arguing against triggering by stress transfer (Oates et al., 2004). LME also

correlated well with an active fault identified in the seismic catalog (Oates et al., 2004).

5. Analysis

Relocation We use the Double-Difference earthquake location method to obtain the highest precision of spatial

offset between the earthquake hypocenters (Waldhauser and Ellsworth, 2000). Use of relative arrival

time data for closely-spaced events suppresses the effect of unmodeled velocity structure on

hypocentral offsets because the ray-paths of paired events are almost identical. The method also

allows the use of differential travel times, which can be measured much more precisely than arrival

time onsets, resulting in more accurate differential locations.

We used the cross-correlation technique to revise relative arrival times. Cross-correlation

derived time delays subtracted from time delays that are generated using manual phase picks. The

time window range is -10 to 240ms framing the manually picked P- and S-wave arrival times. The

cross-correlation coefficient threshold for data selection is previously investigated comparing cross-

correlation coefficient and the hypocentral-precision/data-misfit obtained from the corresponding

data set. The best average precision of hypocenters is obtained for the cross-correlation coefficient

range of 0.6-0.75. However, selecting the highest possible threshold (0.75) significantly reduces the

number of data used (Bulut et al., 2011). Therefore, the data vector is restricted to cross-correlation-

coefficients > 0.6. We adopt a 1-D reference velocity model that was optimized earlier for the target

area (Rivas et al., 2000). The uncertainties of initial locations were typically 150m in horizontal plane

and depth axis, respectively. Using the Double-Difference method lowered the average misfit from

64ms to 17ms resulting in improved precision of the relative hypocentral locations to 8.8m for the

horizontal and depth axis, respectively. Nevertheless, there is a trade-off in absolute locations

between the hypocentral accuracy and the velocity model. This is partly resolved in horizontal scale

since the stations are well distributed at surface. However, depth axis still remains relatively

unresolved since the depth aperture of the network is not that much advanced. One-sided control of

the hypocenters for depth axis introduces an inherent uncertainty into the event depths. Therefore,

we note that there might be significant vertical offsets from their true locations even though

covariance-derived hypocentral errors are negligibly small. Fig. FEE presents the spatial distribution

of 393 relocated seismic events out of initial 581 earthquakes.

Figure FEE. The seismic catalog relocated using DD algorithm . The lines display the relative shift in location between the original and double-difference-relocated catalogs. Color reflects different families of events determined with the cluster analysis (1-9). Events marked as circles occurred during three injection phases into TR8A.

Cluster analysis For 393 relocated events we performed the hierarchical cluster analysis. The measure of a ‘distance’

was the weighted sum of an actual Cartesian distance between two events i and j and the similarity

of their waveforms estimated as√ ⁄ ∑

, where is the cross-correlation coefficient

between S-phase waveforms of event i and j recorded at station k. We intended to investigate

whether the spatial clusters also display similar waveform characteristics. We arbitrarily set up the

cut-off for family selection based on the visual inspection of generated dendrogram. The cophenetic

correlation coefficient is 0.94 suggesting obtained clustering solution reflect the actual data

characteristics. The epicentral distribution of families is shown in Fig. FEE.

The performed cluster analysis divided the dataset into 9 families, however only 3 families

(F3, F1, F7) contain more than 15 events. The remaining families hold either shallow activity within

Berlín HFR site (F8, F2, F5) or clusters of deep events (F9, F8) probably related to the normal fault

system “Central Graben” located at the depth of approximately 1500m (cf. Fig. 12 of Rivas, 2000).

Finally, family F6 (not shown) carries a cluster of seismic events related to volcanic located ESE from

Berlín HFR site.

Source parameters Source parameters were calculated following the slightly modified spectral fitting procedure already

presented in Kwiatek et al., 2011 and Yamada et al., 2007. For the analysis only waveforms with a

signal-to-noise ratio of 20 dB or more were selected. The original 3-component waveforms were de-

trended and then filtered using 2Hz, 4th order Butterworth’s high-pass filter. Waveforms were

analyzed with a window length of 0.44 s including a 0.04 s period prior to either P or S wave onsets.

The windows were smoothed using von Hann’s taper (11% of the selected window). The ground

velocity spectra ( ) were calculated from three components of the sensor using the multitaper

method (Percival and Walden, 1993).

The ground velocity spectra were modeled using Boatwright’s source model (Boatwright,

1978):

( )

( ( )

) (

), (1)

where R is the source-receiver distance, M0 is the seismic moment, fC is the corner frequency, QC is

the quality factor and RC is the P- or S- wave average radiation pattern correction factor. Following

Boore and Boatwright (1984) we used =0.52 and =0.63 for P and S waves, respectively, as focal

mechanisms of recorded seismicity are unknown. VC stands for the P- or S- wave velocity in the

source area (typically =4200m/s and =2435m/s for most of seismic events, cf. Tab. 2 in Rivas et

al., 2000 for details). We assumed density =2430 kg/m3 (Oates et al., 2004).

We compared the logarithm of the modeled ground P and S velocity spectra (1) with the

observed spectra, ‖ ( ( )) ( ( ))‖ , and solved the optimization problem

using grid search technique with M0, FC and QC as unknown variables. The resulting parameters were

further optimized using simplex algorithm (Nelder and Mead, 1965; Lagarias et al., 1998).

We used ground velocity spectra corrected for attenuation to calculate the radiated energy EC from P

and S waves following Boatwright and Fletcher (1984):

, (2)

where JC stands for the energy flux (e.g. Snoke, 1987). The values of radiated energy were corrected

for the limited frequency band according to Ide and Beroza, (2001). The total energy was the sum

of energy radiated from P and S phases.

The review of the initial results clearly demonstrated that we are facing the serious inversion

stability problems related to the unknown attenuation in the investigated area (cf. Fig. FKK).

Although estimation of seismic moment seems to be consistent between different stations (as well

as to the values available from original industrial catalog), the source radius estimates suffer from the

trade-off between the corner frequency and generally unknown quality factor, especially for smaller

events. We therefore decided to suppress the propagation effects using modified spectral ratio

technique (e.g. Imanishi and Ellsworth, 2006) and refine the calculated source parameters.

In the first step for each family we selected a few events with reliably calculated source

parameters (i.e. characterized by a high number of stations used to calculate source parameters and

low dispersion of estimated source characteristics). These events served as “benchmarks” with fixed,

invariant source parameters. We assumed all events in the family can be described by a point source

using Boatwright’s source model (eq. 1) and that they display similar focal mechanism/complexity of

the rupture process. From the cluster analysis we know that each family is consistent in terms of

waveform similarity. Therefore it is feasible focal mechanisms and rupture processes within family

are similar as well. We also assumed that the family is a group of events located close by. Therefore,

any two events from the same family display a negligible difference in propagation effects from their

sources to the specific receiver. This is again secured by the distance criteria from the cluster

analysis. For any two events from the same family we could create a ratio of the ground velocity

spectra:

( ) ( )

( )

(

(

⁄ )

( ⁄ )

)

(3)

As a result, the propagation effects are effectively removed. In our case spectral ratio presented in

eq. 3 (called “link”) is created only for events meeting additional criteria: (a) inter-event distance not

larger than 200m and (b) cross correlation coefficient greater than 0.8. This is in order to secure the

similarity of mechanism, rupture process and colocation of two events forming spectral ratio. Links

may be created between benchmark event and non-benchmark event as well as between two non-

benchmark events. Eventually, any non-benchmark event that fits the criteria is connected to at least

one benchmark event using a limited number of links, i.e. it is either directly or indirectly connected

to some benchmark event(s). The advantage of the method is that it allows using many spectral

ratios between various events to produce stable estimates of source parameters. The shortcoming of

the procedure is that events with complex source rupture processes (especially if they are considered

benchmarks) may introduce an unintentional bias to other linked events. Fortunately, as suggested

by Shearer at al., (2004) this effect is suppressed and averaged over many linked events and weights

used in the following inversion procedure.

The inversion problem relies on minimizing the following cost function:

( ) ∑ ∑ ‖ ( )

( )‖ ( ) (4)

where summation goes through all accepted pairs (i,j) of spectral ratios recorded at stations k. The

additional weighting coefficient wij depends on the distance and cross-correlation coefficient

between events i and j. The inversion problem presented in eq. 4 was solved using Very Fast

Simulated Re-Annealing algorithm (e.g. Ingber, 1989, 1993). We applied this technique to 8 families.

We refined estimates of seismic moment and corner frequency of 331 events out of initial 393

relocated events. The remaining events were not processed mainly because of not fulfilling the

distance criteria and therefore lack of linkage to remaining events from the specific family.

After refinement, the source radius was calculated assuming circular source model of

Madariaga (1976). We calculated the static stress drop following Eshelby’s formula (Eshelby, 1959):

. We calculated moment magnitude using standard relation: W

log

(Hanks and Kanamori, 1979).

Figure FKK. The dependence between seismic moment (moment magnitude) and corner frequency (source radius) for the 331 relocated earthquakes calculated using spectral fitting (a) and refined with spectral ratio (b) method. Thin dashed lines present the lines of constant static stress drop assuming Madariaga source model. (a) The filled and empty circles correspond to source parameters calculated with N>=10 (high quality) and N<10 stations (low quality). (b) The color as well as the shape of points corresponds to the family number (cf. Fig. FEE). Thick dashed lines denote the decrease in the scatter of the stress drop thanks to the application of spectral ratio refinement.

Fig. FKK presents the comparison between the corner frequency (source radius) and the seismic

moment (moment magnitude) before and after the refinement of the source parameters using the

spectral ratio method. The original data before refinement display two features unusual for reliably

developed source parameters. First of all, we observe an extreme variation in the stress drop

reaching 3 orders of magnitude from 0.1MPa to 100MPa which is too high in comparison to the

typically observed scatter of 1-1.5 order of magnitude for a single dataset in a single dataset (cf.

Fig. 8 in Kwiatek et al., 2011). Secondly, we noticed the corner frequency of analyzed seismicity does

not exceeding 50Hz and for the same source radius we observe a great variety of the moment

magnitude. Overall, the results of spectral fitting procedure could be initially interpreted as a

breakdown in the self-similarity of earthquake rupture process reported in volcanic area (cf. Del

Pezzo et al., 2004, Galluzzo et al., 2009). However in our case the apparent breakdown is caused by

the trade-off between attenuation and the corner frequency as well as quite limited observable

range of frequencies. The application of spectral ratio technique resulted in an immediate

improvement in the source parameters, decrease in the scatter of the stress drop values and

suppression of dependency of the static stress drop on seismic moment.

6. Discussion Our primary focus is to investigate the interaction between the fluid injection and the

microseismicity. For the time period of our analysis, we have injection data only from injection well

TR8A. However, the injections were performed at the same time in other injections wells southward

from TR8A (Oates et al., 2004) in the central part of the field under the gravitational feed, with much

lower injection rates and well head pressures. The important fact is that the overall amount of fluids

injected in the whole area does not change i.e. the fluids reinjected into TR8A at high pressures and

injection rates are taken from the total volume of the fluids available for all reinjection wells in the

investigated area (Oates et al., 2004). Therefore, at the times of high injections into TR8A, the

reinjections into other wells were reduced. It seems that recorded seismicity clearly correspond to

the described ‘balancing’ injection scheme, at least for the periods of injection into TR8A. During

injection periods into the well TR8A nearly all seismic activity occurs close by injection well TR8A

(Fig. FEE, circles, family 1) with very weak seismic activity detected outside of this area at the same

time. After the shut-in of TR8A in injection phases I1 and I2 we observe the immediate outburst of

the seismicity in the central part of the geothermal site (Fig. FEE, circles, family 3), where remaining

injection wells are located. This is also visible in Fig. FHH which displays the temporal evolution of

seismicity together with injection rates and well head pressure. Therefore it seems that if the total

amount of reinjected water in the reservoir does not change, the spatial and temporal distribution of

the seismicity is directly related to area of high injection in phases I1 and I2. Interestingly, the shut-in

TR8A after I3 resulted in no increase in seismic activity in the central part of the reservoir,

contradicting the observations confirmed for injection phases I1 and I2. One reason may be the

injections were stopped in all wells after the end of 3rd stage of injection into TR8A. As the detailed

seismic response to the reinjections in the central part of Berlin reservoir cannot be performed due

to missing (re)injection data from other wells, in the following we focus on the 102 relocated

earthquakes related to the three stimulations of well TR8A (dark blue circles in Fig. FEE and FHH,

family 1).

Seismic response to injections Injections into TR8A have been performed at two different intervals (Fig. FDD, I1+I2 and I2 in

Fig. FIIa-c). The first two phases I1 and I2 were performed at the interval 1250-1300m below the see

level and I3 was performed at a depth of 1650-1700 m (cf. Fig. FEE for injection depths).

Figure FHH. Seismic activity recorded during three injection phases PH1-PH3 (top to bottom) into the well TR8A. The changes in well head pressure and injection rate (left axis) are shown using red and deep blue solid lines. Events are shown using different colors corresponding to the family number and the vertical coordinate is related to the moment magnitude (right axis). Numbers below events from family name correspond to those presented in adequate subfigures of Fig. FII. Green stars marks the outbursts of the seismicity in the central part of Berlin HFR site (see text for details), whereas red star denote the LME and its aftershock sequence.

Figure FII. Detailed spatio-temporal distribution of seismic events occurred during three stimulation phases (subfigures a, b and c) into well TR8A. The events occurring during the actual stimulation phase as shown as black dots. The radius of colored circles surrounding these events corresponds to the moment magnitude, whereas color reflects the injection rate at the moment specific event occurred (warmer color – higher injection rate). The numbers and gray vectors mark the migration of seismic events in time and space. Grayed dots correspond to all seismic events recorded during three stimulation phases for comparison. Empty grayed points show remaining activity that occurred outside of injection periods. Known surface traces of faults are marked as solid red lines. The dashed red lines mark faults inferred using head-on and mise-a-la-masse techniques. Possible locations of other faults inferred from head-on profiles are shown using dotted red lines finished with arrows on both sides. The wells are shown using dark blue and injection intervals I1+I2 and I3 are shown as blue thick line over the well trace. The sketch in the top-right corner provides a simplified and interpreted version of each subfigure (see text for details).

First injection phase I1 (Fig. FHHa, FIIa) began with a few short lasting stimulations (stage 1A

in Fig. FHHa) and injection rates not exceeding 50l/s. At this stage no seismicity occurred close by

TR8A. The injection rate was increasing stepwise reaching the level about 105l/s on 11th of July, 2003

when the injection was stopped (stage 1B in Fig. FHH). We found first events during stage 1B (events

E1-E12) occurred firstly close by the injection well and then propagated consequently towards SE

until the time of the shut-in. Interestingly, the moment magnitudes in 1B phase were slowly

decreasing with time despite increasing injection rate. The physical origin of this behavior is not

clear. The events recorded at stage 1B form a linear structure propagating SE. Although there are no

surface traces of faults close by TR8A, Santos and Rivas (1999) found evidence for buried faults in the

direct vicinity of TR8A using head-on and mise-a-la-masse techniques. These inferred faults form an

extension to faults surveyed on the surface and they are shown as dashed red lines in Fig. FII and

marked as F1 and F2. It is likely that events from stage 1B of injection occurred on this pre-existing

fault plane, as the faults are dipping nearly vertically (P. Santos, personal comm.) and the existing

shift of ca. 50m between the inferred fault plane and actual distribution of relocated seismic events

may be simply related to uncertainties of the velocity model. Therefore, the generation mechanism

of these events probably involves decrease in the effective normal stress on the fault plane reducing

the resistance against strain energy and leading to accelerating microseismicity progressing outside

of injection point. For a few events that occurred on the inferred fault plane we calculated the

constrained double-couple seismic moment tensors using the amplitudes of the first P-wave phase

from vertical components following the procedure presented in Wiejacz (1992), developed for

mining-induced seismicity. Two examples of fault plane solutions are shown in Fig. FJJ. Although the

results are rather tentative due to the limited number of sensors (effectively 6, as due to location

sensors are grouped in pairs on the focal sphere) and persistent problems with sensor polarities, the

calculated fault plane solutions display strike-slip faulting with minor normal component where the

one fault plane corresponds to the strike of inferred F1 fault plane. This is in agreement to composite

fault plane solutions presented in Oates et al. (2004). Additional confirmation of the spatial

development of the seismicity outside of injection zone over the pre-existing fault plane is directions

of the maximum horizontal stress which is approximately NNW-SSE in this area (Oates et al., 2004).

Therefore, the seismicity develops along the inferred fault F1 according to the direction of the

maximum horizontal stress.

After the shut-in of the well on 11th of July, 2003, the expansion of seismicity towards SE

stops. In the relatively short time period we observe a few seismic events distributed rather

randomly around the injection well (events E13-E15) followed by a sudden outburst of seismicity on

the fault plane F1 (stage 1C), this time NW from the initial point when the events started to

propagate along the fault plane toward SE in previous stage 1B. Events in stage 1C are characterized

by on average larger magnitudes than that from 1B. This somehow supports the frequently observed

phenomena in case of fluid-induced seismicity that the largest events occur after the shut-in of the

well (REFS). It is worth to mention that these events do not propagate outside of the initial point of

injection but rather form a clustered structure that does not display any specific spatiotemporal

pattern. This is likely because due to the shut-in of the well there is no driving force (i.e. injection) to

force seismic events to propagate outside of the injection area. Finally, the last stage of injection is

performed between 14 and 17 of July, 2003 with even higher injection rates reaching 110l/s. This

part induced only two seismic events that occurred in the same area activity observed in stage 1B. It

is worth to notice these events occurred only when the previous highest injection rate from 1B was

exceeded by 5l/s. The final shut-in on 17th of July did not resulted in the outburst of activity close-by

injection well TR8A, as observed in stage 1C. One possible explanation is that… In the intermediate

stages with no injections into TR8A some seismic events (marked with green arrows in Fig. FHH)

occurs in the central part of the geothermal site supporting the idea that the activity is directly

related to areas subjected to strong injections at a certain time.

Figure FJJ. Fault plane solutions derived from the double-couple moment tensors for two events that occurred on the fault plane F1 (cf. Fig. FIIa) during injection I1. The lines presents the relatively high level of uncertainty of calculated focal mechanism.

The second injection phase I2 began in August 2003 (Fig. FHHb and Fig. FIIb). It was composed of

only one major injection with injection rate increasing in two steps. In the midpoint of stage 2B

[name] was performed aiming at *…+. Initial stage 2A did not induced any seismicity in the vicinity of

TR8A, however the second stage 2B with injection rates mostly around 120l/s including oscillating

injection-rate (stage 2C) lead to the occurrence of another sequence of seismic activity presented in

Fig. FIIb. Comparing injection rates and well head pressures from I2 and I1 with occurrence of

seismicity it becomes clear that we observe some type of Kaiser Effect (Holcomb, 1993, Lavrow,

2003). We conclude induced seismicity occurs during injection phases only if the consecutive

injection campaigns display an increase in the injection rate and/or the well head pressure.

The seismic response to I2 is slightly different than for I1. The seismicity begins with two

earthquakes (E2-3) that occurred in the northern part of the investigated area in the vicinity of TR8A.

These two events, as well as some other recorded in the later parts of I2 (events E13-14, E22, E26)

form another SE-NW striking structure that apparently corresponds to another fault plane (F2 in

Fig. FIIb) inferred by Santos and Rivas (1999). Together with seismicity related to fault F2 we again

observe the migration of seismic activity outside of injection area on the fault F1 (E4-E9 in Fig. FIIb)

as well as the decrease in moment magnitudes, exactly in the same manner observed in stage 1B.

The migration finished right after the first drop down in the injection rate from 140l/s to 120l/s (color

arrow in Fig. FHHb). The following oscillations of injection rate in stage 2C resulted in “leaping” of

seismicity between faults F1 and F2 which continued until the end of stage 2C. For the remaining

period of injection after oscillations in the injection rate we do not recorded any activity, once again

confirming the existence of Kaiser Effect. After the shut-in of the well, another outburst of seismicity

in the central part of Berlin field is observed (green arrow in Fig. FHHb), however there is no

seismicity observed close-by injection well TR8A.

It is worth to note that the moment magnitudes observed on fault F1 are similar for both

injections during injection stages whereas moment magnitudes on fault F2 are slightly lower. We

speculate this is because F1 seems to be strongly affected by increased pore pressure in comparison

to the F2 resulting in dominance of larger events on F1. The seismic response in terms of moment

magnitude for the following injection I3 is much weaker.

The third injection was performed in the deeper part of the reservoir (Fig. FHHc and Fig. FIIc). It is

composed of two major stimulations denoted 3A and 3B in Fig. FHHc. The spatial and temporal

pattern is different from injections I1 and I2. This is likely because another formation ca. 500 below

the previous injection point was stimulated (I2 in Fig. FII). The seismicity did not emerge unless

injection rate and well head pressure exceeded 90l/s and 130bars, respectively. The activity during

this stage is composed of the one outburst of the seismicity just W from the injection point I3

(Fig. FIIc, phase 3A) and the formed cluster do not display any particular structure or migration in

time. The location of the cluster suggests that firstly it could be tentatively related to another fault F3

inferred only from the head-on technique (Fig. FIIc). The closest fault parallel to F3 is located SE close

by the heads of injection wells TR1 and TR10 so it is feasible that similar structure indeed exists in the

proximity of TR8. As the relative distances between events are always preserved in the double-

difference technique, we must consider the seismicity recorded during I3 together with earthquakes

recorded during I1 and I2. We see that aligned activity in I1 and I2 is slightly (ca. 80m) south from

inferred fault planes F1 and F2. If we believe that aligned activity from I1+I2 is indeed related to

faults F1 and F2, we may manipulate the velocity model within some constraints in order to shift the

aligned seismicity to inferred faults. This will result in the clustered seismicity being shifted to the

area right in the middle between F1 and F2 where there is no inferred fault once we shift all locations

northwards. If we consider shifting the epicenters westwards, we will find the clustered seismicity

from stage 3A in the direct proximity of injection well. As the clustered seismicity from stage 3A does

not display any particular structure, we speculate that these events occurred in the direct vicinity of

the deeper injection interval rather than on the fault F3.

The Kaiser effect is also apparent for the deeper injection, as the second phase of injection 3B,

performed at generally lower injection rates does not induced the seismicity. After the shut-in of the

well, three more seismic events occurred SW from the TR8A well head. There is no known fault in

this area, however head-on method performed by Santos and Rivas (1999) suggest the existence of

such structure in depth (P. Santos – personal comm.) marked as F4 in Fig FIIc.

Static stress drop and injections We investigated the dependence of the static stress drop on the distance from injection point. If we

consider all three injections into TR8A altogether, the static stress drop values do not show any

specific spatial and temporal pattern. However, interesting patterns appear if we focus only on the

aligned seismic related to the fault plane F1 (stimulations 1-2, Fig. FIIa,b) or close-by injection well

(stimulation 3, Fig. FIIc). Fig. FLL shows the dependence between the static stress drop of seismic

events and their distances to the adequate injection interval. For I1 and I2 we selected seismic events

that occurred in the vicinity of fault plane F1 during injections phases 1B and 1C (cf. Fig. FHHa-b),

respectively (i.e. we did not consider events that occurred after the shut-in). In case of injection I3,

we selected the clustered seismicity that occurred close-by injection interval I3 during phase 3A

(Fig. FHHc).

Despite of a few outlaying values, the static stress drop of events recorded during I1 and I2

and located on the inferred fault F1 (Fig. FLLa-b) seems to increase with the distance from injection

point (as well as the duration of injection). The initial values of the static stress drops are of order of

~1MPa for events located about 100m from injection point. The static stress drop values grow up

slowly with increasing distance reaching 10MPa for microearthquakes located 500m from the

injection point. In both cases the slope of the growth of static stress drop with distance is

comparable, but the actual correlation is better resolved for seismic events recorded during injection

I2. Similar relation has been observed for earthquakes induced by hydraulic stimulations in crystalline

rocks for a deep heat mining project in Basel (Goertz-Allmann et al., 2011), where an increase of

stress drop by a factor of 5 was observed for microearthquakes located at distances 10-300m from

the injection well. The authors attribute observed increase to pore pressure perturbations due to the

injection. The relation between the static stress drop and distance for injection I3 is different. Here,

the seismic activity is possibly not attributed to any existing fault plane and it shows the clustered

structure without any signatures of migration. This is somehow reflected in very consistent values of

the static stress drop which display quite similar values ranging 1-5MPa. Relatively low number of

seismic events and lack of additional geomechanical information refrained us from analyzing the

detailed relations between stress drops and pore pressure perturbations due to injections, as

performed by Goertz-Allmann et al. (2011).

Figure FLL. The dependence between the static stress drops of seismic events recorded during injections I1-I3 and the distance from the adequate injection point. The numbers matches the relevant event numbers in Fig. FHHa-c. Inset: the dashed ellipses denotes locations of plotted seismic events.

MW3.6 LME and injections Two weeks after the injection I2, the large magnitude earthquake occurred in the central part of the

reservoir, ca. 3km from the TR8A. The aftershock sequence is marked with the red arrow in Fig.

FHHb. The aftershock activity following the LME occurs only in the central part of the field with all

events displaying apparently the comparable focal mechanisms, as they all without any exception

belong to the family 3 (cf. Fig. FEE). Similarly to argumentation presented in Bommer et al. (2006),

the refined results do not provide evidence that the multiple stimulations performed in the well

TR8A triggered the LME. However, as we observe strongly and clearly visible spatial and temporal

correlation between the intensity of injection and the seismic activity, the question remains how the

extreme perturbations in the spatial distribution of reinjected water due to multiple stimulations in

TR8A affected the geothermal field as a whole. Unfortunately, due to the lack of injection data from

other wells, we cannot address this issue in more details.

7. Conclusions 1. The seismic dataset composed of 581 events recorded during the period of 16 months

between October 2003 and January 2004 was intensively refined in terms of hypocenter

locations and source parameters. The presented highly precise and detailed analysis of

spatial and temporal development of the induced seismicity and its physical properties could

not be possible without application of double-difference technique (refinement of locations)

and spectral ratio (refinement of source parameters) methods.

2. The seismic activity observed during stimulation phases of TR8A well display clear spatial and

temporal patterns that could be easily related to the amount of water injected into the well

TR8A and other reinjection wells in the investigated area.

3. The Kaiser effect is observed both for stimulation of shallower part of the reservoir (stages I1

and I2) and for the deeper stimulation (stage I3). Generally, the seismicity emerges only after

certain level of the injection rate (or well head pressure) is exceeded (about 70l/s and 90l/s

for I1+I2 and I3, respectively) and the following stimulation episodes in the same interval

requires increased injection rate level to re-activate the seismic activity.

4. During the stepwise increase of injection rates in the shallower part of the reservoir

(especially for stimulation I1) we observe migration of seismicity outside of the initial point

located close by open hole section of the well. The seismicity migrates outside of the initial

point towards SE only when an increase in the injection rate is still observed. The migration

stops when the well is finally shut-in. This behavior is/was observed in other geothermal

reservoirs (e.g. Soultz-sous-Forets, Basel) but here the migration is visible in much smaller

scale with total migration length not exceeding 150m.

5. The locations of seismic events observed during injections I1 and I2 display clear spatial

alignment trending NW-SE. We attributed this aligned activity to the fault system inferred

from the mise-a-la-masse and head-on techniques and already mapped surface fault traces.

This is additionally supported by the calculated fault plane solutions of a few seismic events

that occurred in this area. The derived focal mechanisms display strike-slip behavior with one

fault plane aligned along the fault plane system inferred from the non-seismological studies.

Therefore, the events detected during injections I1 and I2 propagate outside of injection

point along the pre-existing fault plane until the shut-in. The generation mechanism possibly

involves decrease in the effective normal stress due to increase in the fluid pressure on the

pre-existing fault plane. The direction of propagation is also consistent with the direction of

the maximum horizontal stress in the investigated area.

6. The last injection I3 into the deeper formations possibly represents the clustered activity in

the direct proximity of injection area not attributed to any existing fault plane. As the

injection rate is relatively low despite much stronger well head pressure during the third

stimulation, it is likely that the fracture network could not develop distinctively and migrate

outside of the injection point as in the injection I1+I2.

7. After the shut-in of the well during injection I1 we observe firstly a few seismic events

relatively far away from the fault, likely at the rim of pressure gradient zone induced by the

injection that is ellipsoidal in shape with the longest axis trending NW-SE. After that, we

observe an outburst of the seismic activity back on the fault plane F1, but in the opposite

direction NW from the initial injection point. This area did not display seismic activity during

the stimulation. Therefore, we conclude larger magnitude events observed in this area were

triggered by imbalance in the shear stress distribution on the fault plane F1 caused by

previous fluid injection that induced the seismicity in SE part from injection point. The area

where the larger magnitude events occurred corresponds to the undamaged part of the

fault.

8. We observe increase in the static stress drop values with the distance from injection point

for events that migrate along the fault plane F1. This supports observations performed by

Goertz-Allmann et al. (2001) in deep heat mining project in Basel, where similar relation was

observed and attributed to the pore pressure perturbations. As already suggested by

Goertz-Allmann et al. (2001), this suggest static stress drop may be a proxy for monitoring

pore pressure distribution if elastic parameters of the stimulated volume of rocks is know

reasonably well.

9. The refined results do not provide any evidence that the multiple stimulations performed in

the well TR8A triggered the LME of moment magnitude MW3.6 recorded 2 weeks after the

second injection into TR8A. The seismic activity induced by multiple injections into TR8A did

not reach distances larger than 500m from the injection point and we do not observe any

sequence of events migrating further away towards the nucleation area of MW3.6. However,

the injection operations at TR8A supposedly strongly disturbed the distribution of pore

pressure at depth as well as influence the fluid migration pattern. As the fluids reinjected

into multiple wells including TR8A comes from the same pool, it is still likely the TR8A

stimulation could be indirectly responsible for the occurrence of LME. This requires however

the detailed analysis of the response of reservoir to fluid (re)injection(s) in the remaining part

of the reservoir.

Acknowledgments We would like to acknowledge Julian Bommer and Gunter Siddiqi for many valuable comments

regarding data processing and injection operations performed in Berlin HFR site. We would like to

thank Sabrina Andrae and Oliver Germer for improving the quality of the original catalog. Pedro

Antonio Santos is acknowledged for providing additional data and comments that helped us to

finalize the manuscript. We thank NORSAR (Volker Oye, Julie Albaric) for providing the dataset and

for comments regarding the actual data processing. G.K. would like to thank Patricia Martinez-Garzon

for providing translations of some papers and geological profiles. This paper is supported by GEISER

(Geothermal Engineering Integrating Mitigation of Induced Seismicity in Reservoirs) project no.

XXXXX, European Commission.

Bibliography S. Baisch, R. Voros, E. Rothert, H. Stang, R. Jung, and R. Schellschmidt. A numerical model for fluid

injection induced seismicity at Soultz-sous-Forêts. Int. J. Rock Mech. Min. Sci, 47(3): 405–413, 2010.

doi: 10.1016/j.ijrmms.2009.10.001.

J. Boatwright. Detailed spectral analysis of two small New York State earthquakes. Bull. Seismol. Soc.

Am., 68: 1131–1177, 1978.

J. Boatwright and J.B. Fletcher. The partition of radiated energy between P and S waves. Bull.

Seismol. Soc. Am., 71: 361–376, 1984.

M. Bohnhoff, S. Baisch, and H.-P. Harjes. Fault mechanisms of induced seismicity at the super deep

German continental deep drilling program (KTB) borehole and their relation to fault structure and

stress field. J. Geophys. Res., 109(B2): B02309, 2004. doi: doi:10.1029/2003JB002528.

J.J. Bommer, S.Oates, J.M. Cepeda, C. Lindholm, J. Bird, R. Torres, G. Marroquin, and J. Rivas. Control

of hazard due to seismicity induced by a hot fractured rock geothermal project. Eng. Geology,

83(4): 287–306, 2006. doi: 10.1016/j.enggeo.2005.11.002.

J. J. Bommer, G. Georgallides, and I.J. Tromans. Is there a near-field for small-to-moderate magnitude

earthquakes? Journal of Earthquake Engineering, 5 (3), 395—423, 2001

D.M. Boore and J. Boatwright. Average body-wave correction coefficients. Bull. Seismol. Soc. Am.,

74: 1615–1621, 1984.

F. Bulut, W.L. Ellsworth, M. Bohnhoff, M. Aktar and G. Dresen. Spatiotemporal earthquake clusters

along the North Anatolian Fault Zone offshore Istanbul, Bull. Seismol. Soc. Am., 101 (4): 1759-1768,

2011. doi: 10.1785/0120100215.

S.D. Davis and C. Frohlich. Did (or will) fluid injection cause earthquakes? — Criteria for rational

assessment. Seism. Res. Lett., 64: 207–224, 1993.

J.D. Eshelby. The determination of the elastic field of an ellipsoidal inclusion, and related problems.

Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences,

241(1226): 376–396, 1957. doi: 10.1098/rspa.1957.0133.

K.F. Evans, H. Moriya, H. Niitsuma, R.H. Jones, W.S. Phillips, A. Genter, J. Sausse, R. Jung and R. Baria.

Microseismicity and permeability enhancement of hydrogeologic structures during massive fluid

injections into granite at 3 km depth at the Soultz HDR site. Geophys. J. Int. 160, 389-412, 2004.

K.F. Evans, A. Zappone, T. Kraft, N. Deichmann, and F. Moia. A survey of the induced seismic

responses to fluid injection in geothermal and CO2 reservoirs in Europe. Geothermics, 41(0):30–54,

2012. doi: 10.1016/j.geothermics.2011.08.002.

H. Fabriol, J. Rivas, S. Handal, and H. Correia. Microseismicity and geothermal observation of Berlín

(El Salvador). Geothermal Res. Coun. Trans., 22: 159–163, 1998.

D. Galluzzo, E. Del Pezzo, M. La Rocca, M. Castellano and F. Bianco. Source scaling and site effects at

Vesuvius volcano. Bull. Seismol. Soc. Am., 99: 1705-1719, 2009

M. Godano, T. Bardainne, M. Regnier, A. Deschamps and M. Valette. Spatial and temporal evolution

of a microseismic swarm induced by water injection in the Arkema-Vauvert salt field (Southern

France). Geophys. J. Int. 188: 274-292, 2012.

B.P. Goertz-Allmann, A. Goertz and S. Wiemer. Stress drop variations of induced earthquakes at the

Basel geothermal site. Geophys. Res. Lett. 38, 1-5, 2011.

T. Fischer, and A. Guest. Shear and tensile earthquakes caused by fluid injection. Geophys. Res. Lett.

38, 2-5, 2011.

H. K. Gupta. Are RIS events of M5 preceded by a couple of foreshocks of m4? Bull. Seismol. Soc.

Am., 82(1): 517–520, 1992.

T. C. Hanks and H. Kanamori. A moment magnitude scale. J. Geophys. Res., 84: 2348–2350, 1979.

D.J., Holcomb. 1993. „General theory of the Kaiser effect”. Int. J. Rock Mech. Min. Sci. &

Geomechanics Abstracts 30 (7): 929–935. doi:10.1016/0148-9062(93)90047-H.

S. Ide and G.C. Beroza. Does apparent stress vary with earthquake size? Geophys. Res. Lett.,

28: 3349–3352, 2001.

L. Ingber. Very fast simulated re–annealing. Math. Comput. Modeling, 12(8): 967–973, 1989.

L. Ingber. Simulated annealing: Practice versus theory. Math. Comput. Modeling, 18(11):29–57, 1993.

G. Kwiatek, M. Bohnhoff, G. Dresen, A. Schulze, T. Schulte, G. Zimmermann, and E. Huenges.

Microseismicity induced during fluid-injection: A case study from the geothermal site at Gross

Schoenebeck, North German Basin. Acta Geophys., 58(6): 995–1020, 2010. doi: 10.2478/s11600-010-

0032-7.

G. Kwiatek, K. Plenkers, and G. Dresen. Source parameters of picoseismicity recorded at Mponeng

deep gold mine, South Africa: Implications for scaling relations. Bull. Seismol. Soc. Am., 101(6):2592–

2608, 2011. doi: 10.1785/0120110094.

J.C. Lagarias, J.A. Reeds, M.H. Wright, and P.E. Wright. Convergence properties of the nelder–mead

simplex method in low dimensions. SIAM J. Optimization., 9: 112–147, 1998.

doi: 10.1137/S1052623496303470.

A. Lavrov. The Kaiser effect in rocks: principles and stress estimation techniques. Int. J. Rock Mech.

and Min. Sci. 40, 151–171, 2003.

R. Madariaga. Dynamics of an expanding circular fault, Bull. Seismol. Soc. Am. 66, 639–666, 1976.

E.L. Majer, R. Baria, M. Stark, S. Oates, J. Bommer, B. Smith, and H. Asanuma. Induced seismicity

associated with enhanced geothermal systems. Geothermics, 36(3): 185–222, 2007. doi:

10.1016/j.geothermics.2007.03.003.

S. Michelet and M.N. Toksöz. Fracture mapping in the Soultz-sous-Forêts geothermal field using

microearthquake locations. J. Geophys. Res. 112, 1-14, 2007.

I. Moeck, G. Kwiatek, and G. Zimmermann. Slip tendency analysis, fault reactivation potential and

induced seismicity in a deep geothermal reservoir. J. Struct. Geol., 31:1174–1182, 2009. doi:

10.1016/j.jsg.2009.06.012.

J. Nelder and R. Mead. A simplex method for function minimization. Comput. J., 7: 308–313, 1965.

S. Oates, B. Hustedt, G. Siddiqi, B. Worrall, and M. Darnet (2004). Microseismic monitoring of the HFR

hydraulic stimulation of well TR8A in the Berlin geothermal field, El Salvador, Shell internal report

J.C. Pechmann, W.R. Walter, S.J. Nava, and W.J. Arabasz. The February 3, 1995, ML5.1 seismic event

in the Trona Mining District of Southwestern Wyoming. Seism. Res. Lett., 66(3): 25–34. doi:

10.1785/gssrl.66.3.25.

D.B. Percival and A.T. Walden. Spectral Analysis for Physical Applications: Multitaper and

Conventional Univariate Techniques. Cambridge University Press, 1993.

E. Del Pezzo, F. Bianco and G. Saccorotti. Seismic source dynamics at Vesuvius volcano, Italy. J.

Volcanology Geoth. Res., 133: 23-39, 2004

J.A. Rivas. Seismic monitoring and its application as an exploration tool in the Berlin geothermal field,

El Salvador. The United Nations University Reports, 17: 355-384, 2000.

P.A. Santos, and J.A. Rivas. Geophysical methods applied in the detection of faults in the Berlín

geothermal field (in Spanish). Presented at the Geophysical Seminar, IPGH-CIG (MOP), El Salvador,

1999.

S. Sasaki. Characteristics of microseismic events induced during hydraulic fracturing experiments at

the Hijiori hot dry rock geothermal energy site, Yamagata, Japan. Tectonophys., 289: 171-188, 1998.

doi: 10.1016/S0040-1951(97)00314-4.

P. Segall, J.R. Grasso, and A. Mossop. Poroelastic stressing and induced seismicity near the lacq gas

field, Southwestern France. J. Geophys. Res. 99, 15423–15438, 1994.

M. Schoenball, T.M. Müller, B.I.R. Müller and O. Heidbach. Fluid-induced microseismicity in pre-

stressed rock masses. Geophys. J. Int. 180, 813-819, 2010.

J.A. Snoke. Stable determination of (Brune) stress drops. Bull. Seismol. Soc. Am., 77:530–538, 1987.

E. Torio-Henríquez, 2007. Petrography and mineral alteration in Berlin geothermal field. Proceedings

of the 32nd Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford, California,

January 22-24, 2007, SGP-TR-183

V. Vavryčuk, M. Bohnhoff, Z. Jechumtálová, P. Kolář and J. Šílený. Non-double-couple mechanisms of

microearthquakes induced during the 2000 injection experiment at the KTB site, Germany: A result of

tensile faulting or anisotropy of a rock? Tectonophys. 456, 74-93, 2008.

F. Waldhauser and W.L. Ellsworth. A Double-Difference earthquake location algorithm: Method and

application to the Northern Hayward Fault, California. Bull. Seismol. Soc. Am., 90(6): 1353 –1368,

2000. doi: 10.1785/0120000006.

P. Wiejacz. Calculation of seismic moment tensor for mine tremors from the Legnica-Głogów Copper

Basin. Acta Geophys. Pol., 40: 103–122, 1992.

T. Yamashita. Pore creation due to fault slip in a fluid-permeated fault zone and its effect on

seismicity: generation mechanism of earthquake swarm, Pure Appl. Geophys., 155, 625–647, 1999.

T. Yamada, J. J. Mori, S. Ide, R. E. Abercrombie, H. Kawakata, M. Nakatani, Y. Iio, and H. Ogasawara.

Stress drops and radiated seismic energies of microearthquakes in a South African gold mine. J.

Geophys. Res., 112: B03305, 2007. doi: 10.1029/2006JB004553.

For Peer Review

Geomechanical modeling of induced seismicity source

parameters and implications for seismic hazard assessment

Journal: Geophysics

Manuscript ID: Draft

Manuscript Type: Technical Paper

Date Submitted by the Author: n/a

Complete List of Authors: Goertz-Allmann, Bettina; ETH Zurich, Swiss Seismological Service, Insitute of Geophysics Wiemer, Stefan; ETH Zurich, Swiss Seismological Service, Insitute of Geophysics

Keywords: earthquake, geothermal, microseismic, risk

Area of Expertise: Passive Seismic

GEOPHYSICS

For Peer Review

Geomechanical modeling of induced seismicity source

parameters and implications for seismic hazard assessment

Bettina P. Goertz-Allmann∗ and Stefan Wiemer∗

∗ Swiss Seismological Service, ETH Zurich, Institute of Geophysics,

(bettina.allmann@sed.ethz.ch, stefan.wiemer@sed.ethz.ch)

(March 19, 2012)

Running head: Modeling induced seismicity hazard

ABSTRACT

We simulate induced seismicity within a geothermal reservoir using pressure-driven stress

changes and seismicity triggering based on Coulomb friction. The result is a forward-

modeled seismicity cloud with origin time, stress drop, and magnitude assigned to each

individual event. Our model includes a realistic representation of repeating event clusters,

and is able to explain in principle the observation of reduced stress drop and increased

b-values near the injection point where pore-pressure perturbations are highest. The higher

the pore-pressure perturbation, the less critical stress states still trigger an event, and

hence the lower the differential stress is before triggering an event. Less critical stress

states result in lower stress drops and higher b-values, if both are linked to differential

stress. We are therefore able to establish a link between the seismological observables and

the geomechanical properties of the source region and thus a reservoir. Understanding the

geomechanical properties is essential for estimating the probability of exceeding a certain

magnitude value in the induced seismicity and hence the associated seismic hazard of the

1

Page 1 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

operation. By calibrating our model to the observed seismicity data we can estimate the

probability of exceeding a certain magnitude event in space and time and study the effect

of the injection depth and crustal strength on the induced seismicity.

2

Page 2 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

INTRODUCTION

Hydraulic fracturing is an increasingly utilized technology to enhance the extraction of

hot water or tight gas from a subsurface reservoir. Fluids are pressed into the reservoir

formation from a treatment well at high pressures to open fractures and hence increase

the permeability of the reservoir. Monitoring the seismic emission associated with the fluid

injection allows to estimate the stimulated reservoir volume, and hence the effectiveness

of the treatment. However, oftentimes little is known about the mechanical properties of

the reservoir rocks, making it difficult to predict the response of the medium to the fluid

injection. On the one hand, one would like to ensure that the fluid injection operation alters

the medium sufficiently to make the reservoir economic, and on the other hand it needs to

be ensured that the magnitude of induced seismic events does not exceed values where

shaking can affect surface infrastructure. A proper estimation of the in-situ mechanical

properties of the reservoir is therefore necessary for an assessment of both the economics

and environmental safety of reservoir treatment as well as the associated seismic hazard at

the surface.

The increase of pore pressure pp due to fluid injection reduces the effective normal stress

and can cause microseismicity in near-critically stressed rocks (Pearson, 1981; Zoback, 2007).

Propagation of the microseismic event cloud through the medium can be described by the

diffusive process of pore-pressure relaxation (Talwani and Acree, 1984). The hydraulic diffu-

sivity of the medium can be estimated from the time-distance dependence of the seismicity

triggering front (Shapiro et al., 1997, 2002). Apart from the spatio-temporal behaviour

of seismicity, source properties of induced seismicity appear to be also influenced by the

hydraulic medium properties. Using data from a geothermal operation in Basel, Goertz-

3

Page 3 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Allmann et al. (2011) showed that the Brune stress drop of induced events is significantly

lower near the injection point and inversely correlates with estimated pore-pressure values.

Similarly, Bachmann et al. (2012) measured the spatial variation of Gutenberg’s b-value on

the same dataset and found that b-values are higher near the injection point where pore

pressures are higher. Seismic source property observations such as the latter two contain

information about the in-situ stress regime: the b-value is linked to the differential stress

(Amitrano, 2003; Schorlemmer et al., 2005), and stress drop denotes the difference in shear

stress on the fault plane before and after the earthquake (e.g., Kanamori and Brodsky,

2004).

The main aim of this paper is twofold: Firstly, we seek a possible explanation for

the observed spatio-temporal behavior of stress drop and b-value. We forward-model the

pressure-induced stress changes and seismicity triggering based on Coulomb friction. We

link both stress drop and b-value to the differential stress before assigning a magnitude and

stress drop to each event. The result is a simple geomechanical model of induced seismicity

that is able to explain the observed source property variations to the first order. Secondly,

we attempt to use the simple geomechanical model to predict aspects of the behavior of

an induced seismicity cloud. The modeled source property variations in space and time

allow us to calculate the probability of exceeding a certain magnitude event and therefore

make a prediction of the seismic hazard of an injection operation. This last step requires

a calibration of the model to observed data. In our case, we use the observed seismicity

of the Basel geothermal stimulation (Haring et al., 2008; Deichmann and Giardini, 2009).

Due to the calibration step, the obtained probabilities are strictly valid only for the Basel

case. However, the main characteristics of the modeled probability curves are independent

of the calibration step and allow us to make some generalized statements. To illustrate the

4

Page 4 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

sensitivity to the underlying stress regime, we present example calculations of the event

probability for a few selected stress scenarios. If calibrated repeatedly using an evolving

measured seismicity during a stimulation operation, the presented method may be used in

the future for a near real-time prediction of induced seismicity.

Triggering of microseismicity based on a pressure perturbation has been modeled in the

past, however no link to seismic source properties was established in previous modeling

approaches. Rothert and Shapiro (2003) developed a numerical model to trigger microseis-

micity due to the process of pore-pressure relaxation and statistically pre-defined critical

zones in the medium. Their model is able to describe the spatio-temporal distribution of in-

duced seismicity reasonably well but it does not include any event size distribution or other

source parameters such as stress drop. A similar approach is used by Langenbruch and

Shapiro (2010) to investigate the behaviour of post-injection seismicity and the approach

by Hummel and Muller (2009) considers non-linear pore fluid pressure diffusion. Schoenball

et al. (2009) find that both the influence of hydraulic properties and the coupling of pore

pressure to the stress field have an effect on the occurrence of microseismicity but again

they do not concentrate on event magnitudes. Kohl and Megel (2007) use a finite-element

approach to model the hydro-mechanical response of the medium. Their approach attempts

to model the opening or shearing of discrete (but stochastically distributed) fractures in the

medium. It can take into account a Coulomb failure criterion to associate seismicity with

shearing and hence model location and time of seismic events. However, it cannot predict

seismic source properties such as magnitude and stress drop. Another study also includes

the opening of stochastically distributed fractures due to fluid flow induced stress changes

to obtain a microseismicity distribution (Bruel, 2007). This study also obtains a magnitude

of each event from the slip and the rupture area but the magnitude distribution is basically

5

Page 5 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

pre-defined by the fracture distribution where each fracture can only fail completely.

In addition to assigning magnitudes and stress drop to each event, our rigorous ap-

plication of Mohr-Coulomb theory can also model repeating events at the same location.

Repeating events are important to properly describe effects very close to the injection point.

The ability to include event magnitudes allows us go one step further and use the model to

calculate the probability of exceeding a certain magnitude event as a function of time and

distance from the pressure source.

MODELING METHODOLOGY

We randomly distribute possible failure points (seeds) in a three-dimensional volume with

a constant background stress regime and constant friction coefficient and cohesion. At each

seed point, we randomly assign values for the minimum and maximum principal stress,

σ3 and σ1, assuming a Gaussian perturbation to a given background stress regime, and

optimally oriented faults. Such a stress perturbation could be interpreted as due to the

variation of elastic parameters (Langenbruch and Shapiro, 2011). We impose a limit on σ1,

such that it cannot exceed a maximum value as defined by the strong crust limit (Zoback,

2007) and on σ3, such that it cannot be smaller than the hydrostatic pore pressure ph.

The stress distribution for one example seed point is depicted in Figure 1a as a Mohr

diagram. The distribution of σ1 and σ3 is indicated by the histograms for all triggered

seeds (actual distribution after application of limiting constraints) and dashed bell curves

(nominal perturbation of values). Table 1 lists values for the assumed stress regime and

perturbations. These cases are based on the Basel geothermal example with numbers taken

from Haring et al. (2008) who investigated the depth-dependent stress regime in the Basel-1

borehole. At Basel, injection occurred at a depth of 4.5 km within the Granite. Since we

6

Page 6 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

require the medium to be in a stable equilibrium before the pore-pressure perturbation, the

upper bound of σ1 is effectively linked to the coefficient of friction. If we chose a smaller

coefficient of friction, all stable systems (Mohr circle just below the failure criterion) will

have smaller average values of σ1.

[Figure 1 about here.]

In an injection operation, the stress field is mainly modified by the injection pressure.

Increasing the pore pressure pp in the medium causes a reduction of the normal stress to

an effective stress. If the stress is near the critical state (Mohr circle close to the Coulomb

envelope) the reduction of the normal stress may cause the shear stress to exceed the

Coulomb failure envelope and hence trigger an event (Figure 1b). If pp continues to increase,

the failure envelope may be reached again at a later point in time due to the previous event

stress drop, thus triggering a repeating earthquake. We model the spatio-temporal evolution

of the effective stress based on a linear diffusion model in a hydraulically homogeneous and

isotropic model with a linearly increasing pressure source (Dinske et al., 2010),

p(t) = 4πDa0(p0 + ptt) , 0 < t < t0. (1)

For this case, Dinske et al. (2010) formulate an analytical solution to the diffusion equation

(Wang, 2000) that we will use in the following for our modeling. Values for the hydraulic

diffusivity D, the effective source radius a0, the starting pressure p0, the wellhead pressure

gradient pt, and the injection time t0 are listed in Table 1, and are again chosen to closely

follow the Basel case study. With the listed values, equation denotes a simplified approxi-

mation to the actual wellhead pressure in Basel. We assume an injection point in the center

of our volume. An event is recorded once the Mohr circle has reached the failure envelope.

7

Page 7 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Note that the number of seed points can be adjusted to modify the number of induced

events. We chose the number of seeds (30,000) such that the modeled seismicity’s a-value

(Gutenberg and Richter, 1944), describing the overall event productivity of the volume, is

similar to the a-value measured from the Basel event cloud (4.3). The large effective source

radius a0 used in our model compared to Dinske et al. (2010) trades off with the number

of seed points chosen in the model. The smaller the source radius, the more seed points

are required to induce an event cloud with a similar a-value. The choice of seed points is

bounded by considerations detailed in the Appendix. We do not define a minimum and

maximum triggering pore pressure (criticality). Theoretically, a stable stress state can be

infinitesimally close to the failure envelope, and hence the minimum pore pressure required

to trigger an event can be infinitesimally small. In practice, stable systems are systems with

a minimum triggering pressure larger than the average background pressure fluctuations

imposed by, e.g., tidal variations, ambient seismic noise, etc.. Choosing a0 = 70 m and

30,000 seeds results in a triggering front (envelope to all events) that roughly follows the

2000 Pa pressure contour (Figure 2). 2000 Pa is the maximum tidal-induced pressure

variation measured from borehole water level gauges in Basel (Evans, pers. comm.), and

therefore a viable assumption for the minimum triggering pressure.

[Figure 2 about here.]

Schorlemmer et al. (2005) infer that the b-value acts as a stress meter that depends

inversely on differential stress σd. This inverse relationship has also been suggested by

laboratory studies (Amitrano, 2003). To model the size of each induced event we link the

b-value to σd in an inversely proportional relationship. We randomly draw a magnitude

from the underlying Gutenberg-Richter relation. Event stress drop ∆σ is also assumed to

8

Page 8 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

be proportional to σd. After an event is induced due to the increase in pp, σ1 is reduced by

10%±5% of σd. We define the difference in shear stress τs before and after the reduction

of σ1 to be the stress drop (Figure 1b). Due to the stress drop the point in the medium

becomes stable again and a further increase in pp at a later time may repeat the entire

process at this point (Figure 1b). Note that the scaling between ∆σ and σd is arbitrarily

defined. A higher percentage would lead to higher absolute ∆σ values and a larger standard

deviation to a higher scatter in individual ∆σ. The used scaling factor results in a mean

unsmoothed ∆σ of 2.56 MPa and a variation over two orders of magnitudes of individual

∆σ values. This is similar to stress-drop estimates obtained for the Basel induced seismicity

(Goertz-Allmann et al., 2011).

In a next step we calibrate the modeled seismicity to the observed Basel data. To

calibrate the overall magnitude distribution we adjust the linear scaling relation between b-

value and differential stress σd (Figure 3). We make two assumptions: first, we choose σd =

136 MPa as the upper differential stress limit, which corresponds to the mean differential

stress plus respective standard deviations of σ1 and σ3 for the average crust at 4.5 km

depth (see Table 1). Second, we assume that the events at further distance behave similar

to tectonic events with a b-value of one. Events induced at larger distances to the injection

point are mainly events that are near critical failure and thus have on average larger σd.

While we fix the lower b-value limit (bmin) to one at σd = 136 MPa, we loop over possible

upper b-value limits (bmax) at σd = 0 (inset of Figure 3) and compute the L1 misfit

between the mean b-value of the observed data and the mean b-value from 50 model runs,

each with a different realization of random seeds and stresses. The best fit between the

Basel data and the modeled seismicity is observed at the minimum of the quadratic fit,

which corresponds to bmax=4 and therefore a linear scaling relation of b = −0.022σd + 4.

9

Page 9 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Apart from the magnitude calibration, we also want to equalize the number of induced

events between the Basel data and the modeled seismicity. Instead of modifying the seed

density, which also controls the overall number of induced events but is computationally

very expensive, we estimate a constant scaling factor s from 100 model runs with s =

(∑100

i=1 Nobserved/Nmodel)/100, where N is the overall number of induced events. We obtain

s = 0.92 and therefore randomly pick 92% of the modeled seismicity of each run for further

analysis.

[Figure 3 about here.]

The result is a forward-modeled seismicity cloud with location, origin time, stress drop,

and magnitude assigned to each event location in the medium and calibrated to match the

observed Basel seismicity. About 1000 events are induced with about 50% of these events

being repeaters (seed points rupturing more than once, Figure 2). This is consistent with

a multiplet analysis for the Basel microseismic events, which grouped 52% of all events

into multiplets (Kummerow et al., 2011). Most repeaters in our model occur close to the

injection point, where pore pressures are higher, and the number of repeats decreases with

overall pore-pressure perturbations, and hence with distance.

DISTANCE DEPENDENCE OF SOURCE PARAMETERS

Our model is able to explain the two basic seismological observations of reduced stress drops

(Goertz-Allmann et al., 2011) and increased b-values (Bachmann et al., 2012) near the

injection point where pore-pressure perturbations are highest. It also realistically simulates

the occurrence of repeating event clusters, consistent with observations (Kummerow et al.,

2011). Color-coding b-value and stress drop (Figure 4a and b) for one model run shows

10

Page 10 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

the distance-dependence of both compared to the observed Basel seismicity (Figure 4c). To

enable a fair comparison, we estimate the b-value from the synthetic seismicity cloud in the

same manner than for the real data: we compute the b-value from the event magnitudes

over a fixed number of 100 surrounding events to evalutate the lateral variation following

Bachmann et al. (2012). Stress drop is spatially smoothed using a median filter over the

closest 20 events. The smoothing of ∆σ is needed to investigate any spatial variability and

some spatial smoothing has also been applied to the observed Basel data (Goertz-Allmann

et al., 2011). We notice that the variance of the b-values and especially of the stress drops

is considerable, even though this is synthetic data and we based our modeling originally

on a comparatively moderate variance of 10 % for the principal stresses. To analyze the

properties of the synthetic cloud in a more robust way, we run 100 model runs, each with

a different realization of the random seeds and stresses. The result of the spatial b-value

and stress drop variations is summarized in Figure 5. While Figure 5a and b show a 3D

display and a cross-sectional view for one model run, Figure 5c shows a grid-stack of the

respective b-value (top) and stress drop (bottom) over 100 model runs. Both, mean b-value

and mean stress drop show a stable distance dependence from the injection point with

high b-value and low stress drop at the close distance. Figure 5d shows a corresponding

plot for a constant b-value of 1.48 and a constant stress drop of 2.5±1.25 MPa, without

any dependence on differential stress. In this case, any lateral changes in b-value or stress

drop are small random fluctuations and no distance dependence of these two parameters is

observed. If we plot the source parameters versus distance (Figure 6), we observe that the

distance dependence is strongest up to 200-300 m and decreases at further distances. This

is especially visible if we compute the mean of the mean values of each distance bin from

100 model runs (Figure 6c and d). The error bars in Figure 6a and b show the standard

11

Page 11 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

error of the mean values and have been computed using bootstrap resampling over 1000

realizations. The error bars in Figure 6c and d show the standard deviation.

[Figure 4 about here.]

[Figure 5 about here.]

[Figure 6 about here.]

The higher the pore-pressure perturbation, the less critical stress states still trigger an

event and hence the lower the differential stress can be before triggering an event (Figure 7).

The inset in Figure 7a shows the decrease of differential stress with increasing pore pres-

sure. This dependence results in lower differential stress close to the injection point and an

increase of differential stress at further distances (Figure 7b). Therefore, less critical stress

states result in lower stress drops and higher b-values, if both are linked to differential stress.

We expect the lowest ∆σ and highest b-value close to the injection point where the pore

pressure is the highest. In addition to the above described effect the distance dependence

of b-value and stress drop is strengthened by repeating events. Each repeat at one location

results in a smaller differential stress than its predecessor due to the previous event stress

drop. This leads to increasingly higher b-value and decreasingly lower stress drop for each

repeater. Since the number of repeaters increases close to the injection point (Figure 2) the

distance dependence of both stress drop and b-value is emphasized.

[Figure 7 about here.]

12

Page 12 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

BREAK IN SELF-SIMILARITY OF SEISMICITY

Many source parameter studies indicate that ∆σ is scale invariant for natural seismicity

(e.g., Abercrombie, 1995; Allmann and Shearer, 2009) as well as for induced seismicity

(e.g., Kwiatek et al., 2010; Goertz-Allmann et al., 2011). However, there are also various

studies that found an increase of stress drop with magnitude (e.g., Mayeda and Walter,

1996) and this topic remains a controversial question with important implications for seismic

hazard analysis.

Our model suggests a connection between b-value and stress drop since both parameters

are linked to differential stress. We therefore indirectly input a non self-similar stress drop

scaling with magnitude in our model. Figure 8 shows the relation between stress drop

and magnitude for different assumptions. No dependence of stress drop and magnitude is

observed if the stress drop is linked to differential stress but includes a 5% scatter (Figure 8a)

and we cannot distinguish the result from a constant stress drop model (Figure 8c). A slight

increase of stress drop with magnitude is observed for the raw stress drop values only if the

additional 5% scatter of the stress drop is omitted in the modeling (top of Figure 8b). This

can be better seen if we directly compare the mean values (Figure 8d). If a spatial smoothing

is applied to the individual stress drops (bottom row of Figure 8), the expected dependence

with magnitude becomes indiscernible. This shows that even synthetic stress drop scaling

is difficult to resolve. It appears to be an extremely challenging task to extract empirically

sound scaling relations from real data, where we have to deal with considerable additional

scatter due to noise and measurement errors. Furthermore, our result suggests that a spatial

smoothing of parameters should not be applied if scaling relations are investigated.

[Figure 8 about here.]

13

Page 13 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

TIME AND LOCATION OF LARGE MAGNITUDE EVENTS

Several conditions must be met for significant (damaging) earthquakes to occur. There must

be a fault large enough to allow significant slip, there must be stress present to cause the slip

along the fault and this stress must exceed the strength of the fault. It is not clear today

what is the upper possible magnitude limit (Mmax) for induced events in a comparatively

shallow reservoir setting (¡ 5 km depth). However this is an important input parameter to

probabilistic seismic hazard assessment.

Using our model we investigate the probability of an event exceeding a certain magnitude

with respect to injection time and distance from the injection point (Figure 9). We evaluate

the events for a- and b-values within specific time and distance bins. For each bin the

probability p of a certain magnitude M event is defined as (Wiemer, 2000),

p = 1 − e−(a−Mb). (2)

We choose time bins of 105 s, moving at 104 s intervals, and distance bins of 100 m, moving

across distance in 10 m increments. In order to obtain more stable results we analyze the

induced seismicity from 100 model runs and stack individual probabilities. This allows us

to compute a standard deviation to each mean probability estimate. The probability of a

large magnitude event is determined by two factors, i) the overall number of events per bin

(i.e. the a-value), and ii) the b-value in this bin. We find that the mean probability of

a large magnitude event increases with time up to the shut-in time (upper row Figure 9).

This can partially be explained by the steadily increasing number of induced events up to

shut-in and therefore increasing a-values, which is reflected by increasing probabilities up

to the shut-in for a constant input b-value model. A varying b-value causes a substantial

increase of the mean probability for the time period right after shut-in. The probability of

14

Page 14 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

a large magnitude event also increases to much further distances from the injection point

for a varying b-value compared to a constant b-value (bottom row Figure 9). The increased

probability at later times and larger distances is especially prominent for larger magnitudes

(Figure 9c). Note that the large difference in absolute probability before the shut-in time

between the constant and the varying b-value model mostly depends on the choice of a

constant b-value of 1.48. A larger constant b-value would bring the probability levels of the

two models closer at early times. The result of an increased probability of larger magnitude

events at later times just after the shut-in and at larger distances to the injection point

is consistent with observations of large magnitude events during hydraulic stimulations in

geothermal systems not only in Basel (Deichmann and Giardini, 2009), but also elsewhere

(Charlety et al., 2007; Evans et al., 2012).

[Figure 9 about here.]

DEPENDENCE ON DEPTH AND CRUSTAL STRENGTH

To test a more general applicability of the simple geomechanical model for the purpose

of forward-modeling expected seismic responses to fluid injections, we now investigate the

influence of some of the modeling parameters onto the resulting seismicity cloud and source

parameters. First, we calculate one model with an injection at a shallower depth of only

2.5 km, with correspondingly smaller overall stress magnitudes (Haring et al., 2008) and

with only half the pore pressure. In a second step, we vary the strength of the crust by

adjusting σ1 and the coefficient of friction µ. Input parameters of the three additional

models are shown in Table 1.

We keep the scaling relation between differential stress and b-value, respectively stress

15

Page 15 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

drop, exactly the same as before, even though this dependence was obtained heuristically on

the basis of the Basel observations. Varying differential stresses between the three different

scenarios therefore lead to varying absolute values of average stress drop and b-value. Our

main interest is the relative variation of the source parameters and probabilities with time

and distance. These relative variations can be extracted reliably from the modeling runs,

and some more general postulates about induced seismicity can be derived despite variations

in absolute values. It is unclear at this point whether any scaling between differential stress

and b-value or stress drop would be varying in different geologic situations. However, for

the time being we consider it a realistic assumption since it explains the observation at

Basel quite well.

At shallow depth and for a weak crust, we observe overall much higher b-values above

two (stars and diamonds, respectively, Figure 10) compared to the original model at 4.5 km

depth (circles), for which we obtain a good fit to the observed Basel seismicity (squares).

The strong crust model results in a much lower b-value (inverted triangles). The stronger

the distance dependence of the b-value the worse the fit of the data to a constant average

b-value. This is particularly evident for the strong crust model (inverted triangles). The

distance dependence of the b-value is greatly reduced for the shallow and weak crust models

due to the reduced range of differential stresses (Figure 11). Interestingly, the stress drop

does not show a strong difference in the distance dependence between the different models.

[Figure 10 about here.]

In a next step, we impose a depth-dependent gradient onto the background stress field

for the average crust case and an injection source at 4.5 km depth. Following Haring et al.

16

Page 16 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

(2008) the background stress field changes as follows:

σh(z) = 10MPa

kmz (3)

σ1(z) = 42MPa

kmz ± 10% (4)

σ3(z) = 17MPa

kmz ± 10% (5)

Since we leave the relation between differential stress and b-value the same we would expect

to observe a depth dependence of the b-value on top of the radial dependence from the in-

jection point. Figure 12 shows that this is indeed observed. However, the depth dependence

is a secondary effect to the radial dependence and therefore difficult to observe in practice.

For the synthetic data, we need to correct for the radial dependence first before we see a

variation of b-values from 1.4 near the top of the seismicity cloud to 1.2 for the deepest

events (diamonds in Figure 12a). A similar, weak depth dependence is also imprinted onto

the stress drops (Figure 12b). Figure 12c and d show the depth variation of b-value and

stress drop of the observed Basel data. As expected, the large scatter in the data does not

allow us to observe such rather subtle depth variations in reality. While at least the mean

b-values per depth bin (white squares in Figure 12c) are not incompatible with a depth

variation, the scatter is too large to reliably resolve a depth dependence with statistical

significance. The observed stress drops show no indication of a depth dependence.

[Figure 11 about here.]

[Figure 12 about here.]

In summary, our results suggest that the probability of large events (M≥4) is reduced

if we either drill less deep, or drill into softer formations such as sediments, instead of

17

Page 17 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

granite (Figure 13a and b). In sedimentary hydrofrac stimulations for enhanced oil- or gas

extraction, b-values above 2 have been reported (Maxwell et al., 2009; Wessels et al., 2011).

For the strong crust model, the probability of a M≥4 event is overall higher compared to

the average crust model at 4.5 km depth (Figure 13c). Nevertheless, our heuristic finding

of a relation between b-value and differential stress prevents us from drawing a general

conclusion about the behavior of b-values and magnitude probabilities anywhere else than

Basel on which we calibrated the modeling.

[Figure 13 about here.]

However, the common feature of most modeled scenarios is the observation of the highest

probability of a large magnitude event after well shut-in and at larger distances to the

injection point. To show this more quantitatively, we calculate normalized cumulative

probabilities in time and radial distance to the injection point (Figure 14) for the different

model set-ups. For each case we compare the cumulative probability ratios before and

after the shut-in time as well as closer and further away from a 300 m distance to the

injection point. Individual values are listed in Table 2. Note that the 300 m distance level

is arbitrarily chosen to separate the induced seismicity into two equal groups. For example

we find that our model predicts a 61.5% higher probability of a M≥4 event after the shut-

in time compared to before the shut-in time and a 53.2% higher probability of an event

further away than 300 m if b-value is coupled to differential stress. A constant b-value

always predicts higher probabilities before the shut-in time and closer than 300 m to the

injection point compared to the model where b-value is linearly linked to differential stress.

Whereas the distance dependence is reduced for the shallow crust model, we also find that

the weak and strong crust models predict higher probabilities of a M≥4 event after shut-in

18

Page 18 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

and at larger distances. Two important consequences arise from this observation: (i) it

is insufficient to shut-in an injection operation upon observation of a threshold magnitude

event (so-called traffic-light system), since the largest magnitude probability is yet to come

after the shut-in, irrespective of the applied threshold magnitude. (ii) increased probabilities

at larger distances from the well means that a much larger area around the injection well

is potentially affected by increased seismic hazard.

[Figure 14 about here.]

Our study suggests that a-priori information about the stresses in the area of interest

are important for some initial estimation of the expected seismic hazard. Careful pre-

operation site characterization of the underground to image possible fault zones capable of

larger-magnitude seismicity can impose additional constraints on the maximum expectable

magnitude in or near the stimulation volume. Furthermore, the implementation of a near

real-time probabilistic seismic hazard prediction model can be used to replace previously

used traffic light systems. Our model can also be used to investigate the effect on the

induced seismicity and the respective seismic hazard using varying stimulation approaches

such as a different injection flow rate or a different wellhead pressure.

CONCLUSIONS

We propose a simple geomechanical model for induced seismicity that links differential

stress to b-value and stress drop. With this model, we can forward-model the induced

seismicity response to a hydraulic injection in space and time. In addition, we can forward

model source parameters such as stress drop and magnitude of events in a stochastic way.

The latter allows us to estimate the seismic hazard associated with an injection operation

19

Page 19 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

by calculating the probability of exceeding a certain magnitude event. Our model can

explain the observation of increasing stress drops and decreasing b-values as a function of

radial distance from the injection well. The probability of exceeding a certain magnitude

event is larger after well shut-in, and reaches further out from the injection point. The

overall procedure can be coupled with incoming seismicity data from an ongoing stimulation

operation and thus allow to estimate the seismic hazard of an injection operation in real

time. Furthermore, the model can be used to estimate the seismic response, or ”fracability”

of a medium before the stimulation such that critical parameters (i.e. injection volume,

time, and pressure) can be planned and optimized for particular projects.

ACKNOWLEDGMENTS

We are grateful to GeoPower Basel for permission to publish these data. This research is

supported by the project GEOTHERM that is funded by the ETH Competence Centre for

Environment and Sustainability. This work is also supported by the European Commission

through funds provided within the FP7 project GEISER, grant agreement no. 241321-2.

We wish to thank Alexander Goertz, Valentin Gischig, and Corinne Bachmann for fruitful

discussions that helped to improve the manuscript.

APPENDIX A

CONSTRAINTS ON THE SEED DENSITY

We assume that every seed point in our volume represents a disk-shaped fracture of random

orientation with a radius that depends on the assigned magnitude. Magnitudes are ran-

domly assigned, but follow a Gutenberg-Richter distribution with a prescribed b-value. The

20

Page 20 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

higher the seed density (seed points per volume), the more are the fractures intersecting

each other, up to the point where intersecting fractures are completely separating individual

parts of the rock frame from each other. The fracture porosity at which this is reached is

termed critical porosity φc (Nur et al., 1998); it is the point beyond which the fractured

material would fall apart. Obviously, we have to choose the seed density such that the

fracture porosity stays below this critical value.

Assuming circular ruptures, we can define a moment-dependent fault radius (Eshelby,

1957),

r =

(

7

16

M0

∆σ

)1/3

, (A-1)

with M0 the seismic moment, and ∆σ the stress drop, which we will assume constant for

simplicity in this exercise. We know that log10 M0 ∝ 1.5Mw, and hence log10 r ∝ 0.5Mw,

with the Mw distribution following a Gutenberg-Richter statistic above the magnitude of

completeness. Consequently, the radius r will follow a similar distribution, and we have to

find a meaningful average radius r over this distribution.

For disk-shaped fractures, the crack density can be given as (Kachanov, 1992)

ρ =N

V0r3 , (A-2)

where V0 denotes a reference volume, r is the average radius of a circular rupture for a given

event distribution, and N is the overall number of cracks (seeds) in the volume. In order

to define a porosity to a medium cracked in such a manner, we need to assign an opening

width or aperture d to these cracks. For an ellipsoidal shape of the crack, the porosity can

then be stated as (Saenger et al., 2004),

φ =4N

3V0πr2d . (A-3)

21

Page 21 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

[Figure 15 about here.]

To define a lower critical porosity limit for such a medium, we consider a simplistic thought

experiment with uniformly oriented and regularly distributed fracture sets (Figure 15). We

attempt to fill a cube of side length n · a with a network of non-intersecting fractures with

diameter 2r, spaced a distance a from each other. In this manner, we can fill the cube with

1/2 n cracks in the direction of the two long axes of the crack, and n cracks in the direction

of the aperture d � r, such that to fill the whole cube, it requires 1/4 n3 cracks. Assume

now that the medium is intersected by three mutually perpendicular fracture networks with

a total of 3/4 n3 cracks. In order for these fracture networks to fully intersect each other,

the crack radius r needs to be such that it describes the circumscribed circle of a square of

side length a,

a = r√

2 . (A-4)

Using equation A-3, we can now state the porosity for this case as

φc =1√2πα . (A-5)

where α denotes the crack aspect ratio d/r. We assume that for more complicated situa-

tions, e.g. randomly oriented fractures, the critical prorosity would be higher.Combining

equations A-2 and A-3, we can define an upper bound for the crack density ρ, which we

equal to the seed density in our modeling,

ρ ≤3

4√

2≈ 0.53 , (A-6)

and with that define an upper bound for the number of seeds in our modeling as

N ≤0.53V0

r3. (A-7)

22

Page 22 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

To give and example, for an average rupture radius r of 20 m, which corresponds very

roughly to assuming a magnitude of completeness (mode of the magnitude distribution) of

1, we could fill our 1km3 modeling volume with a maximum of 66,250 seeds. Using 30,000

seeds, as we have done in our modeling, fulfills criterion A-6, and stays below this simple

critical porosity limit.

[Table 1 about here.]

[Table 2 about here.]

23

Page 23 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

REFERENCES

Abercrombie, R., 1995, Earthquake source scaling relationships from −1 to 5 ML using

seismograms recorded at 2.5 km depth: J. Geophys. Res., 100, 24,015 – 24,036.

Allmann, B., and P. Shearer, 2009, Global variations of stress drop for moderate to large

earthquakes: J. Geophys. Res., 114.

Amitrano, D., 2003, Brittle-ductile transition and associated seismicity: Experimental and

numerical studies and relationship with the b value: J. Geophys. Res., 108.

Bachmann, C., S. Wiemer, B. Goertz-Allmann, and J. Woessner, 2012, Influence of pore

pressure on the size distribution of induced earthquakes: submitted to Geophys. Res.

Lett.

Bruel, D., 2007, Using the migration of the induced seismicity as a constraint for fractured

Hot Dry Rock reservoir modelling: Int. J. Rock Mech. & Min. Sciences, 44, 1106 – 1117.

Charlety, J., N. Cuenot, L. Dorbath, C. Dorbath, H. Haessler, and M. Frogneux, 2007, Large

earthquakes during hydraulic stimulations at the geothermal site of Soultz-sous-Forets:

Int. J. Rock Mech. Min. Sciences, 44, 1091–1105.

Deichmann, N., and D. Giardini, 2009, Earthquakes induced by the stimulation of an en-

hanced geothermal system below Basel (Switzerland): Seismol. Res. Lett., 80.

Dinske, C., S. Shapiro, and M. O. Haring, 2010, Interpretation of microseismicity induced

by time-dependent injection pressure: SEG Expanded Abstracts, 2125 – 2129.

Eshelby, J., 1957, The Determination of the Elastic Field of an Ellipsoidal Inclusion, and

Related Problems: Proc. R. Soc. London, Series A, 241, 376 – 396.

Evans, K., A. Zappone, T. Kraft, N. Deichmann, and F. Moia, 2012, A survey of the

induced seismic response to fluid injection in geothermal and CO2 reservoirs in Europe:

Geothermics, 41, 30 – 54.

24

Page 24 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Goertz-Allmann, B., A. Goertz, and S. Wiemer, 2011, Stress drop variations of induced

earthquakes at the Basel geothermal site: Geophys. Res. Lett., 38.

Gutenberg, B., and C. Richter, 1944, Frequency of earthquakes in California: Bull. Seismol.

Soc. Am., 34, 185 – 188.

Haring, M., U. Schanz, F. Ladner, and B. Dyer, 2008, Characterisation of the Basel 1

enhanced geothermal system: Geothermics, 37, 469 – 495.

Hummel, N., and T. Muller, 2009, Microseismic signatures of non-linear pore-fluid pressure

diffusion: Geophys. J. Int., 179, 1558 – 1565.

Kachanov, M., 1992, Effective elastic properties of cracked solids: critical review of some

basic concepts: Applied Mechanical Review, 45, 304 – 335.

Kanamori, H., and E. Brodsky, 2004, The physics of earthquakes: Rep. Prog. Phys., 67,

1429 – 1496.

Kohl, T., and T. Megel, 2007, Predictive modeling of reservoir response to hydraulic stimu-

lations at the European EGS site Soultz-sous-Forets: Int. J. Rock Mech. & Min. Sciences,

44, 1118 – 1131.

Kummerow, J., S. Shapiro, H. Asanuma, and M. Haring, 2011, Application of an arrival

time and cross corelation value-based location algorithm to the Basel 1 microseismic data:

Presented at the 73th EAGE Conference & Exhibition incorporating SPE EUROPEC.

Kwiatek, G., K. Plenkers, M. Nakatani, Y. Yabe, G. Dresen, and JAGUARS-Group,

2010, Frequency-magnitude characteristics down to magnitude -4.4 for induced seismicity

recorded at Mponeng gold mine, South Africa: Bull. Seismol. Soc. Am., 100, 1165–1173.

Langenbruch, C., and S. A. Shapiro, 2010, Decay rate of fluid-induced seismicity after

termination of reservoir stimulations: Geophysics, 75, MA53–MA62.

——–, 2011, Geomechanical interpretation of pore pressures triggering microseismicity: in

25

Page 25 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

EAGE Expanded Abstracts, 73.

Maxwell, S., M. Jones, R. Parker, S. Miong, S. Leaney, D. Dorval, D. D’Amico, J.Logel,

E. Anderson, and K. Hammermaster, 2009, Fault activation during hydraulic fracturing:

SEG Expanded Abstracts, 1552 – 1556.

Mayeda, K., and W. R. Walter, 1996, Moment, energy, stress drop, and source spectra of

western United States earthquakes from regional coda envelope: J. Geophys. Res., 101,

11,195 – 11,208.

Nur, A., G. Mavko, J. Dvorkin, and D. Galmudi, 1998, Critical porosity: A key to relating

physical properties to porosity in rocks: The Leading Edge, 17, 357 – 362.

Pearson, C., 1981, The relationship between microseismicity and high pore pressures during

hydraulic stimulation experiments in low permeability granitic rocks: J. Geophys. Res.,

86, 7855 – 7864.

Rothert, E., and S. Shapiro, 2003, Microseismic monitoring of borehole fluid injections:

Data modeling and inversion for hydraulic properties of rocks: Geophysics, 68, 685 –

689.

Saenger, E., O. Kruger, and S. Shapiro, 2004, Effective elastic properties of randomly

fractured soils: 3d numerical experiments: Geophys. Prospecting, 52, 183 – 195.

Schoenball, M., T. Muller, B. Muller, and O. Heidbach, 2009, Fluid-induced microseismicity

in pre-stressed rock masses: Geophys. J. Int.

Schorlemmer, D., S. Wiemer, and M. Wyss, 2005, Variations in earthquake-size distribution

across different stress regimes: Nature, 437.

Shapiro, S., E. Rothert, V. Rath, and J. Rindschwentner, 2002, Characterization of fluid

transport properties of reservoirs using induced microseismicity: Geophysics, 67, 212 –

220.

26

Page 26 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Shapiro, S. A., E. Huenges, and G. Borm, 1997, Estimating the crust permeability from

fluid-injection-induced seismic emission at the KTB site: Geophys. J. Int., 131, F15–F18.

Talwani, P., and S. Acree, 1984, Pore pressure diffusion and the mechanism of reservoir-

induced seismicity: Pure and Applied Geophysics, 122, 947–965.

Wang, H. F., 2000, Theory of linear poroelasticity with applications to geomechanics and

hydrogeology: Princeton Univ. Press.

Wessels, S., M. Kratz, and A. D. L. Pena, 2011, Identifying fault activation during hydraulic

stimulation in the barnett shale: source mechanisms, b values, and energy release analysis

of microseismicity.: SEG Expanded Abstracts, 1463 – 1467.

Wiemer, S., 2000, Introducing probabilistic aftershock hazard mapping: Geophys. Res.

Lett., 27, 3405–3408.

Zoback, M. D., 2007, Reservoir geomechanics: Cambridge University Press.

27

Page 27 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

LIST OF FIGURES

1 a) Example stress distribution for an average crust at 4.5 km depth at onepoint in the medium depicted by the Mohr diagram (shear stress τs versusnormal stress σn). The bold dashed line shows the failure envelope. Dis-tribution of stress values around the mean σ3 and σ1 are indicated by thedotted line (planned) and the histogram (actual). The vertical dashed lineshows the strong crust upper limit for σ1. ph denotes the hydrostatic porepressure. b) The black Mohr circle shows the original stress distribution atone point in the medium. The gray circle shows the effective stress of an in-duced event with the depicted critical pore pressure. The black dotted circleshows the post-event stress regime (decrease of σ1 only) due to the indicatedevent stress drop ∆σ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2 Time-distance plot of the induced seismicity. The color denotes the numberof repeats of an event. Note that the colorscale is clipped at 5 repeats. Thedashed line shows the shut-in time. The solid contour lines show isobars inMPa. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Calibration of magnitudes between the modeled seismicity and the Baselobserved data. Modulus of the overall b-value difference between the Baseldata and the model versus possible upper b-value limits (bmax). The circlesshow the mean value from 50 model runs with standard deviation. The solidline shows a quadratic fit with its minimum determined at bmax=4 (dashedline). The respective best fitting scaling relation is shown. The inset shows asketch of the underlying b-value versus differential stress σd scaling relationsusing different upper b-value limits. The lower b-value limit is fixed to oneat σd = 136. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Time-distance plot of the induced seismicity. The color denotes a) b-valueand b) smoothed log stress drop for one model run. The dashed line denotesthe shut-in time. Stars mark all events with M≥2.5. c) Observed Baselseismicity for all events above the magnitude of completeness (Mc = 0.9).Color denotes the b-values estimated at a certain event location by Bachmannet al. (2012). Note that the colorbar is clipped at both ends for a bettercomparison with the modeled seismicity. . . . . . . . . . . . . . . . . . . . . 34

5 Spatial b-value (upper panels) and smoothed stress drop (lower panels) dis-tributions of induced events in a) three-dimensional view for one model run,b) cross-section for one model run, c) grid-stack of all cross-sections over100 model runs, and d) grid-stack of all cross-sections over 100 model runsusing a constant input b-value and stress drop as described in the text. Thecross-sections include all events within 100 m to a plane through the injectionpoint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

28

Page 28 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

6 a) b-value, and b) stress drop versus distance from the injection point for onemodel run. The squares in a) and b) show the mean values over constantlog distance bins with respective standard errors from bootstrap resampling.c) Mean b-value (crosses), and d) mean stress drop (crosses) versus distancefrom the injection point from 100 model runs. The squares in c) and d) showthe mean of the mean values from 100 model runs with standard deviation. 36

7 a) Differential stresses of all induced events (colored dots) plotted at theminimum distance of the Mohr circle to the failure envelope (red dashedline). The maximum and minimum differential stresses are indicated by theblack circles. The inset shows differential stress versus pore pressure. b)Cross-section through injection point of stacked differential stress from 100model runs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

8 Stress drop versus magnitude for a) the original model were stress drop islinked to differential stress and a 5% scatter is included, b) the same than a)but without the scatter, c) a constant stress drop of 2.5±1.25 MPa and d) thecomparison of the mean values of a) to c). Upper row shows raw stress dropsand bottom row spatially smoothed stress drops. The bold symbols show themean value within 0.2 magnitude bins with respective standard error frombootstrap resampling using 1000 realizations. . . . . . . . . . . . . . . . . . 38

9 Probability of an event exceeding a magnitude a) M3, b) M4, and c) M5to occur at a certain time (top row) and distance from the injection point(bottom row). Error bars show the standard deviation computed from 100model runs. The dashed line marks the shut-in time in time, and the locationof the largest observed Basel event in distance. The different colors denotethe model were the b-value is linked to differential stress (white), and themodel with a constant input b-value (grey). . . . . . . . . . . . . . . . . . . 39

10 Gutenberg-Richter relations of one model run for different model set-upscompared to the observed Basel data (squares). The circles show the averagecrust model at 4.5 km depth. The diamonds show the weak crust model.The inverted triangles show the strong crust model, and the stars show theshallow crust model. See Table 1 for input parameters. The magnitude ofcompleteness Mc is marked. . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

11 a) b-value, and b) stress drop versus distance from the injection point forvarious model set-ups. The symbols show the mean over constant log dis-tance bins of the mean values from 100 model runs with respective standarddeviations for the average crust model at 4.5 km depth (circle), the shallowcrust model at 2.5 km depth (stars), the weak crust model (diamonds), andthe strong crust model (inverted triangles). . . . . . . . . . . . . . . . . . . 41

29

Page 29 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

12 a) b-value and b) smoothed stress drop versus focal depth for the modeledseismicity (average crust, injection at 4.5 km). Circles show the data (meanfrom 100 model runs with standard deviation) without correction for thedistance to the injection point. Diamonds show the same data after correctionfor the radial distance dependence. c) Observed b-value versus focal depth forthe Basel seismicity estimated by Bachmann et al. (2012). d) Observed stressdrop versus focal depth for the Basel seismicity estimated by Goertz-Allmannet al. (2011). The bold squares in c) and d) show the mean value within0.1 km depth bins with respective standard error from bootstrap resamplingusing 1000 realizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

13 Probability of an event exceeding a magnitude M 4 for a) a shallow crustmodel at 2.5 km depth b) a weak crust model, and for c) a strong crustmodel to occur at a certain time (top row) and distance from the injectionpoint (bottom row). Error bars show the standard deviation computed from100 model runs. The dashed line marks the shut-in time in time, and thelocation of the largest observed Basel event in distance. . . . . . . . . . . . 43

14 Normalized cumulative probability of an event exceeding a certain magnitudeversus time (a and c) and distance to the injection point (b and d). a) and b)Comparison of an average crust model at 4.5 km depth for varying b-value(black curves) and constant b-value (grey curves). c) and d) Comparison ofan average crust at 4.5 km depth (solid black), shallow crust (dashed grey),weak crust (dotted grey), and strong crust model (solid grey). The verticallines marks the shut-in time (a and c) or the 300 m distance (b and d). . . 44

15 Sketch to illustrate the simplest deterministic case which we use to estimatea lower limit of critical fracture porosity. . . . . . . . . . . . . . . . . . . . . 45

30

Page 30 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review0 50 100 150 200

0

20

40

60

80

τmax

σ1 − p

hσ3 − p

h

τs=0.85σ +7

n

τ s [M

Pa]

σn [MPa]

0 50 100 150 2000

20

40

60

80

pp=8.31 MPa

Δσ=2.1 MPa

τ s [M

Pa]

σn [MPa]

a)

b)

0

40

80

120

160

200

counts

Figure 1: a) Example stress distribution for an average crust at 4.5 km depth at one pointin the medium depicted by the Mohr diagram (shear stress τs versus normal stress σn).The bold dashed line shows the failure envelope. Distribution of stress values around themean σ3 and σ1 are indicated by the dotted line (planned) and the histogram (actual). Thevertical dashed line shows the strong crust upper limit for σ1. ph denotes the hydrostaticpore pressure. b) The black Mohr circle shows the original stress distribution at one pointin the medium. The gray circle shows the effective stress of an induced event with thedepicted critical pore pressure. The black dotted circle shows the post-event stress regime(decrease of σ1 only) due to the indicated event stress drop ∆σ.

31

Page 31 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

0 1 2 3 4 5 6 7 8

x 105

0

200

400

600

800

1000

time [s]

dis

tan

ce [

m]

0.0

02

0.002

0.002

0.002

0.0

1

0.01

0.01

0.01

0.1

0.1

0.1

0.1

1

1

1 1

10

10

1001001000

# repeats1 2 3 4 5+

Figure 2: Time-distance plot of the induced seismicity. The color denotes the number ofrepeats of an event. Note that the colorscale is clipped at 5 repeats. The dashed line showsthe shut-in time. The solid contour lines show isobars in MPa.

32

Page 32 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.05

0.1

0.15

0.2

0.25

0.3

0.35

| bobse

rved-b

model |

bmax

1

136

2

5.5

0diff. stress σd [MPa]

b-v

alu

ebest fit: b=-0.022σd+4

bmin

}bmax

Figure 3: Calibration of magnitudes between the modeled seismicity and the Basel observeddata. Modulus of the overall b-value difference between the Basel data and the model versuspossible upper b-value limits (bmax). The circles show the mean value from 50 model runswith standard deviation. The solid line shows a quadratic fit with its minimum determinedat bmax=4 (dashed line). The respective best fitting scaling relation is shown. The insetshows a sketch of the underlying b-value versus differential stress σd scaling relations usingdifferent upper b-value limits. The lower b-value limit is fixed to one at σd = 136.

33

Page 33 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review0 1 2 3 4 5 6 7 8

x 105

0

0.2

0.4

0.6

0.8

1

# ev:964

time [sec]

dis

tance [

km

]

b−value

1.2 1.4 1.6 1.8 2 2.2

0 1 2 3 4 5 6 7 8

x 105

0

0.2

0.4

0.6

0.8

1

# ev:964

time [sec]

dis

tance [

km

]

log10 stress drop [MPa]

0.2 0.3 0.4 0.5

0 1 2 3 4 5 6 7 8

x 105

0

0.2

0.4

0.6

0.8

1

# ev:988

time [sec]

dis

tance [

km

]

b−value

1.2 1.4 1.6 1.8 2 2.2

Figure 4: Time-distance plot of the induced seismicity. The color denotes a) b-value andb) smoothed log stress drop for one model run. The dashed line denotes the shut-in time.Stars mark all events with M≥2.5. c) Observed Basel seismicity for all events above themagnitude of completeness (Mc = 0.9). Color denotes the b-values estimated at a certainevent location by Bachmann et al. (2012). Note that the colorbar is clipped at both endsfor a better comparison with the modeled seismicity.

34

Page 34 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

−0.40

0.4

−0.40

0.4

0

0.4

x [km]y [km]

z [

km

]

# ev:910

−0.2 0 0.2 0.4−0.4

−0.2

0

0.2

0.4 # ev:549

x [km]

y [

km

]

b−value1.2 1.6 2

−0.40

0.4

−0.40

0.4

−0.4

0

0.4

x [km]y [km]

z [

km

]

−0.2 0 0.2 0.4−0.4

−0.2

0

0.2

0.4

x [km]

y [

km

]

log10 stress drop [MPa]0.2 0.3 0.4 0.5

−0.4

−0.2

0

0.2

0.4

y [

km

]

−0.4 −0.2 0 0.2 0.4

x [km]

−0.4

−0.2

0

0.2

0.4

y [

km

]

−0.4 −0.2 0 0.2 0.4

x [km]

1.44 1.48 1.52

0.37 0.42

−0.4 −0.2 0 0.2 0.4

−0.4

−0.2

0

0.2

0.4

x [km]

y[k

m]

−0.4 −0.2 0 0.2 0.4

−0.4

−0.2

0

0.2

0.4

x [km]

y[k

m]

b−value

log10 stress drop [MPa]

d)

log10 stress drop [MPa]0.2 0.3 0.4 0.5

log10 stress drop [MPa]0.2 0.3 0.4 0.5

b−value1.2 1.6 2

b−value1.2 1.6 2

a) d)b) c)

Figure 5: Spatial b-value (upper panels) and smoothed stress drop (lower panels) distri-butions of induced events in a) three-dimensional view for one model run, b) cross-sectionfor one model run, c) grid-stack of all cross-sections over 100 model runs, and d) grid-stackof all cross-sections over 100 model runs using a constant input b-value and stress drop asdescribed in the text. The cross-sections include all events within 100 m to a plane throughthe injection point.

35

Page 35 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review10

110

210

31

1.5

2

2.5

distance [m]

b−

valu

e

a)

101

102

103

100.1

100.3

100.5

distance [m]

smo

oth

ed

str

ess

dro

p [

MP

a] b)

101

102

103

1

1.5

2

2.5

b−

valu

e

distance [m]

101

102

103

100.1

100.3

100.5

smo

oth

ed

str

ess

dro

p [

MP

a]

distance [m]

c)

d)

Figure 6: a) b-value, and b) stress drop versus distance from the injection point for onemodel run. The squares in a) and b) show the mean values over constant log distance binswith respective standard errors from bootstrap resampling. c) Mean b-value (crosses), andd) mean stress drop (crosses) versus distance from the injection point from 100 model runs.The squares in c) and d) show the mean of the mean values from 100 model runs withstandard deviation.

36

Page 36 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review0 20 40 60 80 100 120 140 1600

20

40

60

80

σn [MPa]

τ s [MPa]

σd [MPa]

50

100

150

0 10 20

50

100

150

σd

[M

Pa

]

pore pressure [MPa]x−direction [km]

y−direction [km]

−0.5 0 0.5

−0.4

−0.2

0

0.2

0.4

σd [MPa]70 80 90 100 110 120 130 140

a) b)

Figure 7: a) Differential stresses of all induced events (colored dots) plotted at the minimumdistance of the Mohr circle to the failure envelope (red dashed line). The maximum andminimum differential stresses are indicated by the black circles. The inset shows differentialstress versus pore pressure. b) Cross-section through injection point of stacked differentialstress from 100 model runs.

37

Page 37 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

0

0.2

0.4

0.6

0.8

1

log

10

∆σ

[MPa]

1 1.5 2 2.5 3 3.50

0.2

0.4

0.6

0.8

1

log

10

∆σ

(sm

ooth) [MPa]

magnitude1 1.5 2 2.5 3 3.5

magnitude1 1.5 2 2.5 3 3.5

magnitude1 1.5 2 2.5 3 3.5

magnitude

varying stress drop

no additional scatter

constant stress drop

a) b) c) d)

Figure 8: Stress drop versus magnitude for a) the original model were stress drop is linked todifferential stress and a 5% scatter is included, b) the same than a) but without the scatter,c) a constant stress drop of 2.5±1.25 MPa and d) the comparison of the mean values of a)to c). Upper row shows raw stress drops and bottom row spatially smoothed stress drops.The bold symbols show the mean value within 0.2 magnitude bins with respective standarderror from bootstrap resampling using 1000 realizations.

38

Page 38 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

0 1 2 3 4 5 6 7 8x 105

0

0.05

0.1

0.15

0.2

0.25

pro

babil

ity

time [s]

M > 3

0 100 200 300 400 500 6000

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

pro

babil

ity

distance [m]

0 1 2 3 4 5 6 7 8x 105

0

0.005

0.01

0.015

0.02

0.025

pro

babil

ity

time [s]

M > 4

0 100 200 300 400 500 6000

0.005

0.01

0.015

0.02

0.025

0.03

0.035

pro

babil

ity

distance [m]

0 1 2 3 4 5 6 7 8x 10

5

0

0.5

1

1.5

2

2.5

3x 10−3

pro

babil

ity

time [s]

M > 5

0 100 200 300 400 500 6000

0.5

1

1.5

2

2.5

3 x 10−3

pro

babil

ity

distance [m]

shut-in

Basel ML3.4

varying bconstant b

varying bconstant b

varying bconstant b

a) b) c)

Figure 9: Probability of an event exceeding a magnitude a) M3, b) M4, and c) M5 to occurat a certain time (top row) and distance from the injection point (bottom row). Error barsshow the standard deviation computed from 100 model runs. The dashed line marks theshut-in time in time, and the location of the largest observed Basel event in distance. Thedifferent colors denote the model were the b-value is linked to differential stress (white),and the model with a constant input b-value (grey).

39

Page 39 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review1 1.5 2 2.5 3 3.5 4

0

0.5

1

1.5

2

2.5

3

3.5

magnitude M

log

(N)

observed: log(N)=4.3-1.48 M

avgerage: log(N)=4.36-1.55 M

weak crust: log(N)=4.96-2.24 M

strong crust: log(N)=3.88-0.91 M

shallow crust: log(N)=4.92-2.43 M

Mc

Figure 10: Gutenberg-Richter relations of one model run for different model set-ups com-pared to the observed Basel data (squares). The circles show the average crust model at4.5 km depth. The diamonds show the weak crust model. The inverted triangles showthe strong crust model, and the stars show the shallow crust model. See Table 1 for inputparameters. The magnitude of completeness Mc is marked.

40

Page 40 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

101

102

103

100

stre

ss d

rop [

MP

a]

distance [m]

average at 4.5 km

shallow

weak

strong

101

102

103

0.5

1

1.5

2

2.5

3

3.5

b−

valu

e

distance [m]

a)

b)

Figure 11: a) b-value, and b) stress drop versus distance from the injection point for variousmodel set-ups. The symbols show the mean over constant log distance bins of the meanvalues from 100 model runs with respective standard deviations for the average crust modelat 4.5 km depth (circle), the shallow crust model at 2.5 km depth (stars), the weak crustmodel (diamonds), and the strong crust model (inverted triangles).

41

Page 41 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review4 4.2 4.4 4.6 4.8 51

1.2

1.4

1.6

1.8

2

b−value

depth [km]3.5 4 4.5 5

0.6

0.8

1

1.2

1.4

1.6

observed b−value

depth [km]

3.5 4 4.5 510

−1

100

101

102

observed stress drop [MPa]

depth [km]

a)

b)

c)

d)

with no distance correction

with distance correction

4 4.2 4.4 4.6 4.8 52

2.5

3

3.5

4

smoothed stress drop [MPa]

depth [km]

Figure 12: a) b-value and b) smoothed stress drop versus focal depth for the modeledseismicity (average crust, injection at 4.5 km). Circles show the data (mean from 100model runs with standard deviation) without correction for the distance to the injectionpoint. Diamonds show the same data after correction for the radial distance dependence. c)Observed b-value versus focal depth for the Basel seismicity estimated by Bachmann et al.(2012). d) Observed stress drop versus focal depth for the Basel seismicity estimated byGoertz-Allmann et al. (2011). The bold squares in c) and d) show the mean value within0.1 km depth bins with respective standard error from bootstrap resampling using 1000realizations.

42

Page 42 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

0 1 2 3 4 5 6 7 8x 10

5

0

1

2x 10

−4

Pro

babil

ity

time [s]

weak crust

0 100 200 300 400 500 6000

1

2x 10

−4

Pro

babil

ity

distance [m]

0 1 2 3 4 5 6 7 8x 10

5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Pro

babil

ity

time [s]

strong crust

0 100 200 300 400 500 6000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Pro

babil

ity

distance [m]

0 1 2 3 4 5 6 7 8x 10

5

0

0.5

1

1.5

2

2.5

3

3.5

4x 10

−5

Pro

babil

ity

time [s]

shallow crust

0 100 200 300 400 500 6000

0.5

1

1.5

2

2.5

3

3.5

4x 10

−5

Pro

babil

ity

distance [m]

a) b) c)

shut-in

Basel ML3.4

Figure 13: Probability of an event exceeding a magnitude M 4 for a) a shallow crust modelat 2.5 km depth b) a weak crust model, and for c) a strong crust model to occur at a certaintime (top row) and distance from the injection point (bottom row). Error bars show thestandard deviation computed from 100 model runs. The dashed line marks the shut-in timein time, and the location of the largest observed Basel event in distance.

43

Page 43 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review0 1 2 3 4 5 6 7 8

x 105

0

0.2

0.4

0.6

0.8

1

no

rmali

zed

cu

mu

lati

ve p

rob

abil

ity

time [sec]

M≥3

M≥4 varying b

M≥5

M≥3

M≥4 constant b

M≥5

}}

shut-in

0 1 2 3 4 5 6 7 8

x 105

0

0.2

0.4

0.6

0.8

1

no

rmali

zed

cu

mu

,ati

ve p

rob

abil

ity

time [sec]

M≥4: average

M≥4: shallow

M≥4: weak

M≥4: strong

shut-in

c)a)

0 100 200 300 400 500 6000

0.2

0.4

0.6

0.8

1

no

rmali

zed

cu

mu

lati

ve p

rob

abil

ity

distance [m]

M≥3

M≥4 varying b

M≥5

M≥3

M≥4 constant b

M≥5

}}

0 100 200 300 400 500 6000

0.2

0.4

0.6

0.8

1

no

rmali

zed

cu

mu

lati

ve p

rob

abil

ity

distance [m]

M≥4: average

M≥4: shallow

M≥4: weak

M≥4: strong

d)b)

Figure 14: Normalized cumulative probability of an event exceeding a certain magnitudeversus time (a and c) and distance to the injection point (b and d). a) and b) Comparisonof an average crust model at 4.5 km depth for varying b-value (black curves) and constantb-value (grey curves). c) and d) Comparison of an average crust at 4.5 km depth (solidblack), shallow crust (dashed grey), weak crust (dotted grey), and strong crust model (solidgrey). The vertical lines marks the shut-in time (a and c) or the 300 m distance (b and d).

44

Page 44 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

na

a

a2r

Figure 15: Sketch to illustrate the simplest deterministic case which we use to estimate alower limit of critical fracture porosity.

45

Page 45 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

LIST OF TABLES

1 Mean medium parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2 Comparison of normalized cumulative probabilities pcum of exceeding a cer-tain magnitude event in time and distance to the injection point for variousmodel set-ups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

46

Page 46 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Table 1: Mean medium parameters.Model avg. crust avg. crust weak crust strong crust

Depth 4.5 km 2.5 km 4.5 km 4.5 km

σ3 [MPa] 75 42 75 75

σ1 [MPa] 185 105 147 232

max σ1 [MPa] 232 129 232 232

std. dev. (σ1, σ3) 10 %

ph [MPa] 45 25 45 45

µ 0.85 0.85 0.6 1.0

No. of seeds 30,000

Cohesion [MPa] 7

a0 [m] 70

p0 [MPa] 9.5

pt [Pa / s] 48

t0 [s] 4.32 · 105

D[m2/s] 0.05

47

Page 47 of 48 GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review

Table 2: Comparison of normalized cumulative probabilities pcum of exceeding a certainmagnitude event in time and distance to the injection point for various model set-ups.magnitude M ≥ 3 ≥ 4 ≥ 5 ≥ 3 ≥ 4 ≥ 5 ≥ 4

cumulativeprobability [%] varying b constant b 2.5 km weak strong

before shut-in 55.3 38.5 23.2 75.5 73 68.4 65.3 30.8 45.2

after shut-in 44.7 61.5 76.8 24.5 27 31.6 34.7 69.2 54.8

≤ 300 m 54.7 46.8 39.4 72.5 70 64.3 60.1 41.6 50.1

≥ 300 m 45.3 53.2 60.6 27.5 30 35.7 39.9 58.4 49.9

48

Page 48 of 48GEOPHYSICS

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

AC

Ka

b

c

a

ARAA

KGILC

1

ctrcftrt(fl

rG

0d

Geothermics 41 (2012) 30– 54

Contents lists available at SciVerse ScienceDirect

Geothermics

j o ur nal homep age : www.elsev ier .com/ locate /geothermics

survey of the induced seismic responses to fluid injection in geothermal andO2 reservoirs in Europe

eith F. Evansa,∗, Alba Zapponeb, Toni Kraftb, Nicolas Deichmannb, Fabio Moiac

Engineering Geology, Swiss Federal Institute of Technology (ETH), Sonneggstrasse 5, 8092 Zürich, SwitzerlandSwiss Seismological Service, Sonneggstrasse 5, 8092 Zürich, SwitzerlandRSE-Ricerca Sistema Energetico S.p.A., Milan, Italy

r t i c l e i n f o

rticle history:eceived 17 February 2010ccepted 13 August 2011vailable online 21 October 2011

eywords:eothermal

njection-induced seismicityarge-magnitude events (LME)O2 sequestration

a b s t r a c t

The paper documents 41 European case histories that describe the seismogenic response of crystallineand sedimentary rocks to fluid injection. It is part of an on-going study to identify factors that havea bearing on the seismic hazard associated with fluid injection. The data generally support the viewthat injection in sedimentary rocks tends to be less seismogenic than in crystalline rocks. In both cases,the presence of faults near the wells that allow pressures to penetrate significant distances verticallyand laterally can be expected to increase the risk of producing felt events. All cases of injection intocrystalline rocks produce seismic events, albeit usually of non-damaging magnitudes, and all crystallinerock masses were found to be critically stressed, regardless of the strength of their seismogenic responsesto injection. Thus, these data suggest that criticality of stress, whilst a necessary condition for producingearthquakes that would disturb (or be felt by) the local population, is not a sufficient condition. Thedata considered here are not fully consistent with the concept that injection into deeper crystallineformations tends to produce larger magnitude events. The data are too few to evaluate the combined

effect of depth and injected fluid volume on the size of the largest events. Injection at sites with lownatural seismicity, defined by the expectation that the local peak ground acceleration has less than a 10%chance of exceeding 0.07 g in 50 years, has not produced felt events. Although the database is limited,this suggests that low natural seismicity, corresponding to hazard levels at or below 0.07 g, may be auseful indicator of a low propensity for fluid injection to produce felt or damaging events. However,higher values do not necessarily imply a high propensity.

. Introduction

Induced seismicity is recognised as a possible hazard in practi-ally all engineering endeavours where stress or pore pressure inhe subsurface is altered. This can be taken as a reflection of theealization that has dawned in the past 20 years that the Earth’srust generally supports high shear stress levels and is often close toailure. Historically, the most damaging events, which have some-imes caused fatalities, are associated with the impoundment ofeservoirs (Gupta, 1992). However, earthquakes of a size sufficiento cause damage have also been associated with mining activityGibowicz, 1990), long-term fluid withdrawal (Segall, 1989) anduid injection (Nicholson and Wesson, 1990).

Given that massive fluid injections to stimulate crystallineocks have routinely been performed during Engineered/Enhancedeothermal System (EGS) projects – formerly called Hot Dry Rock

∗ Corresponding author. Tel.: +41 44 633 2521; fax: +41 44 633 1108.E-mail address: keith.evans@erdw.ethz.ch (K.F. Evans).

375-6505/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.oi:10.1016/j.geothermics.2011.08.002

© 2011 Elsevier Ltd. All rights reserved.

(HDR) projects – since the early 1970s, it is perhaps surprisingthat the issue of seismic hazard associated with these operationshas only recently come to the fore. This is because the pioneer-ing EGS developments at Fenton Hill (USA), Rosemanowes (UK),Hijiori (Japan) and Soultz (France) (3.5 km reservoir) did not pro-duce events large enough to disturb the local population, whereasmore recent attempts to develop systems at 4.5–5.0 km depth atSoultz, Cooper Basin (Australia) and Basel (Switzerland) producedevents approaching or exceeding magnitude 3. There are also oneor two instances where small but felt events have been associatedwith the operation of deep (∼3 km) hydrothermal systems.

The recent increase of interest in developing deep geothermalsystems and of sequestering large quantities of CO2 undergroundmakes it desirable to identify factors that influence the differentseismogenic responses to fluid injection at various sites. This paperdocuments the first stage of an on-going study that seeks to deter-

mine such factors through examination of incidences where fluidinjection has taken place without generating seismic events thatwere felt by the local population, as well as cases where it has. Wemake no distinction between induced and triggered seismicity.

K.F. Evans et al. / Geothermics 41 (2012) 30– 54 31

Fig. 1. Location of injection sites discussed in the paper shown on a Mercator projection. There are four sites in the Munich area that are considered: Riem, Pul-l obal sh of loa 50 ye

hdilonTWare

taabiarlfwia

sbsgtP

ach, Unterschleissheim and Unterhaching. The background is taken from the glttp://www.seismo.ethz.ch/static/GSHAP. The color scale denote the GSHAP indexcceleration (PGA) in %g on stiff soil that has a 10% probability of being exceeded in

The scope of this study is limited to reporting European caseistories of injection-related seismicity, including the project toevelop an EGS beneath the city of Basel in 2006. Most, if not all

ncidences of injection into crystalline rocks are included, regard-ess of whether events were felt. However, at this stage our coveragef injection into sedimentary rocks is not complete, especially foron-geothermal injections that did not produce felt earthquakes.wo sites where CO2 is injected into sedimentary rock are included.e exclude from consideration other types of induced seismicity

ssociated with such activities as fluid withdrawal, excavation oreservoir impoundment since these may involve somewhat differ-nt mechanisms.

The case histories of injections into igneous and sedimentaryarget rocks are presented in separate sections and are orderedlphabetically according to country. This structuring has the dis-dvantage that case histories from the same geological provinceut different countries are not reported contiguously, although the

ncidences are few. The locations of the sites are shown in Fig. 1, and summary listing of parameters for the sites is given in Table 1. Aock mass will be referred to as ‘critically stressed’ if the shear stressevel it supports would produce failure of an optimally orientedracture whose strength is described by Coulomb friction criterionith a coefficient 0.65. The maximum and minimum principal hor-

zontal stresses will be denoted by SHmax and Shmin, respectively,nd the vertical stress by SV.

For the purpose of comparing the natural seismic activity at aite, we used an index of estimated local seismic hazard which isased on an assessment for all of Europe that conforms to a single

tandard. This hazard is quantified in Table 1 and Fig. 1 by the peakround acceleration (PGA) in units of ‘g’, the acceleration of gravityaken from the compilation of Giardini et al. (1999). The value ofGA for a site denotes the acceleration level on stiff soil that has

eismic hazard map of the Global Seismic Hazard Assessment Program (GSHAP)cal seismic hazard from natural earthquakes defined in terms of the peak groundars (equivalent to a recurrence period of 475 years).

a 10% probability of being exceeded in 50 years (equivalent to arecurrence period of 475 years). Regions of low hazard are charac-terised by values below 0.08 g, high hazard areas have PGAs above0.24 g, and moderate hazard regions present intermediate values.It should be emphasised that this value is not necessarily a measureof the seismogenic response of the ground to injection, but rather,is a conveniently available index of natural seismic hazard at a sitethat is based on the peak earthquake-induced shaking that is likelyto occur at the site. Consequently, PGA values will be high for locali-ties with no natural seismicity if it neighbours a region where largeevents occur. Such situations are noted in the text.

Throughout this report, seismic magnitudes are given aseither local magnitude, ML, duration magnitude, MD or moment-magnitude, MW. Whenever possible, macroseismic intensities aregiven in terms of the EMS-98 scale (Grüntal et al., 1998) anddenoted ‘Io(EMS)’. Most macroseismic intensities given in thispaper were derived before EMS-98 was defined. These values aredenoted as ‘Io’ and should be broadly consistent with the EMS-98classification. The most likely difference is that indistinct valuessuch as Io = IV–V or Io = VI–VII would most probably be assessed asIo(EMS) = IV and Io(EMS) = VI respectively in the EMS-98 classifica-tion.

2. Geothermal injection case histories: igneous rocks

2.1. France

2.1.1. Le Mayet de Montagne, France

The site is located 25 km south-east of Vichy, France, at the

northern fringe of the Massif Central where the granite outcrops.It was established in 1985 as an EGS test site with two boreholesdrilled to 800 m depth. The local stress state is characterised as

32K

.F. Evans

et al.

/ G

eothermics

41 (2012) 30– 54

Table 1Summary of site and injection parameters and the associated maximum earthquake magnitudes for the cases discussed in the text. The injection parameters refer to the operation (i.e. stimulation or circulation) associated withthe largest magnitude event detected.

Setting

Location/Countrya Depthb

(km)Rockc Stressd

regimeCriticalstress?

PGAe

(% g)Date Inject.f type Qmaxg

(l/s)Pw-max(MPa)

Vinjh (m3) Max MLi Well sep.j

(km)Comments

Geothermal: igneous rocksLe Mayet/FR 0.75 Granite NF Yes 7 1987 Stimulation 73 25 200 N-Felt Attempted hydrofractureSoultz/FR 3.5 Granite SS/NF Yes 8 1993 Stimulation 38 10 20 × 103 1.9 Event 9 days after shut-inSoultz/FR 5.0 Granite SS/NF Yes 8 2003 Stimulation 90 18 37 × 103 2.9 Event after shut-inBad Urach/DE 4.3 Gneiss SS/NF Yes 9 2002 Stimulation 50–4 34 5.6 × 103 1.8 Initial inj. rate drops from 50

to 10 l/s in hrs.KTB/DE 9.0 Gneiss SS Yes 6 1994 Inject. test 9 55 200 1.2KTB/DE 3.0/6.0 Gneiss SS Yes 6 2000 Inject. test 1.2 30 2 × 103 0.5Landau/DE 3.0 Gr/SS/Carb. 8 2007 Circulation 70 6.0 Balanced 2.7 1.3 Wells intersect faults. Event

after 1.9 yrs.Krafla/IS 2.0 Basalt 24 2002–2004 Circulation 45 0.1 ≤2.0 Operational geothermal fieldLaugaland/IS 1.8/2.8 Basalt 14 1997–1999 Circulation 6–21 3.4 <−1 Recharge depleted reservoirSvartsengi/IS 2.0 Basalt 40 1993 5-month inj. 30 3.4 200 × 103 <−1 Recharge depleted reservoirHellisheidi/IS 2.5 Basalt 49 2003 Drill&Stim. 50 1.7k 2.4 Well HE-8: event at 7 kmMonte Amiata/IT 3.0 Metam. SS/NF Yes 19 1969 Circulation 3.5 Operational geothermal fieldFjällbacka/SE 0.5 Granite TF Yes 3 1989 Stimulation 21 13 200 −0.2 Felt event during circulationBasel/CH 5.0 Granite SS Yes 15 2006 Stimulation 55 30 12 × 103 3.4Rosemanowes/UK 2.5 Granite SS Yes 4 1987 Circulation 33 11 2.0 0.17

Geothermal: Sedimentary rocksSimbach-Brunau/AT 1.9 Carb. SS/TF 5 2001 Circulation 74 0.1k Balanced N-Rep 2.1 Both wells intersect faultsAltheim/AT 2.2 Carb. SS/TF 6 2001 Circulation 81 <1.7 Balanced N-Rep 1.6 Both wells near faultsGeinberg/AT 2.1 Carb. SS/TF 6 1998 Circulation 21 <0.2 Balanced N-Rep 1.6 Injection well near a faultBad Blumau/AT 2.6 Carb. 7 1999 Circulation 30 <0.7 ∼Balanced N-Rep 1.8 Both wells near faultsThisted/DK 1.25 SS 4 2001 Circulation 56 1.7 Balanced N-Rep 1.5Margretheholm/DK 2.5 SS 2 2004 Circulation 65 7.0 Balanced N-Rep 1.3Paris/FR 1.16-1.98 Carb. 4 1971- Circulation 83 3.5 Balanced N-Rep 1.0–1.4 Faults not targetedNeustadt-Glewe/DE 2.4 SS 2 1995 Circulation 31 0.8k Balanced N-Rep 1.5Waren/DE 1.55 SS 2 1984 Circulation 14 6.0k Balanced N-Rep 1.3 P-inject. reduced to 1.6 MPa in

1986Neubrandenburg/DE 1.25 SS 2 1989 Circulation 28 1.1k Balanced N-Rep 1.2Gross Schönebeck/DE 4.0 SS/Volc. NF Yes 1 2007 Stimulation 150 59 13 × 103 −1.1 Data from stimulation of 2nd

wellHorstberg/DE 4.0 SS 2 2003 Stimulation 50 32 20 × 103 <0Straubing/DE 0.8 Carb. SS/TF 5 1999 Circulation 45 1.5k 90% injected N-Rep ∼1.7 Injection well intersects minor

faultMunich-Pullach/DE 3.4/3.4 Carb. SS/TF 5 2005 Circulation 32 4.0k Balanced N-Rep ∼1.8 Neither well intersects faultsMunich-Riem/DE 2.7/3.0 Carb. SS/TF 5 2004 Circulation 75 2.5k Balanced N-Rep ∼2.2 Production well intersects

faultUnterhaching/DE 3.6/3.35 Carb. SS/TF 5 2007 Circulation 120 2.5k Balanced 2.4 4.0 Both wells intersects faultsUnterschleissheim/DE 1.6/1.6 Carb. SS/TF 5 2003 Circulation 100 <1.0k Balanced N-Rep ∼2 Both wells near faultsBruchsal/DE 2.0/2.5 SS 7 2008 Circulation 24 0.5k Balanced N-Rep 1.4 Both wells intersect faults

K.F.

Evans et

al. /

Geotherm

ics 41 (2012) 30– 54

33

Larderello-Travale/IT 2.0 Carb. SS/NF Yes 16 1977 Circulation 3.0 Operational geothermal fieldLatera/IT 1.0 Carb. SS/NF Yes 19 1984 Injection Balanced 2.9 2.0 Event near active injection

well L2.Torre Alfina/IT 2.0 Carb. SS/NF Yes 20 1977 Injection 40 1.2 4.2 × 103 3.0 Max. event larger than

backgroundCesano/IT 2.0 Carb. SS/NF 14 1978 Injection 15 7.5 2 × 103 2.0 Injection into well RC-1Bialy-Dunajec/PO 2.4 Carb. NF Yes 11 2001 Circulation 186 3.8k Balanced N-Rep 1.2-1.7 In operation since 1992Uniejów/PO 2.0 SS 2 2001 Circulation 19 0.3k Balanced N-Rep 1.0 System extended in 2004Riehen/CH 1.25/1.55 Carb. SS 15 1989 Circulation 18 1.2 Balanced N-Rep ∼1

CO2 sequestration: sedimentary rocksKetzin/DE 0.65 SS 2 2008 Injection 0.8 kg/sl 1.7k N-Rep CO2 is a gas at formation P–TSleipner/NO-Offshore 0.8–1.1 SS 1996 Injection 32 kg/sl <3.4k N-Rep CO2 is supercritical at

formation P–T

a AT: Austria; CH: Switzerland; DE: Germany; DK: Denmark; FR: France; IS: Iceland; IT: Italy; PO: Poland; UK: United Kingdom.b Where two values are given: the first denotes the depth of the injection well.c Carb.: carbonate; GR: granite; Meta: metamorphics; SS: sandstones; Volc.: volcanics.d TF: thrust faulting (i.e. S1 is vertical); SS: strike-slip (i.e. S2 is vertical); NF: normal faulting (i.e. S3 is vertical).e Peak ground acceleration (% of the acceleration due to gravity (9.81 mm/s2)) with a 10% probability of being exceeded in 50 years as estimated on the basis of natural seismicity. Values of 7 or less denote low hazard: 8–23

denote moderate hazard: 24–48 denote high hazard: and 49 and greater denote very high hazard.f Stimulation: relatively short high-pressure injection to enhance rock mass permeability; Drill: injection during drilling due to fluid losses; Circulation: simultaneous injection and production from doublets or triplets whose

flow volumes may or may not be equal. ‘CO2’: injection of liquid carbon dioxide.g The injection or circulation parameters reflect peak operational values.h The fluid volume injected is difficult to estimate for circulations lasting more than a few months. Balanced: injection and production rates are equal.i N-Rep: No events reported either by local population or a regional/local network. N-Felt: events of uncertain magnitude recorded by a local network but none were felt by the local population.j Separation at reservoir depth between injection and production wells for circulation operations. No separation is given for single-well injection tests.k Value indicates estimated downhole injection pressure above the natural formation pressure.l Flow rates and volumes are specified in terms of mass because the fluid is a gas under formation pressure and temperature conditions.

3 therm

sirmt

bhaiwMoth1pfwtTwJ

2

(tRosamwormmotTo

d1ldci

itmTrdt1hdnp21id

4 K.F. Evans et al. / Geo

trike-slip with a Shmin/SV ratio of 0.55, and thus the rocks are crit-cally stressed. Natural seismicity is low, although the historicalecord indicates that two events estimated to be of moment-agnitude MW 4.3 have occurred at distances of 16 and 27 km from

he site over the last 130 years (Grünthal et al., 2009).Numerous injection experiments were conducted at depths

etween 400 and 800 m in attempts to develop an undergroundeat exchanger. Microseismicity was recorded with a 15-stationrray. The first stimulation at the site was conducted by inject-ng 61 m3 of water through a double packer at 443 m depth at a

ellhead pressure of 19 MPa. Six seismic events of local magnitudeL −1.3 to −0.9 were recorded (Talebi and Cornet, 1987). Numer-

us stimulation injections followed. One of the largest involvedhe injection of 200 m3 of fluid at 73 l/s and up to 25 MPa well-ead pressure (Table 1), and resulted in a shut-in pressure of6 MPa (Cornet, 1989), much greater than the 4 MPa wellheadressure needed to attain Shmin within the reservoir. Relativelyew events were recorded during the stimulations, although 140ere detected at depths of 400–800 m during the various circula-

ion tests conducted on the doublet system (Evans et al., 1992).he event magnitudes have not been reported, although noneere felt by the on site project personnel (F. Cornet, pers. comm.,

une 2009).

.1.2. Soultz-sous-Forêts, FranceThe Soultz-sous-Forêts site is located in the Upper Rhine Graben

URG), some 40 km NNE of Strasbourg, France (other sites inhe URG described in this paper are Bruchsal, Landau, Basel andiehen). At the Soultz site, the granitic basement lies below 1.4 kmf sediments. Graben-parallel faults produce a horst-and-grabentructure within the basement. Fracture zones within the basementre high-angle and strike approximately normal to the E–W averageinimum stress direction. The magnitude of Shmin is about 0.5SV,hich is typical of a graben setting. Thus, the rock mass, and many

f the large-scale structures within it, is critically stressed at alleservoir depths (Evans, 2005; Valley, 2007). The region has low-to-oderate seismic hazard (Table 1). In 1954 a series of events withagnitudes up to ML 4.8 and intensities up to Intensity Io(EMS) VI

n the European Macroseismic Scale (EMS-98) occurred 10–20 kmo the southeast of Soultz towards Seltz/Wissenbourg (Helm, 1996).he hypocentre depth is uncertain although the macroseismicbservations suggest several kilometres.

The development of the project site began in 1987 with therilling of a 2 km deep well to explore the granitic basement below.4 km. Subsequently, a doublet system was developed and circu-

ated at 3.0–3.5 km depth in 1992–1997, and a triplet at 4.5–5.0 kmepth developed and tested between 1998 and 2009. Long-termirculation of the deep system with power production commencedn 2010 (Genter et al., 2010).

All wells were subjected to massive hydraulic injections of typ-cally 20,000–40,000 m3 at flow rates of 40–80 l/s and pressureshat reached the minimum stress value (Cornet et al., 2007). Seis-

ic activity was monitored with downhole and surface networks.ens of thousands of events associated with each stimulation wereecorded by the downhole array. For the reservoir at 3.0–3.5 kmepth, the largest event that occurred during the first massive injec-ion into the rock mass in 1993 was assigned a magnitude of ML.7 from the surface array (Helm, 1996) and occurred during theighest-rate injection of 36 l/s. However, a ML 1.9 occurred some 9ays after shut-in. Neither of these events were felt by site person-el. At Soultz, magnitudes greater than ML 2.0 can felt by the nearbyopulation under ideal conditions (N. Cuenot, pers. comm., May

010). Fluid circulation of the 3.5 km deep system for 4 months in997 was balanced (i.e., injection = production), and hence did not

nvolve a component of net injection. No seismicity was detecteduring this test (Baria et al., 1997).

ics 41 (2012) 30– 54

Stimulation of the wells of the deeper reservoir involved com-parable injection volumes to those used in the upper reservoir,and again pressures appear to have been limited near the mini-mum principal stress level by natural processes (Valley and Evans,2007). Stimulation of the first well that penetrated the 5 km reser-voir, GPK2, began in June 2000 with the injection of 22,000 m3 ofwater at rates of up to 50 l/s. Wellhead pressure rose to 14.5 MPa atshut-in. Some 700 events with magnitudes between ML 1.0 and 2.5occurred during injection, but a magnitude ML 2.6 occurred some 10days after shut-in (Dorbath et al., 2009). Stimulation of the seconddeep well, GPK3, took place in May 2003 and involved the injec-tion of 34,000 m3 of brine and water into GPK3, for the most part at50 l/s with occasional increases for a few hours of up to 90 l/s whichproduced the maximum wellhead pressure of 17.9 MPa. Midwaythrough the stimulation, some 3400 m3 of water was simultane-ously injected into GPK2 for ∼40 h at a rate of 20 l/s (Baria et al.,2004). GPK2 wellhead pressure rose to 7.9 MPa. Some 200 eventswith magnitudes between ML 1.0 and 2.5 occurred during this injec-tion (Charléty et al., 2007). A magnitude ML 2.9 event occurred 2days after shut-in, despite an attempt to avoid this by a stepwiseinjection rate reduction (Baria et al., 2004).

The stimulation of the third deep well, GPK4, began inSeptember 2004. Injection rate was maintained at 30 l/s with a fewshort increases of 2 h duration to 44 l/s. Peak wellhead pressure was17.5 MPa. The injection was terminated after injecting 9000 m3 dueto a pump failure. The stimulation program resumed in February2005 when a further 12,500 m3 of water was injected at up to 45 l/sand peak wellhead pressures of 18.5 MPa (Dorbath et al., 2009).Some 128 events with magnitudes between ML 1.0 and 2.7 wererecorded during injection, but none larger than ML 2.0 occurredduring shut-in.

Of the three deep Soultz wells, GPK3 appeared to be the mostprone to produce large events in response to injection. Dorbathet al. (2009) found the b-values for the GPK2 and GPK3 seismicityto be 1.23 and 0.94, respectively (although only over the respectivelimited magnitude ranges of ML 1.0–1.9 and 1.0–2.3), and suggestedthat the difference reflected the activation of a major fracture zoneintersecting GPK3. In 2005, the 3-well system was subjected to a6-month close-loop circulation test at 15 l/s using only buoyancydrive to produce from GPK2 and GPK4. The fluid was injected intoGPK3 at a wellhead pressure that progressively increased from 4to 7 MPa. Seismicity began soon after injection commenced, and atotal of 32 events exceeding ML 1.2 were recorded during the entireperiod, the largest being ML 2.3.

A 2-month closed-loop circulation test of wells GPK2–GPK3 wasperformed in 2008 at 25 l/s using a production pump in GPK2. Noseismicity was observed for 5 weeks during which time the GPK3injection pressure rose steadily to 6 MPa. Seismicity began oncethat pressure was exceeded, and included four events having mag-nitudes in the range ML 1.3–1.4 (Cuenot et al., 2010).

2.2. Germany

2.2.1. Bad Urach, GermanyThe site is located near the centre of a geothermal anomaly in

the foothills of the Swabian Alb mountains. The orientation of SHmaxof N82◦E is well defined by breakouts that are observed below1900 m depth in the wells at the site (Heinemann et al., 1992).The magnitudes of the horizontal principal stresses are not knownwith confidence, although the frequency of breakouts and theirconsistent orientation suggests they are significantly different. A‘hydrofracture’ stress test on a 1 m naturally fractured interval at

3350 m in Urach-3 yielded an instantaneous shut-in pressure (ISIP)of 41–50 MPa (Rummel et al., 1991), implying Shmin is about 50–55%of the vertical stress. However, this value is not obviously consistentwith the 33 MPa overpressure sustained in the 168 m open-hole

therm

iuw4anbe

3thttlswrV(sectittwtaw

opasmwidh

4dalhsesiio

tlbfaiswsrtw

K.F. Evans et al. / Geo

nterval during the hydrotesting phase. The area has moderate nat-ral seismicity. Several events of intensity Io IV–VI have occurredithin 15 km of the site in the past 200 years. However, it lies only

0 km from a region where events of up to ML 5.5 have occurred,nd hence the PGA takes a relatively high value. In 2002, a seismicetwork of five 3-component sensors was installed in 250 m deeporeholes to monitor the stimulation of the deepened hole (Schanzt al., 2003).

The HDR project was initiated in 1977 with the drilling of a334 m deep borehole, Urach-3, through 1.6 km of sediments intohe basement of metamorphic gneisses (Dietrich, 1982). Bottom-ole temperature was 143 ◦C. Seven inch casing was cementedo 3320 m, leaving 14 m of 8–1/2 in. open hole. Further access tohe formation was provided by three 5 m long perforated intervalsocated 25, 47 and 58 m above the casing shoe. Each interval wasubject to series of small-volume (<100 m3) stimulation injections,hich included gel and proppant for the perforated intervals, at

ates up to 20 l/s to establish a hydraulic linkage between them.ery high wellhead pressures of up to 66 MPa were required

Schädel and Dietrich, 1982). Post-stimulation interval transmis-ivities were of the order of 10−6 m2/s (Stober, 1986) implying anquivalent porous medium (EPM) permeability of 8 × 10−16 m2. Airculation loop was established by running a packer on produc-ion tubing into the well and setting it below the lowest perforationnterval. Most circulation tests involved injection into the rock masshrough the three perforated intervals in the annulus and produc-ion from the open hole section through the tubing. Fluid lossesere low but system impedance was high and increased during

he test sequence (e.g. 35 MPa wellhead pressure required to injectt 0.5 l/s in test 22) (Schädel and Dietrich, 1982). No seismic eventsere felt on-site during the injections (Stober, 2011).

In 1983, the hole was deepened to 3488 m, leaving 168 m ofpen hole. Initial hydraulic testing of the combined open hole anderforated intervals combined showed that the transmissivity ofll intervals had been significantly reduced during the deepening,uggesting mud invasion of the fractures had reduced their per-eability (Stober, 1986; Stober and Bucher, 2000). Transmissivityas also pressure-dependent, and markedly increased at wellhead

njection pressures above 17 MPa, indicating the onset of fractureilation. No limiting injection pressure, as might be expected forydrofracture growth, was reached up to 33 MPa (Stober, 1986).

In 1992, the well was further deepened with a 5–7/8 in. bit to395 m TVD (true vertical depth) as the first well of an intendedoublet. Bottom-hole temperature was 170 ◦C. Following logging,

drill string was lost in the well (Tenzer et al., 2000). The topies within the casing at 3234 m and obstructs access to the openole, although it remains hydraulically open. Injection tests showedimilar behaviour to that seen before the latest deepening, withvidence of fracture dilation above 17 MPa and no limiting pres-ure being reached up to a maximum wellhead pressure attainedn the tests of 25 MPa (Stober, 2005). The low-pressure transmissiv-ty was of the order 5 × 10−7 m2/s, implying an EPM permeabilityf 9 × 10−18 m2.

In September 2002, the 1125 m long open hole together withhe perforated intervals was subjected to a large-volume stimu-ation injection of 5600 m3 of brine and water. The stimulationegan with two ∼1 h, high-rate injections of heavy brine preparedrom high-grade salt. In the second of these, ∼150 m3 was injectedt 35–40 l/s and a wellhead pressure of 34 MPa, the maximummposed by casing limitations (Tenzer et al., 2004). The main watertimulation began immediately afterwards at a rate of 50 l/s and aellhead pressure of 34 MPa, but injection rate had to be progres-

ively reduced to 10 l/s over several hours due to rapidly increasingeservoir impedance. The first seismicity was detected 9 h afterhe start of operations (Schanz et al., 2003), and was coincidentith the injection impedance increase, suggesting that the initial

ics 41 (2012) 30– 54 35

high-rate phase had served to inflate the reservoir created in pre-vious phases of the project. After 5 days, a further slug of brine wasinjected to achieve higher downhole pressure. Injectivity declinedfurther, necessitating a further reduction in injection rate to 5 l/sand eventually 4 l/s to prevent wellhead pressure exceeding the34 MPa casing limit (Tenzer et al., 2004). This may have been dueto clogging of the flow paths by sediment from the salt used to pre-pare the brine (U. Schanz, personal comm., April 2011). Injectionwas paused after 7.4 days and the well vented in two ∼4 h rela-tively high-rate production periods during which fluid volumes of∼210 and ∼140 m3 were recovered at rates which declined to 12and 10 l/s. Each period was followed by a ∼4 h period of shut-in(note that the venting rate of 90 l/s reported by Stober (2011) forthis phase is in error (I. Stober, personal comm., March 2011)). Well-head pressure rose within ∼2 h to stable levels of 14.7 and 12.3 MPaduring the first and second periods respectively. Given the fluidvolumes recovered and the low porosity nature of the rock, theseobservations qualitatively indicate that reservoir overpressures inexcess of 10 MPa extended significant distances from the wellboreand were not limited to the near-field. A total of 420 events weredetected during the injection. These had moment-magnitudes of−0.6 to 1.8, and extended more than 500 m from the well (Tenzeret al., 2004).

A step-rate injection test was conducted 2 weeks after the stim-ulation. A total of ∼1200 m3 of water was injected at rates of 2.4 and1.8 l/s and wellhead pressures of up to 17.0 MPa. Following shut-in,wellhead pressure dropped quickly within a few hours to 13 MPabut then declined more slowly to reach 7.5 MPa after 6 days (Baischet al., 2004; Schanz et al., 2003). Transmissivity estimates lie inthe range 0.5–1.5 × 10−6 m2/s, only marginally higher than beforethe stimulation (Stober, 2011). A further set of injections was per-formed in summer 2003, the largest of which involved the injectionover 10 days of ∼1950 m3 of water at a rate of 2.1 l/s and wellheadpressure of up to 18.3 MPa. A total of 218 events were detected(Tenzer et al., 2004). The seismic energy release rate continued toincrease with pressure once 15 MPa was exceeded (Baisch et al.,2004).

In 2006, drilling of a second borehole, Urach-4, commenced withthe intention of forming a doublet. The drilling was stopped at adepth of 2793 m due to budgetary considerations (Stober, 2011).The condition of the wells was evaluated in 2008 with a view todeveloping a duplet between the 100 m separated wells at ∼2500 mdepth. Urach-4 could not be logged below 1600 m because of gelleddrilling fluid. The costs of cleaning the well were considered pro-hibitive and so the proposal was abandoned (Cammerer et al.,2009). The lost drill string remains in Urach-3.

2.2.2. KTB borehole, GermanyThe 9.1 km deep German Continental Deep Borehole (i.e. the

KTB borehole) that penetrates gneisses and amphibolites was com-pleted in 1994 in SE Germany. Natural seismicity at the site isvery low. However, it lies just south of a region of moderate nat-ural seismic activity characterised by swarms of events whoseintensity rarely exceeds Io V within 30 km of the site. The rocksare in a critical stress state, at least between 3 and 7.5 km depth(Brudy et al., 1997). Shortly after completion, about 200 m3 ofbrine was injected at up to 9 l/s and 55 MPa wellhead pressureinto the 70 m open-hole section at well bottom. Microseismicitywas monitored using a temporary surface network of 73 short-period stations augmented by an instrument located at 3.8 kmdepth in the KTB pilot borehole. A group of 400 events of magni-tude less than ML 0 extended several hundred metres above the

injection interval, and a cluster that included the largest eventof ML 1.2 occurred at 8.6 km depth (Zoback and Harjes, 1997). Afew events were observed as much as 1.5 km above the injectioninterval.

3 therm

a633l6aomap9bi

2

ttsiislsmircoesUsseaus

optd1tfLawes

ftft2Lmi2oce

6 K.F. Evans et al. / Geo

In 2000, a larger injection of 4000 m3 of water was performedt flow rates of up to 1.2 l/s and wellhead pressures of 30 MPa over0 days. The injection was monitored by a surface network of 40-component stations augmented with a downhole instrument at.8 km in the pilot borehole. A total of ∼2800 seismic events were

ocated. The vast majority were found to be clustered at 3.3, 5.4,.6 km, as well as near the hole bottom at 9.0 km. These depthsre believed to reflect points where fluid flow into the rock massccurred, through casing leaks or open hole. The maximum eventagnitude was ML 0.5 (Baisch and Harjes, 2003). More recently,

10-month injection test was conducted into the 4.0 km deepilot hole at a constant rate of ∼3 l/s and a wellhead pressure of–12 MPa. Some 3000 events were recorded (Shapiro et al., 2006),ut they were of lower magnitude than those in the earlier exper-

ments (Kümpel et al., 2006).

.2.3. Landau, GermanyThis dual use (electricity/district heating) project is located in

he Upper Rhine Graben some 35 km NE of Soultz (Fig. 1). The injec-ion occurs into both basement and the lowermost units of theedimentary section, and so this site could equally well be includedn Section 3 on sedimentary rock (Schindler et al., 2010). The localitys cut by high-angle faults that strike N–S (Illies and Greiner, 1978),ub-parallel to the NNE–SSW strike of the graben. There is no pub-ished stress information for the site. However, given the grabenetting, the focal mechanism solutions of local earthquakes and theeasured stress state at the Soultz EGS site, it is likely that Shmin

s oriented E–W to SW–NE, and is substantially less than SV. Theegion has low-to-moderate natural seismic activity with histori-al events that produced maximum intensities of up to Io VII–VIIIn the European Macroseismic Scale (EMS-98). An event with anstimated maximum intensity of Io VII occurred some 10 km to theouth of Landau, near Kandel in 1903 (Ahorner et al., 1970b). For thepper Rhine Graben region, a maximum intensity Io(EMS) VII for

hallow events corresponds to a magnitude ML of 4.0–4.5. Macro-eismic observations suggest a depth of a few kilometres (Ahornert al., 1970b). The site borders an oilfield that has produced oil from

formation at 1.1 km depth since 1955 (Doebl, 1968) and contin-es to do so today. Long-term fluid extraction can be a source ofeismicity (Van Eck et al., 2006).

During 2005 and 2006, two boreholes were drilled deviated inpposite E–W directions so as to penetrate faults that cut relativelyermeable carbonates and sandstones at the base of the sedimen-ary section and the uppermost level of the basement at ∼3 kmepth (Schindler et al., 2010). Well separation at reservoir depth is.3 km. The faults are believed to intersect some distance north ofhe site, and so flow within the reservoir is complex and probablyault-controlled (Baumgärtner et al., 2010b). The injection well, Gt-a2, was subjected to a hydraulic stimulation at rates of up to 190 l/snd wellhead pressures of 13.5 MPa (Hettkamp et al., 2007). Thereas no felt seismicity associated with this operation (Baumgärtner

t al., 2010a). The production well, Gt-La1, did not require hydraulictimulation because it intersected a highly transmissive fault.

The doublet was tested during 2007, and power productionrom a 3.8 MWe ORC plant was demonstrated in November ofhat year. A balanced circulation rate of ∼65 l/s was maintainedrom February to November 2008, during which time the injec-ion pressure declined from 6.0 to 3.0 MPa (Baumgärtner et al.,010a). Following the installation of a downhole pump in Gt-a1, circulation at 70 l/s resumed in February and continued untilid-September 2009. Injection pressure during this time steadily

ncreased from 4.0 MPa to almost 5.5 MPa (Baumgärtner et al.,

010a). In February 2008, two small earthquakes with magnitudesf ML 1.7 and 1.8 were recorded in the area by the Seismologi-al Services of Baden-Württemberg and Rheinland-Pfalz. Anothervent of ML 1.7 occurred in the area in October 2008, and three

ics 41 (2012) 30– 54

further events of magnitudes ML 1.6–1.9 were detected in the areaon 9 May 2009. Although these events occurred in the area of Lan-dau, their depths were not well constrained. On 15 August 2009,an ML 2.7 event that was felt by the population of Landau occurredshortly after operation of the system had been halted for mainte-nance operations. The event hypocentre was located by an expertgroup as lying 1.5–2.0 km north of the plant at a depth of 2.3–3.3 km(Bönnemann et al., 2010). Thus the failure area could lie either inthe sediments or the basement or both. A further seven eventswere recorded on the same day. The plant resumed operation inNovember 2009 with the maximum injection pressure lowered to4.5 MPa.

A similar dual use project is being developed about 5 km southof Landau near Insheim. Drilling and initial testing of the sec-ond well of the doublet was completed in 2009 (Baumgärtneret al., 2010a). In September 2009 the Seismological Services ofBaden-Württemberg and Rheinland-Pfalz recorded five events ofmagnitude approximately ML 2.0 within a few kilometres of the site(LED-LGRB, 2010). The relationship of these events to operations atthe site is unclear.

2.3. Iceland

Geothermal well injection in Iceland is used both for reser-voir stimulation and pressure maintenance or recharge. In mostcases, injection takes place at pressures less than several MPa oris gravity-driven since the basalt reservoirs tend to have relativelyhigh natural transmissibility. The reservoirs and hence the injec-tion horizons also tend to be comparatively shallow, the deepestinjection well being 2.8 km. Many geothermal areas experienceintermittent phases of natural seismic activity that are clearly unre-lated to fluid injection or other human activities. Here we describefour sites where the seismic response to injection has been docu-mented.

2.3.1. KraflaGeothermal power has been generated at Krafla since 1977.

A 60 MWe power station is located in the Krafla caldera, whichwas the source of a series of eruptions and intrusions associatedwith crustal accretion that took place between 1975 and 1984(Gudmundsson, 2001). The magmatic events were accompaniedby considerable seismicity which subsequently declined to lowlevels. Analysis of focal mechanisms of earthquakes up to magni-tude ML 2.1 occurring 1 year after the termination of the eruptionsindicated that a heterogeneous stress field prevailed at that time(Foulger et al., 1989). Some of the events were believed to resultfrom heat-mining operations. In 2004 a passive seismic experimentwas performed around the main injection well KG-26 of the high-temperature Krafla-Leirhnúkur field, which is located within thecaldera, 1 km north of the power station. This well supplies fluidto the lower reservoir at a depth of 2.0–2.1 km (Tang et al., 2005a),and had been used as an injector more-or-less continuously since2002 at rates of up to 70 l/s and wellhead pressures of 0.3 MPa.In 2004, the rate and wellhead pressure were 45 l/s and 0.1 MPa(Á. Gudmundsson, pers. comm., July 2010). The seismic activityand velocity structure (including anisotropy) near the well weremonitored for 2 months with two 20-station 3-component arrays(Onacha et al., 2005). During this period, injection into K-26 washalted for 11 days to study the effect on seismicity and velocitystructure. An average of four locatable seismic events per day ofmagnitudes less than ML 2.0 were detected (Tang et al., 2008). Thehypocentres lay between 1 and 3 km depth, and defined a predom-

inantly E–W trend near the well (Kahn, 2008; Tang et al., 2008).Consequently, it is likely the majority of events were associatedwith injection, although no obvious change in event frequencyaccompanied the halt in injection (Tang et al., 2005b). However,

therm

siot

2

d2hpatb(

2

r2a5awiue

2

aneonsmTpbte

oHwaidepiwecsamdaa

2iow

K.F. Evans et al. / Geo

hear velocities along local raypaths around the well increased dur-ng the halt in injection, implying the closure of fractures that werepen during injection (Tang et al., 2005a). This is consistent withhe observation of shear-wave velocity anisotropy.

.3.2. LaugalandThis is a low-temperature field in central N-Iceland where pro-

uction had resulted in a pressure drop of 3.5 MPa (Axelsson et al.,000). To recharge the reservoir, 15–20 ◦C water from the districteating system was injected at rates of 6–21 kg/s and wellheadressures of up to 2.8 MPa into two deep wells (1.6 and 2.8 km) over

2-year period. Microseismic activity was monitored throughouthe operations by a network with a detection threshold of ML −1,ut no events that could be ascribed to the injection were recordedAxelsson et al., 2000).

.3.3. SvartsengiExploitation of this high-temperature field since 1976 had

esulted in a 2 MPa drop in reservoir pressure (Brandsdóttir et al.,002). In 1984, intermittent injection of 70–80 ◦C waste water into

2.0 km deep well commenced with year-averaged rates as high as5 l/s. A single seismometer installed on site in 1984 failed to detectny events within the field through 2001. The single instrumentas supplemented by a 16-instrument array for a 5-month period

n 1993 when 217,000 m3 of water was injected at rates up to 30 l/snder gravity, but again no events were recorded (Brandsdóttirt al., 2002).

.3.4. HengillThis extensive geothermal area is located at a triple-junction in

volcanically active region. Two geothermal fields lie immediatelyorth and south of the Hengill volcano. The area is characterised bypisodes of natural seismic swarm activity, and two ML 5.0 eventsccurred in 1998 (Agustsson and Halldorsson, 2005). Consequently,umerous studies have been conducted in the area to assess theeismic hazard (Agustsson and Halldorsson, 2005) and for geother-al exploration purposes (Arnason et al., 2010; Tang et al., 2006).

he field to the north, called Nesjavellir, is one of the highest tem-erature geothermal systems under exploitation in Iceland and haseen producing power since 1987 (Arnason et al., 2010). The fieldo the south, called Hellisheidi, is also high-temperature and wasxplored somewhat later. It began producing power in 2006.

Significant injection-induced microseismicity appears to haveccurred during the drilling and stimulation of a 2.8 km deep wellE-8 in the Hellisheidi field in 2003 (Bjornsson, 2004). The areaithin a few kilometres of the well had been microseismically

ctive between 1995 and 1999, possibly associated with activ-ty at Mt. Hengill, but had been largely quiet since then. Duringrilling, water was lost into the formation at rates of 20–50 l/s,ssentially constituting injection (Bjornsson, 2004). Drilling wasaused at 2500 m and the hot well stimulated by intermittently

njecting cold water at up to 60 l/s for several days. This coincidedith the detection by the national seismic network of a series of

arthquakes below the well at depths provisionally given in theatalogue as 4–6 km. Shortly after drilling recommenced, anothereries of events occurred near the well, the largest of which had

magnitude ML 2.4 and a provisional depth of 7 km. After severalonths of shut-in, cold water was injected for 15 days at 50 l/s with

ownhole pressures of 1.7 MPa (Fig. 4 of Bjornsson (2004)). This wasccompanied by events with magnitudes in the ML −0.2–1.2 ranget depths between 4 and 6 km.

Similar behaviour was observed during injections into the

.0 km deep well HE-21 in the same field in February 2006. Follow-

ng completion, the well had a low injectivity, and was stimulatedver a 3 days period by first circulating and then injecting coldater at progressively higher rates and downhole pressures up to

ics 41 (2012) 30– 54 37

1.8 MPa above the formation pressure (Axelsson et al., 2006). By theend of the operation, some 31 l/s of the fluid entered the formationat a downhole pressure of 1.4 MPa above the formation pressure(Mortensen et al., 2006). During this time, several small events ofmagnitude up to ML 2.0 were detected close to the well by thenational network of the Icelandic Meteorological Office (Axelssonet al., 2006; Vogfjörd and Hjaltadóttir, 2007). Consequently, two 3-component seismic stations were temporarily installed to improveevent detectability and location accuracy (K. Vogfjörd, pers. comm.,July 2010). Thermal stimulation operations resumed after 1 weekof shut-in with injection for 2 days at 65 l/s and a downhole over-pressure of 3.4 MPa (Mortensen et al., 2006). Most fluid enteredthe formation through a fracture zone at 1850 m. No events weredetected during injection although a few that were too small tolocate occurred 1–2 days later (Vogfjörd and Hjaltadóttir, 2007). Afew days later, a 2.5 day injection was conducted. The details of thisinjection are not currently available, although it appears that flowrates were not significantly higher than the earlier injection due tolimitation of the water source, which was two nearby groundwaterwells. Seismicity began after 24 h of injection and included a ML 2.7event. A total of 50 events could be located, the vast majority ofwhich lay between 1.0 and 2.5 km depth and defined a structurethat extended from the well towards the northeast (Vogfjörd andHjaltadóttir, 2007).

The four Icelandic examples above emphasise the importance ofstress criticality in injection-induced microseismicity. The Lauge-land and Svartsengi reservoirs had suffered pressure depletion,and thus injection would serve to restore the stress-state topre-exploitation conditions, whereas the Hellisheidi and Kraflareservoirs are located in zones which are undergoing, or haverecently undergone, tectonic activity, and hence the stress stateis more likely to be critical.

2.4. Italy

2.4.1. Monte Amiata areaThis area is located at the southern boundary of Tus-

cany, and contains the Piancastagnaio and Bagnore fields (Billiet al., 1986). Both fields have two reservoirs: a shallow one inevaporites/carbonates at 0.6–1.0 km depth, and a deeper, water-dominated reservoir in the metamorphic basement at 2.5–3.5 kmwith temperatures of 300–350 ◦C (Barelli et al., 2010; Bertiniet al., 2005). The level of background seismicity at the site is sub-stantial (Batini et al., 1980b, 1990), and tends to mask potentialinduced events. Studies of historical seismicity since year 1000AD indicate that a relatively large event occurred in the areathat produced intensity IX MCS shaking (Batini et al., 1990). An8 MWe power plant supplied with fluid from the deeper reser-voir was installed at Piancastagnaio in 1969 and was expanded to88 MWe in the middle-to-late 1990s through the addition of 20 MWplants.

A seismic network of 10 stations was installed at the site in1982 and operated until 1992 when the network geometry waschanged (Moia, 2008). Seismicity is generally shallower than 8 km,and tends to occur in swarms with many small events (Moia, 2008).A magnitude ML 3.5 event occurred within the reservoir region in1983 (Moia et al., 1993). However, it could have been a naturalevent that would have occurred in the absence of injection activ-ity (Batini et al., 1990; Moia et al., 1993). Examination of the INGV(Istituto Nazionale di Geofisica e Vulcanologia) catalogue showsthat a further MD 3.5 (duration magnitude) event occurred within10 km of Piancastagnaio at a reported depth of 5 km in 2000 (INGV

Earthquake Catalogue).

The 88 MWe power plant is still in operation, although expan-sion is being slowed by environmental concerns unrelated toseismicity (Cappetti et al., 2010).

3 therm

2

2

ctseIdaf

wawm1

cgAdirm1bwOw5r

2

2

RtmNomSso2Tcov

it1∼3aww

twws

8 K.F. Evans et al. / Geo

.5. Sweden

.5.1. Fjällbacka, SwedenThis was a shallow (∼500 m) facility established on the west

oast of Sweden in 1984 to evaluate the Bohus granite as a poten-ial HDR reservoir (Wallroth, 1992; Wallroth et al., 1999). Naturaleismicity is low, although the historical record indicates that sev-ral events of magnitude approaching ML 4.0 and intensities up too 5 occurred within 25 km of the site. The stress regime at reservoirepth is ‘thrust’, and near-critical in as much as an overpressure of

few MPa above hydrostatic is sufficient to produce shearing of aracture set (Jupe et al., 1992).

Borehole Fjb1 was drilled to about 500 m depth and stimulatedith numerous injections culminating with 200 m3 of gel at 21 l/s

nd 13 MPa wellhead pressure. A 16-station microseismic networkas installed that recorded 74 events during the stimulation withagnitudes ranging between ML −1.3 and −0.2 (Eliasson et al.,

988).Subsequently, well Fjb3 was drilled through the microseismic

loud about 100 m from Fjb1 and stimulated by injecting 36 m3 ofel with proppant at 16 l/s and wellhead pressures of up to 19 MPa.

further 50 events were recorded. The system was circulated for 40ays during 1989 by injecting water at 1.8 l/s into Fjb3 and produc-

ng Fjb1 against a 0.3 MPa back-pressure. Injection pressure at Fjb3ose quickly to 3 MPa within a few hours and then progressivelyore slowly to reach 5.2 MPa by the end of the test (Eliasson et al.,

990). Only 50% of the injected fluid was recovered, the remaindereing lost to the formation. Several hundred microseismic eventsere recorded at distances up to 400 m from the injection point.ne was felt on site by the project employees but no complaintsere received from the local residents, some of whom lived within

00 m of the site. The event magnitude was not determined becauseecords were clipped (T. Wallroth, pers. comm., May, 2010).

.6. Switzerland

.6.1. Basel, SwitzerlandThe Basel EGS site is located at the southern end of the Upper

hine Graben, where it meets the fold and thrust belt formed byhe Jura mountains. The granitic basement lies under 2.4 km of sedi-

ents. The average orientation of Shmin within the granite section is144◦E ± 14◦ (Valley and Evans, 2009), and is consistent with thatbtained from inversion of focal mechanisms from natural seis-icity (Kastrup et al., 2004). A lower bound on the magnitude of

hmin in the open-hole interval (4629–5000 m) given by 0.69SV iset by the maximum downhole injection pressure of 74 MPa devel-ped during the stimulation injection of well BS-1 (Haering et al.,008). The magnitude of SHmax is currently not well constrained.he region is characterised by moderate seismicity, although theity of Basel was severely damaged by a nearby earthquake thatccurred in 1356 (Meghraoui et al., 2001). For this reason, the PGAalue listed in Table 1 is relatively high.

Borehole BS-1 was drilled to 5 km depth within the city of Baseln 2006 as the first well of an intended doublet. In December 2006,he lowermost 370 m was stimulated by injecting approximately1,500 m3 of water at rates that were progressively increased from1 to 55 l/s over 5 days. Wellhead pressures were restricted to0 MPa by casing limitations (Haering et al., 2008). Seismic activityccompanying the injection was monitored on a six-sensor net-ork of borehole stations installed at depths of 317–2740 m, asell as surface stations of various networks.

Seismicity began at injection pressures of a few MPa indicating

hat the rock mass is critically stressed. More than 10,500 eventsere recorded on the borehole array during the injection phase,ith event rate and magnitude increasing with flow rate and pres-

ure. On the fifth day of injection (8 December 2006), a magnitude

ics 41 (2012) 30– 54

ML 2.6 event occurred within the reservoir. Since this exceeded thesafety threshold for continued stimulation, the rate was reducedto 30 l/s for 5 h before shutting-in. Two events of magnitude ML2.7 and 3.4 occurred during shut-in. Hence, venting was initiatedafter 5 h of shut-in to reduce the pressure as quickly as possible. Inthe following days about one-third of the injected water volumewas allowed to flow back from the well (Haering et al., 2008). Seis-mic activity declined rapidly thereafter, although three events withmagnitudes exceeding ML 3.0 occurred 1–2 months after bleed-off (Deichmann and Giardini, 2009; Mukuhira et al., 2008). Minor,sporadic microseismic activity is still occurring more than 4 yearslater.

The hypocentre locations of 195 induced events that occurredsince the start of injection, and that were strong enough to berecorded by the surface network of the Swiss Seismological Ser-vice, are shown in Fig. 2. The overall hypocentre distribution definesa near-vertical lens-shaped cloud of 1.2 km diameter that strikesNNW–SSE and lies between 4 and 5 km depth. During injection,seismicity migrated away from the borehole, as would be expectedfor a diffusion-regulated process (see Fig. 6 of (Deichmann andGiardini, 2009)), and step-increases in flow rate/pressure tendedto increase event rate (Haering et al., 2008). After shut-in andbleed-off, seismic activity took place mainly at the periphery of thestimulated volume (Deichmann and Giardini, 2009). Fault planesolutions determined for the 28 strongest events are mostly purestrike-slip mechanisms with two pure normal faulting, and onea mixture of the two (Deichmann and Ernst, 2009). These focalmechanisms are largely in accord with the mechanisms of the natu-rally occurring regional seismicity. However, the strike of the nodalplanes of most of the events is oblique to the overall orientation ofthe seismic cloud. In fact, the focal mechanism of the ML 3.4 eventand its neighbours, together with the alignment of these events,suggest that failure occurred on a WNW–ESE striking, near-verticalplane (Deichmann and Ernst, 2009; Kahn, 2008). This suggests thatthe internal structure of the stimulated rock volume is composed ofa complex network of individual fault segments oriented obliquelyto the general trend of the microseismic cloud.

2.7. United Kingdom

2.7.1. Rosemanowes, Cornwall, UKThe Rosemanowes HDR project was active between 1978

and 1991, and culminated in the development and operationof a circulation system at a depth of ∼2 km within the Carn-menellis granite, which extends to the ground surface. Seismicactivity was monitored throughout operations with a surface net-work of 3-component accelerometers and occasionally a stringof hydrophones at reservoir depth (Batchelor et al., 1983). Thestress state is strike-slip and critical, the coefficient of the Coulombfriction strength law required to prevent failure at 2.0 km underambient conditions being 0.85 (Evans et al., 1992). The area has lownatural seismic hazard. The nearest events of note, which includea ML 3.5 event that occurred in 1981, are clustered near the townof Constantine, some 6 km south of the site (Turbitt et al., 1987).

Initially, two wells were drilled to 2050 m and stimulated with avariety of methods, including gel and water injections. For the mainwater stimulation, the wellhead pressure was 14 MPa and flow ratewas 90 l/s. Many tens of thousands of events of magnitude less thanML 0.16 were detected (derived from a moment of 1.8 × 109 Nmgiven by Baria et al. (1985)). None were felt by the local popula-tion or the on-site project staff (CSM-Report, 1989; Turbitt et al.,1987). The seismic activity preferentially grew downward, which

is ascribed to the pore pressure increase required for shear failuredecreasing with depth (Pine and Batchelor, 1984).

In 1985, a third well was drilled through the microseismic cloudto 2.65 km depth and stimulated with the injection of 5700 m3

K.F. Evans et al. / Geothermics 41 (2012) 30– 54 39

Epicentre locations relative to casing shoe

Horizontal location relative to casing shoe

(a) (b)

Fig. 2. Epicentre plot (a) and depth cross-section (b) parallel to the general trend of the epicentre alignment showing the locations of 195 events with magnitudes in therange ML 0.7–3.4, that occurred between December 2, 2006 and November 30, 2007 as a consequence of the reservoir stimulation below the city of Basel (after Deichmannand Giardini, 2009). Average uncertainties of the relative locations of the events, determined with a master-event location technique, are on the order of 50 m horizontallyand 70 m vertically. The location of the vertical well is marked by the black dot at (0,0) in (a), and by the thick (casing) and thin (open hole) vertical line in (b). The size of eachc > 3.0

D occuh The ars

odmcrFlnd(atttwhaatAd(

3

3

3

dtGwitwNt

ircle is proportional to the seismic moment of the event. The four events with ML

ecember 8 is located close to the bottom of the well; the other three ML > 3 eventsad been vented and wellhead pressures had returned to near-hydrostatic levels.

tress in the granite basement.

f intermediate-viscosity gel (0.05 Pa s) at rates up to 260 l/s andownhole pressures up to 35 MPa, which is 12.8 MPa above for-ation pressure (Pine et al., 1987; Parker, 1989). A program of

irculation tests that featured a variety of configurations and flowates commenced in August 1985 and ran until the end of 1989.luid losses averaged about 20%, and thus circulation constitutedong-term net injection. Losses and seismic activity increased sig-ificantly at injection pressures above 10 MPa (about 24 l/s), whenownhole pressures approached Shmin, indicating reservoir growthBaria and Green, 1990). The largest seismic event that can reason-bly be associated with project operations was a magnitude ML 2.0hat occurred in July 1987 and was mildly felt by the local popula-ion within a few kilometres of the site (Turbitt et al., 1987). At theime, the injection rate was 33 l/s at 11.1 MPa wellhead pressure,hich is somewhat less than the peak rate of 38 l/s at 12 MPa thatad been maintained for 2 weeks in April 1986. The event occurredt 3.1 km depth, and reactivated a seismic structure that had beenctive early in the testing (CSM-Report, 1989). Subsequently, injec-ion rate was lowered to 21 l/s for a long-term, constant-rate test.nother event of ML = 1.7 occurred in January 1988, several hun-red metres below the earlier event but was not reported as feltWalker, 1989).

. Injection case histories: sedimentary rocks

.1. Austria

.1.1. Upper Austria (South-German Molasse basin):The region of Upper Austria within 25 km of the German bor-

er is host to a cluster of five doublets that utilise heat fromhe Malm carbonate aquifer. These are Simbach-Braunau, Altheim,einberg, Oberndorf and St. Martin (Goldbrunner et al., 2007). Heree describe the first three of these sites. The only available stress

nformation in the area comes from a short (11 m) drilling-induced

ension fracture imaged in a well at the Simbach-Braunau sitehich indicates an SHmax orientation that is consistent with the–S regional stress trend (Reinecker et al., 2010). Stress magni-

udes are uncertain. The area has low natural seismicity, although

are shown by the bold circles; the ML 3.4 event that occurred just after shut-in onrred in the upper SSW corner of the microseismic cloud, 1–2 months after the wellrows in (a) denote the mean orientation of the maximum horizontal compressive

three events of ML 2.0 and 3.0 and intensity Io V have occurredwithin 30 km of the sites.

3.1.1.1. Simbach-Braunau:. This dual use doublet is constructed onthe border between Germany and Austria and provides districtheating to communities in both countries. Although the plant itselfis in Germany, it is included here under Austria because it shares asimilar geologic setting as the neighbouring plants in Upper Austria.The injection well, Simbach-Braunau Thermal 1, was drilled ver-tical to 1848 m in 1999 and completed with 113 m of 8–1/2 in.open hole in heavily fractured and karstified Malm which con-tains a fault (Goldbrunner, 2005b; Unger and Risch, 2001). TheMalm at this site is separated from basement by 100 m of theDogger formation. Formation pressure is artesian with a shut-inwellhead pressure of 0.23 MPa. The well is highly productive andyields 80 l/s on artesian flow with the choke fully open. The down-hole pressure change on artesian flow was only 0.06 MPa implyinga transmissivity of 0.04 m2/s (Goldbrunner, 2007). The productionwell, Simbach-Braunau Thermal 2, was drilled from the same padand deviated to reach the Malm some 2100 m from the first well ata vertical depth of 1942 m (3203 m MD). The well targeted a small-offset fault in the Malm (Goldbrunner, 2005c) and was completed as6–1/8 in. open hole (Goldbrunner, 2007). The system was commis-sioned in 2001 and operated with a submersible pump in Thermal2 that produced 74 l/s at 80 ◦C. The water was injected into Ther-mal 1 at 55 ◦C. Given the high transmissivity, it is unlikely that thepressure increase in the reservoir during the injection exceeded0.1 MPa. A 200 kWe ORC unit was added in 2009 that produced150 kWe net (Goldbrunner, 2010).

3.1.1.2. Altheim:. The site lies 15 km east of Simbach-Braunau.Development began in 1989 when the first well, Altheim Thermal1, was drilled vertical to 2472 m and reached basement near a fault(Pernecker, 1996). The fractured Malm extended between 2147 and2429 m, and was found to directly overlie the basement. The reser-

voir pressure was artesian and the well flowed at 11 l/s with fullyopen choke. Owing to clogging and ensuing technical difficulties,it was necessary to drill a side-track, Thermal 1a, from 1772 to2300 m TVD. This was 20 m closer to a fault where the rock was

4 therm

mTThitawhat8aasw

3Enh(st2atGTawa(ubhTa(

3

aTsa(

sai2eb2htidoad(p

r

0 K.F. Evans et al. / Geo

ore fractured, which resulted in a higher artesian-flow of 18 l/s.his increased to 46 l/s after acid stimulation (Goldbrunner, 2005b).he well was operated under artesian drive and supplied a districteating system from 1991 to1998 (Goldbrunner, 2005c). Thereafter

t was decided to counter declining reservoir pressure by injectinghe spent fluid. In 1997 the injection well was drilled deviated to

vertical depth of 2165 m (3078 m MD). Very high permeabilitiesere encountered on entering the Malm, possibly indicating a faultad been intersected (Pernecker and Uhlig, 2002). Well separationt the top of the Malm is 1600 m (Goldbrunner, 2005b). In 2000,he production flow rate was increased to the operating level of1 l/s using a submersible electric pump in Altheim Thermal 1a,nd a 500 kW ORC unit was installed. The fluid is produced at 105 ◦Cnd reinjected into Thermal 2 at 70 ◦C and 1.7 MPa wellhead pres-ure (Goldbrunner, 2005b). The system supplies 10 MWt during theinter months and generates electricity during the summer.

.1.1.3. Geinberg (Austria):. This dual use project is located 5 kmNE of Altheim and has a developmental history similar to itseighbour. The first well was a vertical hydrocarbon explorationole drilled in 1976 that entered Malm at 2127 m, near a faultGoldbrunner, 1999), but was terminated at 2166 m due to exces-ive mud losses. The well was reopened and successfully deepenedo 2180 m in 1978 and became Geinberg Thermal 1 (Goldbrunner,005b). The well produced 22 l/s of fluid at more than 100 ◦C underrtesian drive and thereafter supplied a small district heating sys-em. The system was extended to a doublet in 1998 by drillingeinberg Thermal 2, partly to counter declining reservoir pressure.his was deviated to reach Malm at a vertical depth of 2225 m and

distance of 1600 m from the first well (Goldbrunner, 1999). Theell was completed with 276 m of open hole, mostly in Malm, and

fter an acid stimulation the productivity index was 1.0 l/s/MPaGoldbrunner, 1999). An injectivity test of Geinberg Thermal 1sing hot water produced from Thermal 2 showed that 30 l/s coulde injected at 0.2 MPa (Goldbrunner, 1999). The doublet systemas been in operation since late 1998, with 25 l/s produced fromhermal 2 at 105 ◦C under buoyancy drive (Karytsas et al., 2009),nd 21 l/s injected into Thermal 1 at 30 ◦C under gravity driveGoldbrunner, 2005b).

.1.2. Bad Blumau (Styrian basin of south-east Austria):This multiple use doublet is located some 50 km east of Graz

nd exploits heat from Paleozoic dolomites at a depth of ∼2.5 km.here are no published stress measurements within 45 km of theite. Whilst seismic hazard is low, two events of intensity Io VIre reported to have occurred within 40 km in the past 159 yearsGrünthal et al., 2009).

The injection well, Blumau 1a, was drilled as a steeply inclinedide-track from a hydrocarbon exploration well in 1995. It runspproximately parallel to and within 180 m of a WSW–ENE trend-ng growth fault with more than 1 km of throw (Goldbrunner,005a) and is completed in heavily fractured dolomites that arencountered at 2583 m TVD. These lie above phyllites that overlieasement rocks (Goldbrunner, 1999). The production well, Blumau, was drilled vertical to 2843 m and completed with 475 m openole in the fractured dolomites. The open-hole section lies betweenwo faults that cut the wellbore. Well separation at reservoir depths ∼1800 m (Goldbrunner, 2005a). Pumping tests in one well pro-uce perturbations in the other and imply a far-field transmissivityf 5 × 10−5 m2/s (Goldbrunner, 1999). Near-well transmissivitiesre much higher, thanks in part to an acid stimulation, and pro-uction flow rates of up to 80 l/s were obtained during tests

Goldbrunner, 1999). The water is rich in CO2, which aids ‘artesian’roduction through the gas-lift effect.

The system began operation for heating usage in 1999 at a flowate of 30 l/s. Water is produced from Blumau 2 under artesian

ics 41 (2012) 30– 54

flow at a temperature of 110 ◦C. Almost all of the produced wateris injected into Blumau 1 at 50 ◦C or slightly greater temperatureyielding 7.6 MWt (Goldbrunner, 2005a). Injection pressure at thewellhead is less than 0.7 MPa (Goldbrunner, 2005b). Electricity gen-eration began in 2001 using a 250 kWe ORC unit (Legmann, 2003)to produce 180 kWe net (Goldbrunner, 2005a). The heating powerthen dropped to 5.1 MWt.

3.2. Denmark

3.2.1. Thisted:This district heating project is located in NW Jutland in the Dan-

ish basin. It exploits heat from the Upper Triassic Gassum sandstoneaquifer at 1.25 km depth that has a transmissibility of 10−10 m3

(Mahler, 1995). A large salt diapir structure exists immediately tothe NNW of the site (Mahler and Magtengaard, 2005), and so it ispossible that the sedimentary section is hydraulically decoupledfrom the basement. Unpublished analyses of borehole breakoutsin the injection well for the interval 804–1014 m reported in theWorld Stress Map database (Heidbach et al., 2010) indicate a meanSHmax orientation of N77◦E. Natural seismicity at the site is low.Only two documented historical events of estimated intensitiesIo V1 are recognised within 50 km, the nearest being 13 km away(Grünthal et al., 2009).

The project began in 1984 with the drilling of a 3287 m deepvertical exploration well, Thisted-2, which was plugged back to1273 m and a ∼37 m screen installed for production. The injectionwell of the doublet, Thisted-3, was drilled vertical to 1242 m andthe lowermost 50 m screened (Mahler, 1995). Well separation is1.5 km (Mahler and Magtengaard, 2005). The system was initiallycirculated at 10 l/s (Mahler and Magtengaard, 2010). In 1988, fol-lowing an improvement in the surface plant, the circulation ratewas increased to 41 l/s, yielding a total power of 4 MWt. The systemwas further upgraded to 7 MWt in 2001 by increasing the circula-tion rate to 56 l/s. The water is produced at 44 ◦C and cooled to 11 ◦Cbefore injection at a pressure of 1.7 MPa (Mahler and Magtengaard,2005).

3.2.2. Margretheholm (Copenhagen):The project supplies a district heating system from a doublet

completed in the Bunter formation at 2.5 km depth. No salt depositsare present between the Bunter and the Precambrian basement(Dong E&P, 2004). The natural seismicity in the area is low. Onlythree events with intensity Io V–VI are recognised as occurringwithin 50 km of the site since the 17th century (Grünthal et al.,2009).

The first exploration well of this project, Margretheholm-1,was drilled vertically to basement at 2677 m in 2002. Testingshowed the Bunter formation at 2.5–2.6 km had a productivity of22 l/s/MPa and was suitable for reservoir development (Mahler andMagtengaard, 2005). A second well, Margretheholm-2, was devia-tion drilled to 2745 m TVD in 2003 from the same pad so that wellseparation at reservoir depth was 1.3 km. Testing showed a produc-tivity of 17 l/s/MPa and communication with the first well (Mahlerand Magtengaard, 2010). The system began operation in late 2004.Water is pumped to surface at 65 l/s and 1.0–1.5 MPa by a down-hole pump (700 kW) at 650 m depth. It is cooled to 17 ◦C beforebeing injected at pressures up to 7.0 MPa (Mahler and Magtengaard,2010).

3.3. France

3.3.1. Paris basin (Paris):Exploitation of heat from carbonate units beneath Paris for space

heating began in the early 1970s and the system now is secondonly to Reykjavik (Iceland) in terms of size. The primary reservoir

therm

iTd82ai1riato

h2mtrs

3

3

mt

3lhs1aadmtfo2era

Ia(dNleN(diw2p0tta

wc

K.F. Evans et al. / Geo

s the mid-Jurassic Dogger aquifer composed of oolitic limestone.his is exploited over a large part of the eastern Paris region, atepths and temperatures ranging from 1450 to 2000 m and 56 to0 ◦C respectively (Ungemach and Antics, 2006; Ungemach et al.,005). Fifty-five doublets or triplets have been drilled of which 34re currently active (Lopez et al., 2010). The sedimentary sectionn the area is cut by N–S oriented normal faults (Cornet and Burlet,992). There are no reported stress measurements in the area. Theegional stress state is characterised by normal-to-strike-slip fault-ng with mean SHmax oriented approximately N145–150◦E (Cornetnd Burlet, 1992; Vidal-Gilbert et al., 2009). Natural seismicity inhe region is very low with no earthquakes reported within 50 kmf the centre of Paris.

Most well trajectories are deviated from a single drilling pad andave reservoir separations of about 1200 m (Ungemach and Antics,006). The average flow rates are 38–97 l/s, although operation ofost sites follows a seasonal cycle with peak flow rates in the win-

er (Lopez et al., 2010). Corresponding injection pressures usuallyange between 2.0 and 3.0 MPa with a maximum of 4.0 MPa at twoites (P. Ungemach, pers. comm., Oct 2010).

.4. Germany

.4.1. North-German BasinThere are several projects in this basin that mostly exploit

edium-enthalpy geothermal resources hosted in clastic forma-ions.

.4.1.1. Neustadt-Glewe/Waren/Neubrandenburg:. These plants areocated in the northeast of Germany and extract heat from theigh-porosity Rhätkeuper (Triassic) and Hettang (Jurassic) sand-tone aquifers at depths of 1.0–2.5 km. They were developed in the980s and were the first geothermal doublets in Germany (Seibtnd Kellner, 2003). Throughout the area, the sandstone aquifersre underlain by evaporites that include the extensive, thick salteposits of the Zechstein (Permian) (Seibt et al., 2005). The saltay hydraulically isolate and mechanically decouple the units of

he overlying section from the basement. Regional stress data fromormations below the salt indicate an approximately north–southrientation for SHmax and high deviatoric stress (Röckel and Lempp,003). The stress state in the overlying strata shows greater het-rogeneity (Röckel and Lempp, 2003). The natural seismicity in theegion is low with no recognised events located within 40 km ofny of the sites (Leydecker, 2009).

Neustadt-Glewe is the deepest and most westerly of the projects.t utilises the 40–60 m thick Contorta sandstone aquifer that has

porosity of 0.20–0.23 and a permeability of 0.5–1.0 × 10−12 m2

Seibt et al., 2005). The fluid is produced at 99 ◦C from the 2455 meep well NG-1 and injected at 60 ◦C into the 2335 m deep wellG-2, which is separated by1500 m from NG-1. The static fluid

evels in the wells are about 125 m below ground level (Poppeit al., 2000). Acid and hydraulic stimulations were conducted onG-2 in 1993, but details of the treatments could not be located

Heederick, 1997). The system began supplying up to 6 MWt to theistrict heating system in 1995 and a 210 kWe ORC generator was

nstalled in 2003. The initial flow rate of 31 l/s could be injectedith a downhole overpressure of 0.5 MPa (Seibt and Wolfgramm,

008). However, occasional decreases in injectivity due to mineralrecipitation and particle clogging led to temporary increases to.8 MPa. These increases could be remedied with soft acidizationreatments, implying the increased impedance was localised nearhe well. Thus it is probable that the formation pressures in the

quifer beyond the immediate vicinity of the well were unaffected.

Waren began supplying a district heating system in 1984, andas the first operational geothermal plant in Germany. The initial

onfiguration produced water from Contorta sandstone at 1560 m

ics 41 (2012) 30– 54 41

depth (reservoir temperature of 62 ◦C). The water was injected intothe Aalen sandstone at 1160 m depth through a well only 50 m dis-tant (Kabus and Jäntsch, 1995). The system was circulated at 14 l/susing a submersible pump. The wellhead pressure at the injectionwell reached 5 MPa (U. Reimer, pers. comm., Sept. 2010). In 1986 asecond injection well was drilled to 1580 m at a distance of 1.3 kmfrom the production well, and was screened in the Hettangian sand-stone, which lies just above the Contorta. Thereafter, the flow ratecould be increased to 17 l/s (Kabus and Jäntsch, 1995). Productiontemperature is 62 ◦C and injection temperature is no less than 45 ◦C(U. Reimer, pers. comm., Sept. 2010). Injection takes place undergravity drive. As the static water level in the injection well is 110 mbelow ground level, the downhole pressure change under injectionis less than 1.1 MPa (U. Reimer, pers. comm., Sept. 2010).

Neubrandenburg is the most easterly and shallowest of theprojects. In 1985 four boreholes were drilled to make two doubletsaccessing the Hettang and Postera sandstones, which are sandstoneunits above and below the Contorta respectively. The Postera pairwere N1 (production at 1250 m) and N3 (injection at 1250 m) wells,and the Hettang pair were N2 (production) and N4 (injection at1120 m) wells. The two injection wells were 5 m apart and had astatic water level of −62 m (U. Richlak, pers. comm., Sept. 2010). TheHettang production well N2 had technical problems. Thus, between1989 and 2001 the system was circulated at 19 l/s with produc-tion from well N1 (Postera) and injection into the 1.2 km distantwell pair N3 (Postera) and N4 (Hettang) at 0.5 MPa wellhead pres-sures (U. Richlak, pers. comm., Sept. 2010), implying a downholeoverpressure of 1.1 MPa. Production temperature was 54 ◦C.

In 2001, the system was reconfigured to allow waste-heatfrom gas-turbine electricity production in summer to be storedin the Postera reservoir and reused in the winter. Borehole N4(Hettang-injection) was deepened to reach the Postera, and theexisting Postera injection well, N3, was back-filled leaving a doublet(N1–N4) accessing the Postera. Operation of the thermal storagesystem began in April 2004. In summer, water is produced fromN4 at 28 l/s and 54 ◦C, heated to 80 ◦C with waste-heat from a gas-turbine generator, and the water injected into N1. In winter theflow is reversed with production from N1 at 28 l/s and 70–80 ◦C,and injection into N4 at ∼45 ◦C (Seibt et al., 2010). Wellhead injec-tion pressures are always less than 0.5 MPa implying a 1.1 MPadownhole overpressure (U. Richlak, pers. comm., Sept. 2010).

3.4.1.2. Gross Schönebeck. This pilot project located north of Berlinis designed to evaluate the possibility of using the EGS concept toextract heat from low-permeability sedimentary rocks. The reser-voir rocks are the sandstones and conglomerates of the Rotliegendformation and the volcanic rocks of the uppermost Carboniferous,all of which lie below the Zechstein salt layer. Hydraulic tests indi-cate a Shmin/SV ratio of 0.52 in the sandstones at 4150 m depth(Huenges et al., 2006), indicating that the stress state is criti-cal (Moeck et al., 2008). SHmax is oriented N18◦E (Zimmermannet al., 2008). Natural seismicity in the area is low, the closest eventbeing a ML 2.7 event at a distance of 50 km that occurred in 1736(Leydecker, 2009).

The largest volume injection tests were performed in 2003,after the first well (E GrSk3/90) was deepened to 4309 m. Some10,000 m3 of water was injected at rates up to 80 l/s and downholepressures that exceeded Shmin by up to ∼5 MPa. Most fluid enteredthe volcanics below the sandstones. A surface seismic networkconsisting of six 3-component stations in shallow boreholes wasoperational at the site during the injections, but detected no eventsattributable to fluid injection (M. Weber, pers. comm., Oct. 2010).

Subsequently, a second well (Gt GrSk4/05) was drilled and stim-ulation injections performed in the same formations, the largestinvolving the injection of 13,000 m3 of water and proppant intothe volcanic unit at rates of up to 150 l/s and wellhead pressures of

4 therm

5cradrhtt

3sttorii0t(csvgof

2d12dprs23stoh

3esTtB4lS

tiitidtwtad

2 K.F. Evans et al. / Geo

9 MPa (Zimmermann et al., 2008). Such volumes and flow rates areomparable to those used in the hydraulic stimulation of crystallineeservoirs. Seismic activity was monitored by the surface networkugmented by a downhole 3-component sensor placed at 3800 mepth in the first borehole. Only 70 events with magnitudes in theange ML −1.9 to −1.1 were detected, and then only by the down-ole sensor (Kwiatek et al., 2008). This low level of activity despitehe injection of large volumes at high pressure is consistent withhe growth of hydrofractures within the reservoir.

.4.1.3. Ketzin, Germany (CO2 injection):. This CO2 injection testite is located 40 km west of Berlin and is designed to monitorhe behaviour of CO2 injected into the Triassic Stuttgart forma-ion. Natural seismicity is very low, the nearest historical eventsf intensity Io IV or greater being more than 100 km distant. Theeservoir is an 80 m thick fluvial sandstone saline aquifer that hasts top at 630 m depth. It is highly heterogeneous, with porositiesn the range 0.05–0.35 (Norden et al., 2010), and permeabilities of.05–1.0 × 10−12 m2 (Würdemann et al., 2010). The static forma-ion pressure is 6.3 MPa at 642 m, and the temperature is 35 ◦CWiese et al., 2010). Thus, the CO2 is a gas under formation PTonditions. The caprock consists of fine-grained, clay-rich clasticediments of the Weser Formation. Further containment is pro-ided by the Rupelian mudstone which served as the caprock to aas storage reservoir at depths of 250–400 m that was seasonallyperated from 1960s until 2000 (Juhlin et al., 2007). There are noaults within the CO2 reservoir (Juhlin et al., 2007).

The reservoir is penetrated by three vertical wells drilled in007: one 750 m deep injection well (Ktzi-201); and two 800 meep monitoring wells (Ktzi-200 and Ktzi-202) located 50 and00 m from the injector in orthogonal directions (Prevedel et al.,008). All three wells are instrumented with fibre-optic cables foristributed temperature sensing and supporting pointwise tem-erature and pressure measurements, and a 15 electrode verticalesistivity array (Prevedel et al., 2008). The CO2 is injected throughcreens at a depth of 650 m. CO2 injection commenced in June008 at a rate of about 2 ton/h which was gradually increased to

ton/h over 9 months. Pressure at the formation depth increasedteadily to 8.1 MPa after 1 year, which is 1.7 MPa above the forma-ion pressure (Würdemann et al., 2010). A seismic network has beenperational at the site since September 2009. No seismic eventsave been felt at the site (S. Lüth, pers. comm., March 2011).

.4.1.4. Horstberg. This is a geothermal pilot facility designed tovaluate the possibility of using single wells to extract heat fromediments in the North-German Basin (Wessling et al., 2009).he site is located north of Hannover, and the target produc-ion formations are low-permeability sandstone beds within theuntsandstein unit that lies above the Zechstein salt beds at about000 m depth (Orzol et al., 2005). Natural seismicity in the region is

ow, although a very rare event of magnitude ML 4.0 occurred nearoltau, some 26 km from the site in 1997 (Leydecker, 2009).

Massive water stimulations were performed through perfora-ions into several sandstone beds in 2003–2004. These injectionsnvolved volumes of up to 20,000 m3 and rates of 50 l/s. Wellheadnjection pressures of up to 32 MPa were higher than expected onhe basis of other injection operations in the area, possibly reflect-ng stress heterogeneity in the formations above the salt beds due tooming (Jung et al., 2005). Fluid from post-stimulation productionests was injected into the Kalkarenit formation at 1200 m depth,

hich is a karstic carbonate aquifer. The operations were moni-

ored by an 8-station seismic array of 3-component sensors with detection threshold of magnitude ML 0. Only seven events wereetected, and all were too small to be located.

ics 41 (2012) 30– 54

3.4.2. South German–Austrian Molasse basinThis basin is host to numerous medium-enthalpy (100–140 ◦C)

geothermal projects. Most are centred on the Munich area(Schubert et al., 2007; Schulz et al., 2007), the remainder being dis-tributed over a wide area that extends into Austria (see Section3.1.1) (Goldbrunner et al., 2007). The majority of projects targetthe high-transmissibility zones within the Malm carbonate aquifer,which may be either karstic or fault-related, and many are dou-blets that involve injection. Productivities are high, and injectionand production pressures are typically less than a few MPa (Schulzet al., 2007). Stimulation operations tend to use acid rather thanhigh-rate hydraulic injections. For doublets, the production andinjection points within the Malm are two or more kilometres apart,and thus the flow field between the wells within the reservoir canbe complex, and may involve faults.

3.4.2.1. Straubing:. This site lies on the Danube some 10 km southof the Danube fault and 100 km northeast of Munich. It was oneof the first doublets to become operational in the basin. The near-est stress measurement in the World Stress Map (WSM) databaselies 60 km to the south and indicates a NS-orientation for SHmax(Reinecker et al., 2010). The area has low natural seismicity, nearestdocumented earthquake is located at a distance of 33 km and hadan intensity IO V (Leydecker, 2009). The production well, Therme I,was drilled vertical to 824 m depth in 1990 and entered fracturedand karstic Malm at 708 m. The well was artesian with a static well-head pressure of 0.41 MPa (Goldbrunner, 2007) and produced 29 l/sof water at 37 ◦C on unchoked flow. The second borehole, ThermeII, was drilled vertical to 885 m depth at a distance of 1680 m NNWof Therme 1 and penetrated Malm at 715 m, Dogger-age sandstoneand clay (Opalinus) at 842 m and crystalline basement at 863 m.Most of the discharge during drilling came from the Dogger sand-stone. Static wellhead pressure was 0.5 MPa. A production test witha downhole pump yielded 39 l/s with the water level at 203 m (pro-ductivity of ∼15 l/s/MPa), but only 5 l/s could be injected at 2.0 MPawellhead pressure (injectivity of 3.3 l/s/MPa) (Goldbrunner, 2007).To improve injectivity, the 9–5/8 in. well was plugged at 600 m,and an 8–1/2 in. diameter side-track was drilled towards a frac-ture zone that lay to the East. The casing shoe was set when theMalm was reached at 721.8 m TVD, and the hole continued as a6 in. with a sail angle of 77◦ to produce 339 m of open hole in theMalm (Goldbrunner, 2007). The initially poor injectivity was sig-nificantly improved by an acidization treatment with injection at22 l/s at up to 5.0 MPa. Venting produced significant quantities ofclastic material, presumably from the natural fractures that inter-sected the well. Following the treatment, 45 l/s could be injected ata wellhead pressure of 1.9 MPa (injectivity of 32 l/s/MPa). The sys-tem was commissioned in 1999 and operated with 21 l/s of waterproduced from Therme I at 36 ◦C and 19 l/s injected into ThermeIIa at 12 ◦C yielding 1.9 MWt for district heating. Later, the produc-tion flow rate was increased to 45 l/s of which 40 l/s was injected at14 ◦C at a wellhead pressure of 1.8 MPa. There are no reports of feltseismicity associated with this site during its 12 years of operation.

3.4.2.2. Munich area:. In the Munich area, the top of the Malm topdeepens from 1.6 km in the north to 3.5 km in the south, and the unitdirectly overlies the crystalline basement. The stress state in theMalm is strike-slip/thrust with SHmax oriented approximately N–S(Reinecker et al., 2010). No natural seismicity has been recordedwithin 40 km of Munich since records began to be kept in the 19thcentury. Formation pressure in the Malm is sub-hydrostatic, the

equilibrium water levels in the wells declining from 85 m belowground surface in the north to 250 m in the south. Since transmis-sibilities tend to be relatively high, the pressure increase aboveformation pressure in the injection wells at reservoir depth is in

therm

ap

ww(elbeMtSa2w

wi2shstddtfaailoicowwOm0b2(ldt2aeti(

3

oTk(iR

3lt

K.F. Evans et al. / Geo

ll cases less than 5.0 MPa under operational conditions (T. Fritzer,ersonal comm. December 2009).

The construction of doublets in the Munich area began in 2003ith the commissioning of the system at Unterschleissheim whichas soon followed by doublet systems at Riem (2004), Pullach

2005) and Unterhaching (2007) (Schubert et al., 2007; Wolfgrammt al., 2007). The operational parameters of these systems areisted in Table 1. The number of doublet and triplet systemsuilt or planned has accelerated since then, including the deep-st hydrothermal system in Europe at Sauerlach to the south ofunich, which will produce 140 ◦C water from 4.2 km depth for dis-

rict heating and electricity production in 2012 (Pletl et al., 2010).ystems commissioned after 2007 are not included in Table 1 tovoid introducing a local bias. Of the many projects in operation at012, the one at Unterhaching is located near an area with recent,eakly felt seismicity. Thus this project is described in detail.

At Unterhaching, the production (GT1a) and injection (GT2)ells have vertical depths of 3.35 and 3.59 km respectively and

ntersect the Malm at a vertical depth of 3.0 km (Wolfgramm et al.,007). The wells are 4 km apart at the Malm depth, and inter-ect different faults (E. Knapek, pers. comm., June 2010). The wellsave high injectivity/productivity, and pressure perturbations areeen to propagate between the wells in 24 h, presumably throughhe fault system (Wolfgramm et al., 2007). Under operational con-itions, the system to date has been circulated at 120 l/s, theownhole pressure excess above formation pressure in the injec-ion well at this rate being 2.5 MPa. The system began operationor district heating in October 2007 and electricity production wasdded in February 2009. In February 2008, 5 months after oper-tions began, an earthquake of magnitude ML 2.3 was detectedn this formerly aseismic region by the German Regional Seismo-ogical Network. The event was located within several kilometresf Unterhaching and was scarcely felt by the local population liv-ng within 5 km of the injection well GT2 (J. Wassermann, pers.omm., June 2010). A further three events of magnitude ML 2.1–2.4ccurred over a 3-week period in July 2008, the largest of whichas felt by the local population. In July 2008 a 3-station local net-ork was established in the area to improve the location accuracy.n 2nd February 2009, shortly after electricity production com-enced, a series of seven earthquakes of magnitudes between ML

.7 and 2.1 was detected by the local network (Kraft et al., 2009)ut none were felt by the local population (Bayerische-Landtag,009). No change in circulation parameters occurred at this timeE. Knapek, pers. comm., June 2010). Analysis of the data using aocal velocity model indicates that all hypocentres are located at aepth of 3.6 ± 1.5 km and some 0.5 ± 0.3 km west of the Gt-2a injec-ion interval, which lies at a vertical depth of 3.1–3.6 km (Kraft et al.,009). The larger events were also recorded by the national networknd found to have waveforms that were very similar to those of thearlier events, suggesting that all events occurred within 1 km ofhe open-hole section of the injection well (Kraft et al., 2009). Thenjection interval includes a fault with a vertical throw of 238 mWolfgramm et al., 2007).

.4.3. Upper Rhine GrabenThe graben hosts several projects at various stages of devel-

pment that seek to develop hydrothermal reservoirs in faultedriassic sandstone (Bundsandstein) and carbonate (Muschel-alk) units or the immediately underlying weathered basementBaumgärtner et al., 2006). The project at Landau exploits both sed-ments and basement and is described in Section 2, whereas that atiehen can be found in this section under Switzerland.

.4.3.1. Bruchsal, Germany. This dual use geothermal project isocated some 20 km northeast of Karlsruhe, on the eastern flank ofhe graben. It is one of the first geothermal projects in the graben,

ics 41 (2012) 30– 54 43

the first borehole being drilled in 1983. Unpublished boreholebreakout analyses for the geothermal site reported in the WorldStress Map database (Heidbach et al., 2010) indicate SHmax is ori-ented N130–140◦E. Stress magnitudes are uncertain. The site has alow seismic hazard level of 0.07 g, although 10 small events of mag-nitude ML up to 3.3 with intensities up to Io 5 have occurred within25 km of the site since 1970 (Leydecker, 2009). The largest histori-cal event within 30 km of the site is the 1948 Forchheim earthquakethat is 26 km distant and has been assigned a peak intensity of IoVII (Ahorner et al., 1970a).

An exploration borehole, GB1 was drilled vertical to 1932 m in1983 and discovered an exploitable reservoir in the lower Trias-sic Bunter and underlying Permian Rotliegend sandstones whichare cut by elements of the main boundary fault of the graben(Herzberger et al., 2010; Kölbel et al., 2010). In 1985, a second ver-tical well was drilled to 2450 m depth some 1400 m southwest ofthe first well to target the same formations (Kölbel et al., 2010). Thesystem was circulated for 6 weeks at up to 15 l/s in 1987, but tech-nical difficulties and economic considerations led to the suspensionof the project in 1990. It was reactivated in 2002 using the deeper,hotter well, GB2, as the production well. No hydraulic stimulationswere conducted. Static water levels in the wells were 60 m belowthe ground surface. A 550 kWe Kalina plant was installed in 2008and has been in operation since mid-2009. A pump in GB2 produceswater at a rate of approximately 24 l/s and a temperature of 118 ◦C.The cooled water is injected into GB1 under gravity producing anincrease in downhole pressure of approximately 0.5 MPa above thestatic pressure (Kölbel, 2010). There are no reports of felt micro-seismic events that can be ascribed to the operation of the system.A 4-station seismic monitoring network has recently been installedto detect and locate any induced microseismic events that are toosmall to be felt (Kölbel et al., 2010).

For Landau see Section 2.2.3, and for Riehen see Section 3.9.1.

3.5. Italy

3.5.1. Tuscan-Latium geothermal areas, ItalyThe region lies in western Italy between Rome and Pisa, and

includes several geothermal sites that have been explored and insome cases developed since 1970. The area is characterised by highgeothermal gradients, sometimes exceeding 100 ◦C/km, that reflectthe presence of shallow magmatic bodies. The tectonic settingis transcurrent/transtensional with predominant strike-slip fault-ing (Brogi and Fabbrini, 2009). The fields that have been exploredinclude Larderello-Travale and Monte Amiata in Tuscany, and Lat-era, Torre Alfina and Cesano in Latium. All fields except Cesano areassociated with significant natural seismicity (Batini et al., 1980b),suggesting that the reservoirs are likely to be critically stressed. Inthe 1970s, local seismic networks were installed at the sites in orderto systematically assess the seismic response of the various reser-voirs to fluid injection. In most cases the networks were installedprior to injection so that induced events could be distinguishedfrom the background. The networks were operated throughout theexploration phase (1977–1992) and remain active in those fieldsthat are still in production (Larderello-Travale and Monte Amiata).

3.5.1.1. Larderello-Travale. (Shown as “Larderello” in Fig. 1). Thisgeothermal area has long produced steam from folded anhydritesand carbonates that overlie metamorphic basement at a depth of2 km. In the early 1970s, injection of cold condensate from thepower plants was initiated in order to recharge the upper reser-voir, and a seismic monitoring network was installed, in part to

monitor the impact of the injection.

The analyses of seismic data from 1978 to 1982 are presentedby Batini et al. (1980a, 1985). The area has a long history of seis-micity, and therefore many, if not most of the events are likely to

44 K.F. Evans et al. / Geothermics 41 (2012) 30– 54

0 10 20 30 40 50 60

Flow

rate

(l/s

)W

ell-h

ead

pres

sure

(MPa

)N

o of

eve

nts/

hour

10

20

30

40

0

10

20

30

40

60

50

70

80

90

00 10 20 30 40 50Hours Hours Hours

June 1981 May 1982February 1982

10

20

30

40

0

(b)(a) Latera Well L1 injection records and recorded microseismicity Hypocentre locations

X (km)

Y (k

m)

0.0

0.4

0.8

0.0

-0.4

-0.8

0.4-0.4 0.8

Legend June 1981 earthquakes February 1982 earthquakes May 1982 earthquakes Geothermal well L1

0.0 0.4-0.4 0.8

Well L1

1.02.03.04.05.0D

epth

(km

)

X (km)Acidification

2468

10

02468

10

02468

10

0

255075

100125

0255075

100125

0255075

100125

0

0 20 40 60 80 100

Fig. 3. (a) Injection parameters and recorded seismicity (in events/hour), for the injection tests in well L1 at Latera during 1981–1982 (After Carabelli et al., 1984). (b)Hypocentres of earthquakes associated with the injections. The X- and Y-axes point east and north respectively. The locations differ slightly from those given by Carabelliet al. (1984) as station elevation corrections were included in the inversion conducted with the Hypoellipse program, although the velocity model was unchanged. The eventsc n erroa rded o

btweoif1

3

3rAyeeeal

1daui1twnit

luster closely about the well. However, it should be noted that the formal locatiolso for the horizontal. This is because clear P- and S-wave arrivals were rarely reco

e natural. The 5 years of data show large spatial-temporal varia-ions in event rate and b-value. The events are shallower than 8 km,ith 75% located between depths of 3.0 and 5.5 km. The maximum

vent size approached ML 3.2. A clear correlation between volumef water injected and event count is seen, although most of thenduced events appear to be of small magnitude. No change to therequency of events of magnitude ML ≥ 2.0 was evident (Batini et al.,985).

.5.1.2. Monte Amiata area. Listed in Section 2.4.1 on igneous rocks.

.5.1.3. Latera. The reservoir is hosted by fractured, carbonateocks at 0.6–2.0 km depth and has a temperature of 200–230 ◦C.

10-station seismic network began operation late in 1978, about aear before the first injection (Batini et al., 1980b). Several naturalvents having magnitudes in the range ML 0.6–1.7 were detectedach month at distances of 20–35 km from the reservoir (Batinit al., 1980b). We describe experiments at this site in some details they have not been published in the international geothermaliterature.

One of several injection episodes took place during March–April980. A total of 30,000 m3 of water was injected into the 1.4 kmeep well L2 during two periods lasting 8 and 10 days, implyingverage flow rates of 35–45 l/s (Moia, 2008). Wellhead pressure isncertain. Some 24 events with magnitudes ML 1.5–2.0 occurred

n two localised clusters during the 2-month period (Batini et al.,980b). One cluster was located only 200 m south of the injec-ion point at a similar depth, suggesting that the microearthquakes

ere induced. It is of note that a magnitude ML 2.9 event occurredear the L2 well on December 9, 1984 when it was being used to

nject fluid produced from the 2 km distant well L3D. Unfortunately,he circulation parameters are unknown (Moia, 2008). This is the

rs from the program are mostly larger than 1 km for the vertical, and occasionallyn all stations.

largest event that is thought to have been induced by geothermaloperations at Latera.

A more complete investigation of the seismic response of thereservoir to injection was conducted in the 2.8 km deep wellL1 between June 1981 and May 1982. The records for the testsequence are shown in Fig. 3a. The details of the well completionare uncertain although it is known to have been open to theformation in a fracture zone at a depth of 1.7 km (Carabelli et al.,1984). Three separate injections of durations 17–102 h wereperformed months apart at progressively increasing flow ratesof 15, 25 and 83 l/s. The corresponding mean wellhead pressureswere 5.5, 5.0 and 9.0 MPa (Moia, 2008). Wellhead pressure reached7.0 MPa at the start of the first test but then declined to 5.0 MPaover the 61 h of the test. Microseismicity near the well began aftera few hours and stopped after 35 h (Fig. 3b). A total of 223 eventswere recorded with magnitudes less than ML 0.5 (Carabelli et al.,1984; Moia, 2008). The second injection was conduced at a higherrate but comparable wellhead pressure in May 1982 and lasted102 h. Microseismic activity began only after 55 h of injection, bywhich time 1.5 times the net volume of the first test had beeninjected. Sporadic seismic activity persisted until shut-in, resultingin 148 events with a maximum magnitude of ML 0.4. The third andhighest-rate injection was conducted in May 1982 and lasted 17 h.Microseismicity at a high-rate began almost immediately but sub-sided after 6 h. A total of 370 events were detected, the maximummagnitude being ML 0.5 (Carabelli et al., 1984; Moia, 2008). Thehypocentres from all three injections are located between 150 and1500 m from the well, and at depths between 1.5 and 2.0 km, which

is close to the depth of a fault that intersects the well. However, theformal location error is large (see caption to Fig. 3). It is clear thatthe seismicity was induced and possibly reactivated the fault. Thewell was treated with acid shortly after the third injection which

K.F. Evans et al. / Geothermics 41 (2012) 30– 54 45

0 5 10 15 20 25

0Wel

l-hea

d pr

essu

re (M

Pa)

No.

eve

nts/

hour

4

8

0 10 20 30 40 50 60 700 5 10 15 20Hours

1-2 December 1981 21-22 December 198118-19 December 1981

Acidification

begin 17:00on 1.12.81

end 9:00on 2.12.81

Max pressure =14.2 MPa

HoursHours

begin 16:00on 21.12.81

end 11:00on 22.12.81

begin 6:00on 18.12.81

end 8:00on 19.12.81

20

0

40

60

Flow

rate

(l/s

)20

0

40

60

20

0

40

60

12

16

0

4

8

12

16

10

20

30

40

0

10

20

30

40

0

10

20

30

40

0

(a) Latera Well L6 injection records and recorded microseismicity Hypocentre locations(b)

1-2 December 1981

21-22 December 198118-19 December 1981

X (km)

Y (k

m)

0.0

0.0

-0.4

-0.6

-0.2

-0.8

-1.0

-1.2

0.2-0.2 0.4 0.6 0.8 1.0 1.2

Well L6

Well L6

1.02.03.04.05.0D

epth

(km

)

X (km)0.0 0.2-0.2 0.4 0.6 0.8 1.0 1.2

1.4 km depth volcanics1.4 km depth volcanics 1.7 km depth carbonates 1.7 km depthcarbonates

Fig. 4. (a) Injection parameters and recorded seismicity (in events/hour), for the injection tests in well L6 at Latera during 1981 (after (Carabelli et al., 1984)). (b) Hypocentresof earthquakes associated with the injections. The X- and Y-axes point east and north respectively. The locations differ slightly from those given by Carabelli et al. (1984)a Hypol rmal lo

i6

arFdwnwbanwismb2aMadbdcc

opn

3vt∼

s station elevation corrections were included in the inversion conducted with theinear structure extending to the SE from the well. However, as for well L1, the foccasionally also the horizontal.

mproved the injectivity. A subsequent injection at 111 l/s and.0 MPa wellhead pressure produced no detected events (Fig. 3a).

Further studies of the effect of injection on local seismicityt Latera were conducted in well L6 in December 1981. The testecords and the corresponding induced seismicity are shown inig. 4. The first test was performed into volcanic rocks at 1.4 kmepth with the injection of formation fluid at a rate of 66 l/s and aellhead pressure of 7.5 MPa. About 20 events were observed, all ofegative (i.e. very small) magnitude. The second suite of injectionsas conducted on 18–19 December at a depth of 1.7 km into car-

onate rocks. The sequence began with three short ∼2 h injectionst a rate of 8 l/s and wellhead pressure of 13–14 MPa that culmi-ated in an acid treatment and a step-increase in rate to 28 l/s andellhead pressure of 14.5 MPa for 1 h (Moia, 2008). Microseismic-

ty began as soon as the flow rate was increased and stopped withhut-in. A 24 h injection at 28 l/s was then performed. Microseis-ic activity began after 10 h, even though wellhead pressures were

elow 13 MPa, and terminated with shut-in. The final injection of0 h duration was conducted some 2 days later at a rate of 40 l/snd mean wellhead pressure of 14.5 MPa (Carabelli et al., 1984).icroseismicity began after 11 h, reached a peak just before shut-in

nd continued for several hours afterwards. Some 196 events wereetected, 28 after shut-in. Most had negative magnitude, the largesteing ML 0.8. The hypocentre locations lay to the SE of the well atistances of 200–1500 m (Fig. 4b), although the inferred locationsould be significantly influenced by poor network geometry (seeaption to Fig. 4).

The field was subsequently developed and had up to 14 wellsperational from 1999 to 2003 (Bertani, 2005). The wells werelugged in 2008 because of problems related to gas emissions, andot to microseismicity.

.5.1.4. Torre Alfina. This field lies 10 km north of Latera. The reser-oir is a fractured limestone at 0.5–1.7 km depth and 140–150 ◦Cemperature (Billi et al., 1986). Borehole RA1 was drilled to2710 m and a sequence of injection tests performed with fresh

ellipse program, although the velocity model was unchanged. The events define aocation errors from the program are mostly larger than 1 km for the vertical, and

water in January and February 1977 with injection rates in therange 20–40 l/s and wellhead pressures of up to 1.2 MPa. A 4-stationtemporary microseismic network that included three 3-componentinstruments recorded 177 events close to the well (i.e. at 1.4–3.3 kmdepth and 0.2–2.0 km distance). The largest event had a magnitudeML 3.0 and was felt by the local population (Batini et al., 1980b;Moia, 2008). The events occurred only at flow rates greater than25 l/s when injection pressures exceeded 0.7 MPa, and ceased soonafter injection was stopped. Batini et al. (1980b) report that theinjection of fluid produced from well A14 into well A4 under gravitydrive did not result in a detectable change in seismicity.

3.5.1.5. Cesano. This 250 ◦C brine reservoir, located north of Rome,is hosted by fractured carbonates at a depth of 1.5–3.0 km (Billiet al., 1986). A 5-station temporary seismic network was installedin May 1978 and was upgraded to a permanent network in 1979(Batini et al., 1980b). Dispersed, low-level natural seismicity occursat depths of 6–12 km.

A 2 km deep well RC-1 was drilled in 1978 and short injectiontests conducted once the temporary network had been installed.The first test involved a 1-day injection of water at 28 l/s. Wellheadpressure rose steadily from 3.5 to 7 MPa. A few events of maxi-mum magnitude ML 1.6 occurred near the well at the start of thetest, but none thereafter (Batini et al., 1980b). The second injectionwas conducted using a flow rate of 15 l/s sustained for 1.5 days.Wellhead pressure again steadily rose from 4 to 7.5 MPa. Seismic-ity began after 1 day when wellhead pressure reached 7 MPa andcontinued until shut-in. Batini et al. (1980b) indicate that the eventsreached magnitude ML 2.0 and were mostly located within 400 mof the injection well. However, they were recorded on relativelyfew stations and it is likely that the location uncertainty is large.Nevertheless, the events were clearly triggered by injection.

A further set of experiments was conducted in September 1981and April 1982 when a very dense seismic network was opera-tional. Circulation tests with production from well RC-1 (sometimesreferred to as C-1) and injection into well C-5 at rates of up to

4 therm

4b

afhidAciar

3

3

oDTw(aaa1

bdKtta41essbbmp2oaiabIo

3

3

gSsaftSomt

6 K.F. Evans et al. / Geo

2 l/s under gravity drive produced no seismic events that coulde identified as triggered by injection (Cameli et al., 1983).

To summarise the results for the Tuscan-Latium geothermalreas, the impact of injection on seismicity at the various sites is asollows: the effect at Monte Amiata is difficult to assess because ofigh background seismicity (see Section 2); at Larderello-Travale,

njection enhances the number of small-magnitude events butoes not detectably affect the large-magnitude events; at Torrelfina, Latera and Cesano, dedicated injection experiments yieldedlear examples where injection has induced seismicity up to andncluding magnitude ML 3.0. However, the injection experimentslso yielded examples where injection produced negligible seismicesponse.

.6. Lithuania

.6.1. Klaipeda:This district heating project is located on the Baltic Sea coast. The

bjective was to produce 167 l/s of water at ∼40 ◦C from a Lowerevonian Viesvile formation at depths between 990 and 1118 m.he unit consists of sandstone of porosity 0.2–0.3 interspersedith clay-rich packets, and has a permeability of 0.2–6.2 × 10−12 m2

Zinivicius et al., 2003). The basement lies at about 2200 m. Therere no published stress measurements within 50 km of the site. Therea has low natural seismic hazard. The nearest event in the cat-logue of Grünthal et al. (2009) is a Mw 4.2 event at a distance of00 km.

Construction on the system began in 1996. Initially two neigh-ouring vertical production wells KGDP-2P and KGDP-3P wererilled to 1128 and 1225 m respectively, and one injection well,GDP-1I was drilled to 1228 m at a distance of 800–1000 m from

he production wells (Zinivicius et al., 2003). However, it was foundhat the injectivity of the well was insufficient to dispose of fluidt the desired rate of 167 l/s at the maximum injection pressure of.0 MPa. Therefore, a second injection well, KGDP-I4, was drilled to128 m depth even further from the production wells (Ziniviciust al., 2003). All wells are completed with 9–5/8 in. casing andcreened between 1028 and 1128 m. The fluid has 95 g/l dissolvedolids and low gas content. The system began operation in 2001ut capacity declined because of decreasing injectivity, primarilyecause of mineral precipitation in the surface lines, well and for-ation (Seibt and Wolfgramm, 2008). Periodic remedial operations

roduced only temporary improvement (Seibt and Wolfgramm,008). System operation was suspended in 2007 to perform a majorverhaul because flow rate had dropped from 97 to 39 l/s (Ziniviciusnd Sliupa, 2010). Operation resumed in November 2008 withnjection rate into KGDP-I4 increasing from 30 to 50 l/s (Ziniviciusnd Sliupa, 2010). Injection pressures during operation could note determined but presumably approach the design limit of 4 MPa.

nformation is unavailable as to whether a seismic network wasperational and whether any earthquakes were felt.

.7. Norway

.7.1. Sleipner, North Sea, Norway (CO2 injection):This is a pioneering project of Statoil to remove CO2 from natural

as produced from the Sleipner-Vest gas reservoir underlying theleipner-A platform in the North Sea and inject it into the exten-ive, high-permeability, 200 m thick Utsira saline aquifer that lies at

depth of 800–1100 m (Torp and Gale, 2004). The Sleipner-A plat-orm is located near the eastern margins of the southern part ofhe Viking Graben, which is characterised by moderate seismicity.

everal events of magnitude ML 2–3 have occurred within 50 kmf the platform over the past 20 years (ISC, 2010). The Utsira for-ation is a hydrostatically-pressured, unconsolidated sandstone

hat has an estimated porosity of 35–40% and a permeability of

ics 41 (2012) 30– 54

1–8 × 10−12 m−12 (Baklid and Korbøl, 1996; Torp and Brown, 2005).The well entered the Utira formation at 872 m true vertical depth(TVD), and was deviated so that the injection point is approximately3 km away from the platform complex. The lowermost 1365 m iscompleted sub-horizontal with 7 in. liner to 1086 m TVD, and is per-forated over a 138 m interval centred at 1015 m TVD (Hansen et al.,2005). At the P–T conditions of the Utsira, the CO2 is expected to bein the supercritical phase, which is supported by seismic and grav-ity measurements (Zumberge et al., 2006). Injection began in 1996at a rate of 1 Mton/year, or an average of 32 kg/s. The CO2 is injectedat a wellhead temperature of 25 ◦C and a pressure 6.2–6.4 MPa,and is thus close to the liquid/gas phase boundary (Hansen et al.,2005; Eiken et al., 2011). Consequently, downhole injection pres-sure above hydrostatic is uncertain (O. Eiken, pers. comm., April2011). The tables produced by Baklid and Korbøl (1996) suggesta value of 3.5 MPa above hydrostatic pressure. This is probablyan upper bound: other evidence suggests pressure may be onlymarginally above hydrostatic (Eiken et al., 2011). Possible localseismic activity is not monitored at the site. However, there is noevidence from the regional networks of seismicity associated withthe CO2 injection operations (T. Torp, pers. comm., March 2011).

3.8. Poland

3.8.1. Bialy-Dunajec/Banska (Podhale basin):The geothermal plant in the basin is located near Banska where

a doublet began operation in 1992. This was expanded to a 4-wellsystem in 2001, primarily for heating. The reservoir is a heav-ily fractured, karstified, overthrust Triassic carbonate overlain bypost-tectonic, Eocene nummilitic-limestone at depths of 2–3 km(Kêpinska, 2000). The boreholes lie close to the major lineamentof the Bialy-Dunajek fault, which contributes to the high degree offracturing (Kêpinska, 2000; Wieczorek, 1999). The area has moder-ate seismic activity. Events of intensities up to Io V–VI are commonwithin 50 km of the site, with occasional but rare events with inten-sities up to Io VII. Intensity distributions show that at least someevents occur above the basement at depths of a few kilometres(Guterch et al., 2005). An earthquake of magnitude MW 4.5 that pro-duced intensities up to Io 7.0 occurred only several kilometres fromthe site in 2004 (Wiejacz and Debski, 2009). The focal mechanismof the event indicates normal faulting with Shmin oriented approx-imately WNW to NW (Wiejacz and Debski, 2009). The proximityof the event to the site suggests the stress state in the area is criti-cal. The relationship of the earthquake to geothermal operations isunclear as there was no local seismic network.

The production well of the initial doublet, Banska IG-1, wasan exploration borehole drilled to 5261 m in 1981. The well wascompleted in the Triassic/Eocene carbonates by perforating overthe interval 2588–2683 m. Static wellhead pressure was 2.7 MPa,and unchoked artesian-outflows of 17 l/s were obtained (Kêpinska,2003c). The injection well of the doublet, Bialy Dunajec PAN-1,was drilled 1220 m distant from the producer in 1989, and wascompleted in the same unit between 2117 and 2394 m (Kêpinska,2003a). Static wellhead pressure was 2.4 MPa. The doublet wasoperated between 1992 and 2001 at flow rates up to 17 l/s and injec-tion wellhead pressures up to 2.5 MPa (B. Kêpinska, pers. comm.,Oct. 2010). Production temperature was up to 80 ◦C and injectiontemperatures were no less than 50 ◦C (Bujakowski, 2000).

In 2001 the capacity of the system was increased by adding anadditional production and an injection well. These were drilled intothe same formations at locations near the existing wells giving a

pair of injection and a pair of production wells 1.2–1.7 km apart.The second production well, Banska PGP-1, was highly produc-tive and discharged 152 l/s of fluid at 87 ◦C under artesian drive.The second injection well, Bialy Dunajec PGP-2, required 8.4 MPa

therm

wbTiwoiae3

3

oTedo1isnt

iS221a5Ttwit(waT0

3

3

gtostsstsiat

4

tdtb

K.F. Evans et al. / Geo

ellhead pressure to inject 45 ◦C water at 56 l/s (Kêpinska, 2003c),ut this was reduced to 5.0 MPa by acid stimulation (Nagy, 2007).he 4-well system began operation at a peak flow rate of 186 l/sn 2001. The maximum wellhead pressure at both injection wells

as 6.0 MPa (Kêpinska, 2003c). Given a static wellhead pressuref 2.4 MPa, negligible friction losses in the 9–5/8 in. casing, and anncrease in column weight due to cooling of about 0.2 MPa (Dlugosznd Nagy, 1995), the corresponding maximum downhole pressurexcess above formation pressure during injection is approximately.8 MPa.

.8.2. Uniejów (Polish Lowlands):The town lies in the Polish Lowland Province, some 50 km WNW

f Lodz. It is located on the southern margin of the Mid-Polishrough, which is a major continental suture (Dadlez, 2003; Dadlezt al., 1995). The reservoir is a Lower Cretaceous sandstone at aepth of 1.9–2.0 km (Kêpinska, 2010). The basement lies at depthsf 5–6 km and is overlain by Zechstein salt deposits (Dadlez et al.,995). The nearest stress estimate registered in the WSM database

s from a site some 150 km away. Regional data are reasonably con-istent in indicating a pattern of NS compression. The area has lowatural seismic hazard. The nearest events are located some 110 kmo the east and have magnitudes in the range Mw 4.2–4.6.

The injection (AGH-1) and production (AGH-2) wells of thenitial doublet were drilled 1 km apart in 1990–1991 (Sapínska-líwa, 2003). The reservoir temperature was 70 ◦C (Kêpinska et al.,000), and the static wellhead pressure was 0.4 MPa (Kêpinska,003b). Production flow rate from AGH-2 under artesian drive was8.8 l/s at 68 ◦C but could be increased to 33.4 l/s with the aid of

submersible pump (Kêpinska, 2010). Total dissolved solids were–8 g/l. The doublet was commissioned for district heating in 2001.he operational flow rate in 2004 was 18.8 l/s at 68 ◦C, with injec-ion at 42 ◦C giving 2.1 MWt (Kêpinska, 2005). Injection pressureas 0.7 MPa (Sapínska-Slíwa, 2003). The system was expanded

n 2005 with the addition of a second injection well, IGH-1, andhe installation of a submersible pump in the production wellSapínska-Slíwa and Gonet, 2010). The peak flow rate in winteras 28 l/s at a temperature of 69 ◦C with injection temperature of

t most 45 ◦C yielding 2.75 MWt (Sapínska-Slíwa and Gonet, 2010).he injection pressures of the 3-well system were approximately.6–0.7 MPa (Sapínska-Slíwa et al., 2010).

.9. Switzerland

.9.1. Riehen, SwitzerlandThis hydrothermal doublet is located at the southern end of the

raben, some 5 km east of the Basel EGS site. The locality is charac-erised by moderate levels of natural seismicity (see Section 2.6.1n Basel EGS). The system consists of two boreholes drilled to inter-ect the Muschelkalk carbonate at a depth of 1.25–1.55 km wherehey are 1 km apart (Mégel and Rybach, 2000). This geologic unit iseparated from the basement by thick anhydrite beds and possiblyalt that might serve to hydraulically isolate the pore pressure dis-urbance in the reservoir from the basement (Hauber, 1991). Theystem has been in balanced operation at a flow rate of 18 l/s andnjection wellhead pressure of less than 1.5 MPa since 1989. Therere no reports of felt seismicity associated with the operation ofhis dual use (electricity/district heating) plant.

. Discussion

The primary objective of this study is to document case his-

ories of fluid injection with a view to identifying any systematicependence of the seismic response on reservoir injection depth orhe various parameters that constitute ‘geological setting’. It muste recognised from the outset that the available data are not ideal

ics 41 (2012) 30– 54 47

in this regard. Injections that involve a net fluid volume increasewithin the reservoir such as in hydraulic stimulation operationswould, in principle, produce a greater disturbance of pressure in thereservoir and its surroundings than comparable injections that arebalanced by production from the same reservoir, as is the case withmost operating geothermal plants. These cases are distinguishedin Table 1. Even allowing for this, the data are too heterogeneousand too few in number to allow firm conclusions to be drawn onthe basis of single-parameter correlation with seismic response.Multi-parameter correlations such as examining simultaneouslythe dependence of seismic response on depth and volume injectedare probably more appropriate but require more data than the 41sites included here. A further limitation arises from the absence oflocal seismic networks at most sedimentary injection sites. In thesecases, all that is known about the seismic response is whether or nota felt event was generated. In contrast, most igneous injection sitesincluded in the study are EGS sites where seismic networks wereoperational, and thus exact measures of the maximum magnitudeof the induced earthquakes are available.

The index of natural seismic activity adopted in this paper isthe local peak ground acceleration value that has a 10% chanceof being locally exceeded in 50 years, and is denoted as PGA inTable 1 (Giardini et al., 1999). The map of PGA values for West-ern Europe is shown with site locations in Fig. 1. A comparison ofthe maximum magnitudes triggered by fluid injection and the localPGA value is shown in Fig. 5 for igneous and sedimentary sites. Theboundary between low and moderate hazard is taken by Giardiniet al. (1999) as 0.08 g. Although this is somewhat arbitrary, it willbe seen to have some utility. The threshold magnitude for an earth-quake to be felt is taken as ML 2.0, which is the magnitude found forevents in the 5 km reservoir at Soultz to be felt by the local popu-lation (N. Cuenot, pers. comm., May 2010). This corresponds to theboundary between ‘microearthquake’ and ‘earthquake’ proposedby Bohnhoff et al. (2010), although this is also somewhat arbitrary.Thus, the maximum possible magnitude of induced events at siteswithout a local seismic network, and where seismicity was not feltor recorded on regional networks, was set at 2.0. These sites areindicated in Fig. 1 by the ‘error bar’ extending down from ML 2.0 toindicate the range of possible values. The multiplicative factor indi-cates the number of sites that are superposed. Sites where an exactvalue is available for the maximum magnitude of induced eventsare shown by filled circles.

It is evident that there is no simple correlation between the PGAindex of natural seismicity at a site and the maximum ML induced inresponse to injection, either for igneous or sedimentary rocks. Forigneous rocks, where the trends are clearer owing to the availabilityof exact measures of max-ML, all felt events occurred at sites wherethe PGA value was 0.08 g or greater. Most igneous cases denotestimulation injections. Two Icelandic sites at Laugaland and Svart-sengi have relatively high PGA values, but showed very low seismicresponse to injection. This is probably because both involved injec-tion into reservoirs whose pressure had declined by a few MPaowing to earlier production (Brandsdóttir et al., 2002). For injec-tion into sedimentary rocks, only four out of 25 cases produced feltevents, and three of these have PGA values substantially greaterthan 0.07 g. There are three sites where local PGA values exceed0.07 g but no felt events were initiated by the injections, althoughone of these, Cesano, is marginal in as much as an event of ML 2.0was recorded. The other two cases are Riehen in Switzerland, wherethere are evaporites below the reservoir that might inhibit pressurediffusion into the basement, and Bialy-Dunajek in Poland.

The only exception to the rule, based on the current admittedly

limited data, that sites with felt events have PGA values higher than0.07 g is the Unterhaching site where an event of ML 2.4 occurredduring operation despite the area having an expected PGA of 0.05 g.At Unterhaching, injection takes place into a high-angle fault in

48 K.F. Evans et al. / Geotherm

4

3

2

1

0

-1

Max

loca

l mag

nitu

de, M

L

50403020100PGA value (%g)

Igneous rocks

BaselSoultz-5km

Lowhazard

Moderate hazard

Sites with max-ML estimation Sites without max-ML estimation

but no reports of felt events

Landau Hellisheidi

Mt. Amiata

14 data points

SvartsengiLaugaland

Krafla

4

3

2

1

0

-1

Max

loca

l mag

nitu

de, M

L

50403020100PGA value (%g)

Sedimentary rocks

Approximate threshold to be felt

Lowhazard

Moderatehazard

Unterhaching

Latera

Sites with max-ML estimation Sites without max-ML estimation

but no reports of felt events

Torre AlfinaLarderello/Travale

Cesano

x6x5

25 data points

x2x2

(a)

(b)

Fig. 5. Maximum local magnitude of induced earthquakes as a function of the PGAvalue at each of the injection sites. The PGA value is the estimated seismic hazardand is used as a measure of natural seismic activity. The largest magnitude eventsmay occur during either stimulation or circulation operations, as listed in Table 1.The vertical bars denote the range of uncertainty of maximum magnitude at siteswhere no local seismic network was operational and no event was felt (N-Rep inTable 1). Bars marked ‘×4’ indicate that four datapoints are superposed. We assumethat events of ML greater than 2.0 are felt and reported. a) Results for injections intoigneous rocks. All except Laugaland, Svartsengi, Krafla and Monte Amiata are ‘stim-ufi

l4beth

sfpsciompbPsiw

it

lation’ injections involving a net fluid volume increase in the reservoir. b) Resultsor injections into sedimentary rocks. All except Gross Schönebeck and Horstbergnvolve essentially balanced circulation.

imestones just above the basement, and the production well is.5 km distant from the injector, both factors that might reasonablye expected to increase the risk of inducing seismic events. Thevent was small, but nevertheless serves to emphasise the pointhat felt events cannot, a priori, be ruled out in regions with noistoric record of felt seismicity.

The present limited data tentatively suggest that low naturaleismicity levels given by PGA values below 0.08 g may be a use-ul indicator of a low propensity for fluid injection operations toroduce damaging events. However, higher values do not neces-arily imply a high propensity. Other factors can be decisive. Thisonclusion is essentially in accord with the findings of three EGS-nduced seismicity workshops which recommended that the recordf historical seismicity, the data underpinning the local PGA esti-ates, be included in assessments of the potential of planned EGS

rojects to induce felt events (Majer et al., 2007, 2008). It is possi-le that other indices of local natural seismicity may be better thanGA for anticipating the seismic response to injection, such as localeismic moment release or b-values. An evaluation of these indicess beyond the scope of this study, although we note that neither

ould have predicted the event at Unterhaching.The data are too limited to address the question of whether

njection into sedimentary rock tends to be less seismogenichan comparable injection into crystalline rocks in the same

ics 41 (2012) 30– 54

seismo-tectonic and geologic setting. As noted earlier, most datafrom igneous rocks are from stimulation injections that involvea fluid volume increase in the reservoir (Fig. 5a), whereas thevast majority of data from sedimentary rocks are from balancedcirculation at operating geothermal plants (Fig. 5b). Exceptions arethe stimulations of sedimentary reservoirs at Gross Schönebeckand Horstberg, and the long-term CO2 injections at Sleipnerand Ketzin, none of which have produced felt seismicity. Theissue of induced seismicity due to long-term injection (ratherthan circulation) is particularly important for CO2 sequestration.Hydraulic fracturing operations of oil and gas reservoirs are com-monplace but are not usually associated with felt seismicity. Thevolumes injected are small in comparison to those used for EGSstimulations, although the net volumes injected per well for gasshale stimulations are comparable (Cipolla, 2009). Recent gas shalefracturing operations near Blackpool, UK may have been associatedwith ML 2.3 and 1.5 events that reportedly occurred close to theinjection well in April and May 2011 respectively (BGS, 2011).There is no doubt that felt earthquakes can occur in sediments. Inthe present study, Unterhaching and possibly Landau if the failureoccurred in sediments rather than basement, represent examples.Waterflood operations for secondary oil recovery have long beenassociated with felt induced seismicity, the classic example beingRangely (Raleigh et al., 1976). However, there have also beenseveral natural earthquakes of sizeable magnitude in recent years,such as the 1996 ML 5.3 Annecy (France) event and the Fribourg(Switzerland) earthquake sequences with a maximum magnitudeML 4.3, where the failure occurred entirely in sediments at 2–3 kmdepth (Kastrup et al., 2007; Thouvenot et al., 1998). Notewor-thy examples of even shallower seismicity are the earthquakesequences in the Tricastin area of southern France that occurred atdepths of only a few hundred metres (Thouvenot et al., 2009), andthe well-documented cases of rain-induced seismicity in Germanyand Switzerland (Husen et al., 2007; Kraft et al., 2006).

The risk of inducing seismic events is likely to increase wheninjection takes place near to or within fault zones (Davis andFrohlich, 1993), as is often the case in geothermal projects. This isnot only because earthquakes represent the local failure of faults,but also because faults are often highly transmissive and serve tochannel flow. As such, they promote deeper penetration of pressureperturbations, both laterally and vertically, thereby increasing thelikelihood of the perturbation reaching a region where conditionsare close to those required for extensive failure. Large separationbetween injection and production wells will also tend to increasethe spatial extent of pore pressure perturbation within the rockmass. However, it is clear from the data that injection into faultzones does not necessarily produce felt earthquakes. At least 8 ofthe 24 sedimentary sites inject into or close to faults, but only one ofthese (Unterhaching) is associated with felt seismicity. The injec-tion at Landau, which is also associated with induced seismicity,also takes place into or near faults in both the basement and over-lying sedimentary units. The largest events in the Soultz reservoiralso appear to be associated with a fault that intersected the injec-tion well (Dorbath et al., 2009). Interestingly, all wells at Soultzwere intersected by numerous fracture zones that for the most partwere critically stressed (Evans, 2005). These were seismically activeduring the injections, but for the most part the magnitudes of theevents were too small to be felt. The faults are distinguished inSoultz as larger structures that have accommodated greater offset.Whether the faults become activated depends upon whether theyare critically stressed, although the maximum magnitude earth-quake that will result depends upon other factors, such as the scale

of the fault, and stress and strength heterogeneity.

Simple physical considerations might be taken to suggest thathigh injection pressures will increase the risk of producing feltevents. However, felt events will arise only if other factors, such as

K.F. Evans et al. / Geotherm

4

3

2

1

0

-1

Max

loca

l mag

nitu

de, M

L

50403020100Injection pressure minus formation pressure (MPa)

Igneous rocksBaselSoultz-5kmLandau

14 data points

KTB-9km

Bad Urach

Sites with max-ML estimation Sites without max-ML estimation

but no reports of felt events

KTB-6km

Hellisheidi

LaugalandSvartsengi

Rosemanowes

Krafla

Fjallbäcka

Soultz-3.5km

4

3

2

1

0

-1

Max

loca

l mag

nitu

de, M

L

6050403020100Injection pressure minus formation pressure (MPa)

Sedimentary rocks

Approximate threshold to be feltUnterhaching

Sites with max-ML estimation Sites without max-ML estimation

but no reports of felt events

Torre Alfina

Cesano

20 data points

Gross SchönebeckHorstberg

a

b

Fig. 6. Maximum local magnitude of induced earthquakes as a function of the valuesof injection pressure excess over formation pressure listed in Table 1. When avail-able, the downhole injection pressure is used; otherwise wellhead pressure is used.a) Results for injections into igneous rocks. All injections at pressures higher than8 MPa except Rosemanowes correspond to high-rate stimulation injections whichinvolve a fluid volume increase in the reservoir, whereas those below 8 MPa, exceptfor Hellisheidi, are long-term balanced circulations. Note that the data points forLandau and Rosemanowes correspond to circulation parameters at or immediatelybefore the largest magnitude earthquakes occurred, and that stimulation injectionsat substantially higher pressures were performed at both sites without generatingfelt events. b) Results for injections into sedimentary rocks. All data are for balancedcirculations, except for Horstberg and Gross Schönebeck which were both large-vre

ttdmtthsrbdmtginmf

olume stimulation injections. Collectively the results show that there is no simpleelation between injection pressure and the maximum magnitude of the inducedvents.

he prevailing stress state and presence of suitably oriented frac-ure zones or faults, are favourable (Davis and Frohlich, 1993). Thisependence on other factors is evident in Fig. 6, which shows theaximum magnitude of earthquakes induced at the sites as a func-

ion of the injection pressure values listed in Table 1. Fig. 6a presentshe data for injection into igneous rocks. All injections at pressuresigher than 8 MPa, except Rosemanowes, correspond to high-ratetimulation injections which involve a fluid volume increase in theeservoir. Those below 8 MPa, except for Hellisheidi, are long-termalanced circulations. The datapoints for Landau and Rosemanowesenote circulation parameters at or immediately before the largestagnitude earthquakes occurred. At both sites, stimulation injec-

ions at substantially higher pressures were performed withoutenerating felt events (up to 13 MPa for Landau and 14 MPa for

njections at Rosemanowes). The collective data show that there iso simple relation between injection pressure and the maximumagnitude of the induced events. The same conclusion also holds

or the cases of injection into sedimentary rocks shown in Fig. 6b.

ics 41 (2012) 30– 54 49

All data in Fig. 6b are for balanced circulations, except for Horstbergand Gross Schönebeck which both involved large-volume stimula-tion injections.

Significant seismicity, albeit usually at non-damaging magni-tudes, invariably accompanies injection into crystalline rocks. Allof the igneous rock masses (and at least some of the sedimen-tary reservoirs such as Gross Schönebeck) are found to be criticallystressed in the sense that optimally oriented fractures and faultswhose strength is governed by a Coulomb friction criterion witha coefficient of 0.65 would be at or close to failure. The exis-tence of large faults that are critically stressed near a prospectivereservoir has been used as an indicator of the potential of fluid injec-tion to produce damaging earthquakes (Hunt and Morelli, 2006).Whilst this approach may be practical for assessing the poten-tial for inducing large events on major faults that have a mappedsurface expression, it is difficult to apply it to seismic events ofsmaller size, say in the ML 2.5–4.5 range (circular source diametersof 250–2700 m for 1 MPa stress drop), since these may occur onstructures that are wholly buried or too small to have been mapped.For example, the Soultz reservoirs contain critically stressed frac-ture zones or faults that are seismically activated by fluid injection,but the scale of the larger structures, and hence the maximum eventsize that can be generated, is unknown.

It should also be noted that the stress levels supported by geo-logic structures in many of the geothermal reservoirs examinedhere imply equivalent frictional strengths that are significantlyhigher than the threshold of 0.65 used to define criticality in thispaper (e.g. ∼0.95 at Soultz (Evans, 2005)). Thus, criticality in thesense used here is a necessary but not a sufficient condition forfailure. The uncertainty in fault strength, and often also the pre-vailing stresses which are invariably difficult to measure and, likestrength, are subject to heterogeneity, greatly limits the degree towhich it is possible to estimate the ‘proximity to extensive failure’from measurable geomechanical parameters.

Seven of the sedimentary reservoirs described here are under-lain, albeit usually not directly, by extensive evaporite deposits. Thepresence of these units can mechanically decouple and hydrauli-cally isolate the overlying strata from the basement. The carbonatereservoir at Riehen, which is only 5 km from the Basel EGS site,is underlained by anhydrite and possibly salt deposits that mightserve to hydraulically isolate it from basement. This may be a con-tributing factor to the different seismogenic responses, althoughthe different nature of the injections (i.e. stimulation as opposed tobalanced circulation) is possibly the primary reason.

The data presented here are not entirely consistent with theview that deeper injection in crystalline rocks tends to producelarger magnitude events. Events approaching or exceeding ML 3.0were generated through injection at 5.0 km in two EGS crystallinereservoirs (Basel and Soultz – 5 km), but only minor seismicityresulted from injection at 3.3–4.4 km depth in Bad Urach and at6 and 9 km depth in the German KTB site, despite high injectionpressures. Neither of these cases involve the injection of large vol-umes, and it is uncertain whether larger events would have beenregistered had injection continued for longer periods. There are toofew data to draw any firm conclusions in this regard.

Acknowledgements

The work was supported by the CARMA (http://www.carma.ethz.ch) and GEOTHERM (http://www.geotherm.ethz.ch)projects funded by the Competence Centre for Environment and

Sustainability and by the Competence Centre for Energy andMobility of the Swiss Federal Institutes of Technology (ETH), andthe Swiss Federal Office of Energy. We are grateful to the followingpeople for providing unpublished data or information: Kristján

5 therm

ÁABShETGUPMiKG(d(STIoGGDHhRnJaZfL

R

A

A

A

A

A

A

B

B

B

B

B

B

B

0 K.F. Evans et al. / Geo

gústsson (ISOR, Iceland), Gudni Axelsson (ISOR, Iceland), Juliangerer (Stadtwerke München GmbH, Munich, Germany), Stefanaisch (Q-Con, Bad Bergzabern, Germany), Franc ois Cornet (IPGS,trasbourg, France), Nicolas Cuenot (EEIG Heat Mining, Kutzen-ausen, France), Ola Eiken (Statoil, Trondheim, Norway), Lorenzichinger (Hydroisotop GmbH, Schweitenkirchen, Germany),homas Fritzer (Bayerisches Landesamt für Umwelt, Munich,ermany), Wolfgang Geisinger (Geothermie Unterhaching GmbH,nterhaching, Germany), Ásgrímur Gudmundsson (Landsvirkjunower, Reykjavik), Andy Jupe (Altcom Ltd., Penzance, UK),ichael Kaelcke (Innovative Energie für Pullach GmbH, Pullach

. Isartal, Germany), Beata Kêpinska (Polish Academy of Science,raków, Poland), Erwin Knapek (former Mayor of Unterhaching,ermany), Stefan Lüth (GFZ-Potsdam, Germany), Udo Reimer

Stadtwerke-Waren, Waren, Germany), Uwe Richlak (Neubran-enburger Stadtwerke, Neubrandenburg, Germany), Ulrich SchanzGeoenergy-Bayern GmbH, Regensdorf, Bavaria, Germany), Jochenchneider (Hydroisotop GmbH, Schweitenkirchen, Germany),ore Torp (Statoil, Trondheim, Norway), Pierre Ungemach (GPCnstrumentation Process, Paris), Kristín Vogfjörd (Icelandic Mete-rological Office, Reykjavík, Iceland), Thomas Wallroth (Bergab,othenburg, Sweden), Joachim Wassermann (Munich University,ermany) and Michael Weber (GFZ-Potsdam, Germany). We thankavid Bruhn (GFZ-Potsdam, Germany), Agust Gudmundsson (Royalolloway College, London), Reinhard Jung (Jung-Geotherm, Isern-agen, Germany), Grzegorz Kwiatek (GFZ-Potsdam, Germany),üdiger Schulz (Leibnitz Institute for Applied Geophysics, Han-over, Germany), Ingrid Stober (Freiburg University, Germany) and

oachem Poppei (AF-Colenco, Baden, Switzerland) for discussionnd assistance in locating information, and Bjørn Oddsson (ETH-ürich, Switzerland) for Icelandic translations. The comments ofour reviewers and the editors, Drs. Sabodh Garg and Marceloippmann, helped improve the manuscript.

eferences

gustsson, K., Halldorsson, P., 2005. Seismic hazard in the Hengill area based on theSIL earthquake catalogue (First results). Report no. 05015. Vedurstofa, Reykjavik,Iceland, 39 pp.

horner, L., Murawski, H., Schneider, G., 1970a. Die Verbreitung von schaden-vurursachenden Erdbeben auf dem Gebiet der Bundesrepublik Deutschland.Zeitschrift für Geophysik 36, 313–343.

horner, L., Murawski, H., Schneider, G., 1970b. The distribution of earthquakes caus-ing damage on the territory of the Federal Republic of Germany. Zeitschrift fürGeophysik 36, 313–343 (in German).

rnason, A., Eysteinsson, H., Hersir, G.P., 2010. Joint 1D inversion of TEM and MTdata and 3D inversion of MT data in the Hengill area, SW Iceland. Geothermics39, 13–34, doi:10.1016/j.geothermics.2010.01.002.

xelsson, G., Flovenz, O.G., Hjartarson, A., Hauksdottir, S., Sverrisdottir, G., Arnason,F., Arnason, A., Bodvarson, R., 2000. Thermal energy extraction by reinjectionfrom Laugaland geothermal system in North Iceland. In: Proceedings: WorldGeothermal Congress, Kyushu-Tohoku, Japan, May 28–June 10, pp. 3027–3032.

xelsson, G., Thórhallsson, S., Björnsson, G., 2006. Stimulation of geothermal wellsin basaltic rock in Iceland. In: Proceedings ENGINE Workshop 3: Stimulation ofReservoir and Microseismicity, Ittingen, Switzerland, 29 June–1 July, pp. 17–25.

aisch, S., Harjes, H.P., 2003. A model for fluid-injection-induced seismicity at theKTB, Germany. Geophysical Journal International 152, 160–170.

aisch, S., Weidler, R., Voeroes, R., Tenzer, H., Teza, D., 2004. Improving hydraulicstimulation efficiency by means of real-time monitoring. In: Proceedings 29thWorkshop on Geothermal Reservoir Engineering, Stanford University, Stanford,CA, USA, January 26–28, p. 7.

aklid, A., Korbøl, R., 1996. Sleipner Vest CO2 disposal, CO2 injection into a shallowunderground aquifer. In: Proceedings of SPE Annual Technical Conference andExhibition. SPE Paper Number 36600, Denver, Colorado, October 2–9, p. 9.

arelli, A., Ceccarelli, A., Dini, I., Fiordelisi, A., Giorgi, N., Lovari, F., Romagnoli, P.,2010. A review of the Mt. Amiata geothermal system (Italy). In: Proceedings ofWorld Geothermal Congress 2010, Bali, Indonesia, April 25-29, p. 6.

aria, R., Baumgaerner, J., Gérard, A., Jung, R., 1997. European hot dry rock geother-mal research programme 1996-1997. Final Report to the European Commission,Brussels, Belgium, 192 pp.

aria, R., Green, A.S.P., 1990. Microseismics: a key to understanding reservoirgrowth. In: Proceedings of Camborne School of Mines International Hot Dry RockConference, Camborne School of Mines, Redruth, U.K., June 27–30, pp. 363–377.

aria, R., Hearn, K., Batchelor, A.S., 1985. Induced seismicity during the hydraulicstimulation of a potential Hot Dry Rock geothermal reservoir. In: Proceedings

ics 41 (2012) 30– 54

of 4th Conf. on Acoustic Emision/Microseismic Activity in Geologic Structuresand Materials, Pennsylvania State University, State Park, Pennsylvania, USA, pp.327–352.

Baria, R., Michelet, S., Baumgärtner, J., Dyer, B.C., Gerard, A., Nicholls, J., Hettkamp,T., Teza, D., Soma, N., Asanuma, H., Garnish, J., Megel, T., 2004. Microseismicmonitoring of the world’s largest potential HDR reservoir. In: Proceedings 29thWorkshop on Geothermal Reservoir Engineering, Stanford University, Stanford,CA, USA, January 26-28, pp. 194–201.

Batchelor, A.S., Baria, R., Hearn, K., 1983. Monitoring the effects of hydraulic stimu-lation by microseismic event location: a case study. In: Presented at 58th Ann.Tech. Conf. and Exhibition of SPE, San Francisco, California, SPE Paper Number12109, p. 12.

Batini, F., Bufe, C., Cameli, G.M., Console, R., Fiordelisi, A., 1980a. Seismic moni-toring in Italian geothermal areas I: seismic activity in the Larderello-Travaleregion. In: Proceedings Second DOE-ENEL Workshop on Cooperative Research inGeothermal Energy, Report LBL-11555, Lawrence Berkeley Laboratory, Berkeley,CA, USA, October 20–22, pp. 20–47.

Batini, F., Cameli, G.M., Carabelli, E., Fiordelisi, A., 1980b. Seismic monitoring inItalian geothermal areas II: seismic activity in the geothermal fields duringexploitation. In: Proceedings of Second DOE-ENEL Workshop on CooperativeResearch in Geothermal Energy, Report LBL-11555, Lawrence Berkeley Labora-tory, Berkeley, CA, USA, October 20–22, pp. 48–85.

Batini, F., Console, R., Luongo, G., 1985. Seismological study of Larderello-Travalegeothermal area. Geothermics 14 (2–3), 255–272.

Batini, F., Fiordelisi, A., Moia, F., 1990. Main features of the seismicity in the MonteAmiata and Latera geothermal areas (Italy). In: Presented at XXII General Assem-bly European Seismological Commission, Barcelona, Spain, September 17–22,pp. 649–654.

Baumgärtner, J., Teza, D., Hettkamp, T., Gandy, T., Schindler, M., 2010a. Geothermalreservoir development in the Upper Rhine Graben – “Concepts, Techniques andExperiences”, the geothermal projects in Landau and Insheim. In: Presented atSecond European Geothermal Review, Bestec GmbH, Landau, Germany, Mainz,Germany, June 21–23, p. 31.

Baumgärtner, J., Teza, D., Hettkamp, T., Hauffe, P., 2010b. Stimulierung tiefer geother-mischer Systeme, (Stimulation of deep geothermal systems). BBR (Fachmagazinfür Brunnen- und Leitungsbau). Sonderheft Tiefe Geothermie 14–23, 84.

Baumgärtner, J., Teza, D., Hettkamp, T., Homeier, G., 2006. Geothermal explorationin the Upper Rhine valley in Germany. In: Proceedings of the ENGINE LaunchingConference, Orléans, France, February 12–15, p. 24.

Bayerische-Landtag, 2009. Schriftliche Antwort auf Anfrage des AbgeordnetenLudwig Wörner SPD vom 05.03.2009 über Geothermiebohrungen (Record ofproceedings of the 16th Electoral-Period (2008-2013)). Report no. 16/1233.Bayerische Landtag (Parliament of Bavarian States), Munich, Germany, 2 pp.Available at http://www.bayern.landtag.de/cps/rde.

Bertani, R., 2005. World geothermal power generation in the period 2001-2005.Geothermics 34, 651–690, doi:10.1016/j.geothermics.2005.09.005.

Bertini, G., Capetti, G., Fiordelisi, A., 2005. Characteristics of geothermal fields inItaly. Giornale di Geologia Applicata 1, 247–254, doi:10.1474/GGA.2005-01.0-24.0024.

BGS, 2011. British Geological Survey web report of 27th May 2011 event,(http://www.bgs.ac.uk/research/earthquakes/BlackpoolMay2011.html).

Billi, B., Capetti, G., Luccioli, F., 1986. ENEL activity in the research, exploration andexploitation of geothermal energy in Italy. Geothermics 15 (5–6), 765–779.

Bjornsson, G., 2004. Reservoir conditions at 3-6 km depth in the Hellisheidi Geother-mal Field, SW Iceland, estimated by deep drilling, coldwater injection, andseismic monitoring. In: Proceedings 29th Workshop on Geothermal ReservoirEngineering, Stanford University, Stanford, CA, USA, January 26–28, pp. 67–74.

Bohnhoff, M., Dresen, G., Ellsworth, W.L., Ito, H., 2010. Passive seismic monitoringof natural and induced earthquakes: case studies, future directions and socio-economic relevance. In: Cloetingh, S., Negendank, J. (Eds.), New Frontiers inIntegrated Solid Earth Sciences: International Year of Planet Earth. Springer,Dordrecht, Netherlands, pp. 261–285.

Bönnemann, C., Schmidt, B., Ritter, J., Gestermann, N., Plenefisch, T., Wegler,U., 2010. Das seismische Ereignis bei Landau vom 15 August 2009 (Theseismic event near Landau of 15th August 2010). Final report by the expertpanel on seismic risk associated with hydrothermal geothermal plants,Landesamt für Geologie und Bergbau of Rheinland-Pfalz (Regional authorityfor geology and mines of Rheinland-Pfalz), 54 pp. Available at http://www.lgb-rlp.de/einzelansicht.html?&no cache=1&tx ttnews%5BbackPid%5D=1542&txttnews%5Btt news%5D=64.

Brandsdóttir, B., Franzson, H., Einarsson, P., Árnason, K., Kristmannsdóttir, H., 2002.Seismic monitoring during an injection experiment in the Svartsengi geothermalfield, Iceland. Jökull 51, 43–52.

Brogi, F., Fabbrini, L., 2009. Extensional and strike-slip tectonics across the MonteAmiata–Monte Cetona transect (Northern Apennines, Italy) and seismotectonicimplications. Tectonophysics 476, 195–209, doi:10.1016/j.tecto.2009.02.020.

Brudy, M., Zoback, M.D., Fuchs, K., Rummel, F., Baümgartner, J., 1997. Estimation ofthe complete stress tensor to 8 km depth in the KTB scientific drill holes: implica-tions for crustal strength. Journal Geophysical Research 102 (B8), 18453–18475.

Bujakowski, W., 2000. The pilot project of a geothermal heat recuperation cascadesystem for fish and vegetable breeding, the Pas-Meeri Geothermal Laboratory

(Podhale region, Poland). In: Proceedings World Geothermal Congress, Kyushu-Tohoku, Japan, May 28–June 10, pp. 3379–3383.

Cameli, G.M, Carabelli, E., Fiordelisi, A., Graziano, F., 1983. Survey of local seismicactivity during production and reinjection of geothermal fluids in Cesano andLatera areas. In: Proceedings of European Geothermal Update, 3rd International

therm

C

C

C

C

C

C

C

C

C

C

D

D

D

D

D

D

D

D

D

D

E

E

E

E

E

F

G

K.F. Evans et al. / Geo

Seminar on the Results of EC Geothermal Energy research, European Commu-nity, Munich, Germany, November 29–December 1, pp. 232–240.

ammerer, F., Hann, H.P., Martin, M., Obermaier, M., Pechnig, R., Reinold, M., Rupf,I., Schellschmidt, R., Simon, T., Stober, I., Wolpert, P., Zedler, H., 2009. Mach-barkeitstudie für ein Hot Dry Rock Wärme- und Stromnutzungskonzept BadUrach. BMU Report No. FKZ: 0327615A. Stadt Bad Urach, Germany, 118 pp.

appetti, G., Romagnoli, P., Sabatelli, F., 2010. Geothermal power generation in Italy:2005-2009 update report. In: Proceedings of World Geothermal Congress 2010,Bali, Indonesia, April 25–29, p. 8.

arabelli, E., Moia, F., Fiordelisi, A., 1984. Seismic monitoring during geothermalwells stimulation as contribution to the individuation of prevailing fracturationtrends. In: Presented at Seminar on Utilization of Geothermal Energy for ElectricPower Production and Space Heating, Florence, Italy, May 14–17, pp. 1–31.

harléty, J., Cuenot, N., Dorbath, L., Dorbath, C., Haessler, H., Frogneux, M., 2007. Largeearthquakes during hydraulic stimulations at the geothermal site of Soultz-sous-Forêts. International Journal Rock Mechanics Mining Sciences 44, 1091–1105.

ipolla, C.L., 2009. Modelling production and evaluating fracture performance inunconventional gas reservoirs. Journal Petroleum Technology 61 (9), 84–90.

ornet, F.H., 1989. Experimental investigations of forced fluid flow through a graniterock mass. In: Proceedings of 4th Int. Seminar on the Results of EC GeothermalEnergy Demonstration, Florence, Italy, April 27–30, pp. 189–204.

ornet, F.H., Bérard, F.H., Bourouis, S., 2007. How close to failure is a granite rockmass at a 5 km depth? International Journal Rock Mechanics Mining Sciences44, 47–66.

ornet, F.H., Burlet, D., 1992. Stress field determination in France by hydraulic testsin boreholes. Journal Geophysical Research 97 (B8), 11829–11849.

SM-Report, 1989. Camborne Geothermal Energy Project 2C-5 Final Report Phase2C. Report ETSU G 137F, Camborne School of Mines, Penryn, Cornwall, U.K., 42pp.

uenot, N., Dorbath, C., Frogneux, M., Langet, N., 2010. Microseismic activity inducedunder circulation conditions at the EGS Project of Soultz-sous-Forêts (France).In: Proceedings World Geothermal Conference, Bali, Indonesia, 25–29 April, p.9.

adlez, R., 2003. Mesozoic thickness pattern in the Mid-Polish Trough. GeologicalQuarterly 47 (3), 223–240.

adlez, R., Narkiewicz, M., Stephenson, R.A., Visser, M.T.M., van Wees, J.-D., 1995.Tectonic evolution of the Mid-Polish Trough: modelling implications and signif-icance for central European geology. Tectonophysics 252, 179–195.

avis, S.D., Frohlich, C., 1993. Did (or will) fluid injection cause earthquakes? –criteria for rational assessment. Seismological Research Letters 64, 207–224.

eichmann, N., Ernst, J., 2009. Earthquake focal mechanism solutions of the inducedseismicity in 2006 and 2007 below Basel (Switzerland). Swiss Journal Geo-sciences 102, 457–466, doi:10.1007/s00015-009-1336-y.

eichmann, N., Giardini, D., 2009. Earthquakes induced by the stimulation ofan Enhanced Geothermal System below Basel (Switzerland). SeismologicalResearch Letters 80 (5), 784–798, doi:10.1789/gssrl.80.5.784.

ietrich, H.G., 1982. Geological results from the Urach-3 borehole and the corre-lation with other boreholes. In: Hänel, R. (Ed.), The Urach Geothermal Project.Schweizerbart’sche Verlagsbuchhandlung, Stuttgart, Germany, pp. 49–58.

lugosz, P., Nagy, S., 1995. Determination of hydrodynamic parameters of the Pod-hale geothermal reservoir. Bulletin Polish Academy Sciences-Earth Science 43(4), 225–242.

oebl, F., 1968. Die geothermischen Verhältnisse des Ölfeldes Landau/Pfalz. In: Pro-ceedings of Graben Problems: International Rift Symposium, October10–12, E.Schweizerbart’sche Verlagsbuchhandlung, Karlsruhe, Germany, pp. 192–207.

ong E&P, 2004. Geothermal well MAH-2, 5512/11-2 (Margretheholm-2). ReportNo. 17060. Geological Survey of Denmark and Greenland, Copenhagen,Denmark, 18 pp.

orbath, L., Cuenot, N., Genter, A., Frogneux, M., 2009. Seismic response of thefractured and faulted granite of Soultz-sous-Forêts (France) to 5 km deepmassive water injections. Geophysics Journal International 117, 653–675,doi:10.1111/j.1365-246X.2009.04030.x.

iken, O., Ringrose, P., Hermanrud, C., Nazarian, B., Torp, T.A., Høier, L., 2011. Lessonslearned from 14 years of CCS operations: Sleipner. In Salah and Snøhvit. EnergyProcedia 4, 5541–5548, doi:10.1016/j.egypro.2011.02.541.

liasson, T., Sundquist, U., Wallroth, T., 1988. Stimulation experiments with waterand viscous fluid at the HDR research site in the Bohus granite, SW Sweden.Report no. Fj-6. Chalmers University of Technology, Gothenburg, Sweden, 69pp.

liasson, T., Sundquist, U., Wallroth, T., 1990. Circulation experiments in 1989 atFyällbacka HDR test site. Report no. Fj-10. Chalmers University of Technology,Gothenburg, Sweden, 56 pp.

vans, K.F., 2005. Permeability creation and damage due to massive fluid injectionsinto granite at 3.5 km at Soultz: part 2 – critical stress and fracture strength.Journal Geophysical Research 110, 14, doi:10.1029/2004JB003169.

vans, K.F., Kohl, T., Hopkirk, R.J., Rybach, L., 1992. Modelling of energyproduction from Hot Dry Rock systems. Report. Swiss Federal Instituteof Technology/Polydynamics, Zürich, Switzerland, 316 pp. Available athttp://www.geothermal.ethz.ch/publications.htm.

oulger, G.R., Long, R.E., Einarsson, P., Bjornsson, A., 1989. Implosive earthquakesat the active accretionary plate boundary in northern Iceland. Nature 337,

640–642.

enter, A., Evans, K.F., Cuenot, N., Fritsch, D., Sanjuan, B., 2010. The Soultz geothermaladventure: 20 years of research and exploration of deep crystalline frac-tured rocks for EGS development. Comptes Rendus Geoscience 342, 502–516,doi:10.1016/j.crte.2010.01.006.

ics 41 (2012) 30– 54 51

Giardini, D., Grünthal, G., Shedlock, K., Zhang, P., 1999. The GSHAP Global seismichazard map. Annali di Geofisica 42 (6), 1225–1230.

Gibowicz, S.J., 1990. Seismicity induced by mining. Advances in Geophysics 32, 1–74.Goldbrunner, J.E., 1999. Hydrogeology of deep groundwater in Austria. Mitteilungen

der Österreichischen Geologischen Gesellschaft 92, 281–294.Goldbrunner, J.E., 2005a. Bad Blumau (Styria, Austria). The success story of com-

bined use of geothermal energy. Geo-Heat Centre (GHC) Quarterly Bulletin 26(2), 27–30.

Goldbrunner, J.E., 2005b. State, possible future developments and barriers of theexploration and exploitation of geothermal energy in Austria – Country update.In: Proceedings World Geothermal Congress, Antalya, Turkey, April 24–29, p.11.

Goldbrunner, J.E., 2005c. Zur Situation der Tiefen Geothermie in Österreich (Stateof deep geothermal resource utilisation in Austria). Geothermische Energie 48,22–25.

Goldbrunner, J.E., 2007. Thermalwassererschließungen - wissenschaftlicheBegleitung von Bohrungen einschließlich Bohrlochuntersuchungen (Researchon thermal water exploitation). BBR (Fachmagazin für Brunnen- undLeitungsbau) 11, 26–33.

Goldbrunner, J.E., 2010. Austria – country update. In: Proceedings of World Geother-mal Congress, Bali, Indonesia, April 25–29, p. 14.

Goldbrunner, J.E., Shirbaz, A., Heiss, H.P., 2007. Wasserwirtschaftliche Bewertungder Thermalwassernutzungen in Oberösterreich: Berichtszeitraum 2000–2005(Economic evaluation of thermal water usage in Austria). Report Amt derOberösterreich. Landesregierung, Wasserwirtschaft, Grund- und Trinkwasser-wirtschaft, Wasserwirtschaft Land Oberösterreich, L., Linz, Austria, 36 pp.

Grüntal, G., Musson, R.M.W., Schwarz, J., Stucchi, M., 1998. European MacroseismicScale 1998 (EMS-98). Cahiers du Centre Européen de Géodynamique et de Séis-mologie 15. Centre Européen de Géodynamique et de Séismologie, Luxembourg,99 pp.

Grünthal, G., Wahlström, R., Stromeyer, D., 2009. The unified catalogue of earth-quakes in central, northern, and northwestern Europe (CENEC) – updatedand expanded to the last millenium. Journal of Seismology 13 (4), 517–541,doi:10.1007/s10950-008-9144-9.

Gudmundsson, A., 2001. An expansion of the Krafla Power Plant from 30 to 60 MWe:geothermal considerations. Transactions Geothermal Resources Council 25,741–746.

Gupta, H.K., 1992. Reservoir-Induced Earthquakes. Elsevier, Amsterdam, TheNetherlands, 364 pp.

Guterch, B., Lewandowska-Marciniak, H., Niewiadomski, J., 2005. Earthquakesrecorded in Poland along the Pieniny Klippen Belt, Western Carpathians. ActaGeophysica Polonica 53 (1), 27–45.

Haering, M.O., Schanz, U., Ladner, F., Dyer, B.C., 2008. Characterisation of the basel-1enhanced geothermal system. Geothermics 37, 469–495.

Hansen, H., Eiken, O., Aasum, T.O., 2005. Tracing the path of carbon dioxide from agas/condensate reservoir, through an Amine plant and back into the subsurfaceaquifer – case study: the Sleipner area, Norwegian North Sea. In: Offshore Europe2005, SPE Paper Number 96742, Aberdeen, Scotland, UK 6–7 September, p. 15.

Hauber, L., 1991. Ergebnisse der Geothermmiebohrungen Riehen 1 und 2 sowieReinach 1 im Südosten des Rheingrabens (Results of the geothermal bore-holes Riehen 1, 2 and Reinach in the southeastern Rhein Graben). GeologischesJahrbuch E48, 167–184.

Heederick, J.P., 1997. Improvement of the injectivity index of argillaceous sandstone.Report no. NITG 97-260-A. Netherlands Institute of Applied Geoscience, Delft,The Netherlands, 183 pp.

Heidbach, O., Tingay, M., Barth, A., Reinecker, J., Kurfess, D., Müller, B., 2010. Globalcrustal stress pattern based on the World Stress Map database release 2008.Tectonophysics 428 (1–2), 3–15.

Heinemann, B., Troschke, B., Tenzer, H., 1992. Hydraulic investigations and stressevaluation at the HDR test site Urach 3, Germany. Transactions GeothermalResources Council 16, 425–431.

Helm, J.A., 1996. The natural seismic hazard and induced seismicity of the EuropeanHDR (Hot Dry Rock) geothermal energy project at Soultz-sous-Forêts, France.Doctoral thesis. Université Louis-Pasteur de Strasbourg, Strasbourg, France, 197pp.

Herzberger, P., Münch, W., Köbel, T., Bruchmann, U., Schlagermann, P.H.H., Wolf, L.,Rettenmaier, D., Steger, H., Zorn, R., Seibt, P., Möllmann, G.U., Sauter, M., Ghergut,J., Ptak, T., 2010. The geothermal power plant Bruchsal. In: Proceedings WorldGeothermal Congress 2010, Bali, Indonesia, April 25–29, p. 6.

Hettkamp, T., Teza, D., Baumgärtner, J., Gandy, T., Homeier, G., 2007. A multi-horizonapproach for the exploration and exploitation of a fractured geothermal reser-voir in Landau/Palatine. In: Presented at First European Geothermal Review,Mainz, Germany, October 29–31, p. 15.

Huenges, E., Trautwein, U., Legarth, B., Zimmermann, G., 2006. Fluid pressurevariation in a sedimentary geothermal reservoir in the North-German basin:case study gross Schönebeck. Pure and Applied Geophysics 163, 2141–2152,doi:10.1007/s00024-006-0122-2.

Hunt, S.P., Morelli, C.P., 2006. Cooper basin HDR seismic hazard evaluation: predic-tive modelling of local stress changes due to HFR geothermal energy operationsin South Australia. Report 2006/16. University of Adelaide, Adelaide, Australia,50 pp.

Husen, S., Bachmann, C., Giardini, D., 2007. Locally triggered seismicity in the centralSwiss Alps following the large rainfall event of August 2005. Geophysical JournalInternational 171, 1126–1134, doi:10.1111/j.1365-246X.2007.03561.x.

Illies, J.H., Greiner, G., 1978. Rhinegraben and the Alpine system. Geological SocietyAmerica Bulletin 89, 770–782.

5 therm

I

J

J

J

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

L

L

L

2 K.F. Evans et al. / Geo

SC, 2010. On-line bulletin of the International Seismological Centre. InternationalSeismological Centre, Thatcham, UK. Available at http://www.isc.ac.uk.

uhlin, C., Giese, R., Zionck-Jørgensen, K., Cosma, C., Kazemeini, H., Johojuntti, N.,Lüth, S., Norden, B., Förster, A., 2007. 3D beseline seismics at Ketzin, Germany:the CO2SINK project. Geophysics 72 (5), B121–B132, doi:10.1190/1.2754667.

ung, R., Orzol, J., Jatho, R., Kehrer, P., Tischner, T., 2005. The Gensys project: extrac-tion of geothermal heat from tight sediments. In: Proceedings 30th Workshopon Geothermal Reservoir Engineering, Stanford University, Stanford, CA, USA,January 31–February 2, pp. 306–311.

upe, A.J., Green, A.S.P., Wallroth, T., 1992. Induced microseismicity and reser-voir growth at the Fjällbacka Hot Dry Rock Project, Sweden. InternationalJournal Rock Mechanics Mining Sciences & Geomechanics Abstracts 29 (4),343–354.

abus, F., Jäntsch, E., 1995. The geothermal heating plant at Waren-Papenberg:experience and modernisation. In: Proceedings of World Geothermal Congress,Florence, Italy, May 18–31, pp. 2227–2232.

ahn, D., 2008. Hydro-fractured reservoirs: a study using double-difference locationtechniques. Doctoral Thesis. Duke University, Durham, North Carolina, USA, 175pp.

arytsas, C., Mendrinos, D., Goldbrunner, J.E., 2009. Production monitoring at theGeinberg geothermal site. Energy Sources (Part A) 31, 1537–1552.

astrup, U., Deichmann, N., Fröhlich, A., Giardini, D., 2007. Evidence for an active faultbelow the northwestern Alpine foreland of Switzerland. Geophysical JournalInternational 169, 1273–1288, doi:10.1111/j.1365-264X.2007.03413.x.

astrup, U., Zoback, M.-L., Deichmann, N., Evans, K.F., Giardini, D., Michael, A., 2004.Stress field variations in the Swiss Alps and the northern Alpine foreland derivedfrom inversion of fault plane solutions. Journal Geophysical Research 109 (B1),22, doi:10.1029/2003JB002550, B01402.

êpinska, B., 2000. The Podhale low-enthalpy system (S-Poland): geothermal char-acteristics. In: Proceedings World Geothermal Congress, Kyushu-Tohoku, Japan,May 28–June 10, pp. 1343–1348.

êpinska, B., 2003a. Evolution of selected geochemical and reservoir factorsinfluencing the exploitation of the Podhale geothermal system, S-Poland. In: Pro-ceedings of International Geothermal Conference. Session 9, Reykjavik, Iceland,September14–17, pp. 7–14.

êpinska, B., 2003. Geothermal Energy Development in Poland. Reports 2003 Num-ber 2. Geothermal Training Program of the United Nations University, Reykjavik,Iceland, pp. 45–62.

êpinska, B., 2003. The Podhale geothermal system and space heating project,Poland – case study. Reports 2003 Number 2. Geothermal Training Programof the United Nations University, Reykjavik, Iceland, pp. 63–91.

êpinska, B., 2005. Geothermal energy country update report for Poland, 2000-2004. In: Proceedings World Geothermal Congress, Antalya, Turkey, April 24–29,p. 10.

êpinska, B., 2010. Geothermal energy country update report for Poland, 2005-2009.In: Proceedings World Geothermal Congress, Bali, Indonesia, April 25–29, p. 8.

êpinska, B., Bujakowski, W., Ney, R., 2000. Geothermal energy country updatereport from Poland. In: Proceedings World Geothermal Conference, Kyushu-Tohoku, Japan, May 28–June 10, pp. 253–259.

ölbel, T., 2010. Geothermal power plant Bruchsal: construction and initial operat-ing experiences. In: Presented at Second European Geothermal Review, BestecGmbH, Landau, Germany, Mainz, Germany, June 21–23, p. 21.

ölbel, T., Schlagermann, P., Münch, W., Rettenmaier, D., Zorn, R., 2010. Geother-miekraftwerk Bruchsal: erste Messergebnisse (First results of the geothermalpower station at Bruchsal). BBR (Sonderheft (Special edition) Tiefe Geothermie),50–55.

raft, T., Wassermann, J., Deichmann, N., Stange, S., 2009. The 2008 earthquakes inthe Bavarian Molasse Basin – possible relation to geothermics? In: Proceedingsof European Geophysical Union General Assembly Research Abstracts, EuropeanGeophysical Union. Vol. 11, EGU2009-10593-6, Vienna, Austria, April 19–24, p.1.

raft, T., Wassermann, J., Schmedes, E., Igel, H., 2006. Meteorological triggering ofearthquake swarms at Mt. Hochstaufen, SE-Germany. Tectonophysics 424 (3–4),245–258, doi:10.1016/j.tecto.2006.03.044.

ümpel, H.-J., Erzinger, J., Shapiro, S.A., 2006. Two massive hydraulictests completed in deep KTB Pilot hole. Scientific Drilling 3, 40–42,doi:10.2204/iodp.sd.3.05.2006.

wiatek, G., Bohnhoff, M., Dresen, G., Schiulze, A., Schulte, T., Zimmermann, G.,Huenges, E., 2008. Microseismic event analysis in conjunction with stim-ulation treatments at the geothermal reserach well GTGRSK4/05 in GrossSchönebeck/Germany. In: Proceedings 33rd Workshop on Geothermal Reser-voir Engineering, Stanford University, Stanford, CA, USA, January 28–30, pp.391–397.

ED-LGRB, 2010. Yearly earthquake bulletins 1996 to 2008 and recent earth-quake localisations of the Landeserdbebendienst in Baden-Württemberg(LED). Regierungspräsidium Freiburg, Landesamt für Geologie, Rohstoffe undBergbau, Freiburg im Briesgau, Germany. Available at http://www.lgrb.uni-freiburg.de/led pool/led 2 1.htm.

egmann, H., 2003. The Bad Blumau geothermal project: a low temperature, sus-tainable and environmentally benign power plant. Geothermics 32, 497–503,doi:10.1016/S0375-6505(03)00067-1.

eydecker, G., 2009. Earthquake Catalogue for the Federal Republic of Germanyand Adjacent Areas for the Years 800–2007. Federal Institute for Geosciencesand Natural Resources, Hannover, Germany, Available at http://www.bgr.bund.de/cln 109/nn 331926/DE/Themen/Seismologie/Seismologie/Erdbebenauswertung/Erdbebenkataloge/historische Kataloge/germany.html.

ics 41 (2012) 30– 54

Lopez, S., Hamm, V., Le Brun, M., Schaper, L., Boissier, F., Cotiche, C., Giuglaris, E., 2010.40 years of Dogger aquifer management in Ile-de-France, Paris Basin, France.Geothermics 39, 339–356.

Mahler, A., 1995. Geothermal plant in thisted with absorption heat pump and 10years operation without corrosion or reinjection problems in sandstone for 15%saline water. In: Proceedings World Geothermal Congress, Florence, Italy, May18–31, pp. 2161–2166.

Mahler, A., Magtengaard, J., 2005. Geothermal development in Denmark, coun-try update. In: Proceedings World Geothermal Congress, Antalya, Turkey, April24–29, p. 8.

Mahler, A., Magtengaard, J., 2010. Country update for Denmark. In: ProceedingsWorld Geothermal Congress, Bali, Indonesia, April 25–29, p. 9.

Majer, E.L., Baria, R., Stark, M., 2008. Protocol for induced seismicity associatedwith Enhanced Geothermal Systems. Report produced in Task D Annex I (9April 2008). International Energy Agency, Vienna, Austria, 8 pp. Available athttp://www.iea-gia.org/publications.asp.

Majer, E.L., Baria, R., Stark, M., Oates, S., Bommer, J., Smith, W., Asanuma, H., 2007.Induced seismicity associated with enhanced geothermal systems. Geothermics36, 185–222.

Mégel, T., Rybach, L., 2000. Production capacity and sustainability of geothermaldoublets. In: Proceedings World Geothermal Congress, Kyushu-Tohoku, Japan,May 28–June 10, pp. 849–854.

Meghraoui, M., Delouis, B., Ferry, M., Giardini, D., Huggenberger, P., Spottke, I.,Granet, M., 2001. Active normal faulting in the Upper Rhine Graben andpaleoseismic identification of the 1356 Basel earthquake. Science 293 (5537),2070–2073, doi:10.1126/science.1010618.

Moeck, I., Schandelmeier, H., Holl, H.G., 2008. The stress regime in a Rotliegendreservoir of the Northeast German Basin. International Journal Earth Sciences(Geol. Rundsch.) 98, 1643–1654, doi:10.1007/s00531-008-0316-1.

Moia, F., 2008. Individuazione ed applicazione di metodologie di monitoraggio dipossibili fughe di CO2 dai serbatoi di stoccaggio (Identification and applicationof methodologies for monitoring potential CO2 leakage from storage reser-voirs.). Report no. 08001015 CESI RISERCA. ASV Ambiente e Sviluppo Sostenibile,Milano, Italy, 64 pp.

Moia, F., Angeloni, P., Cameli, G.M., Zaninetti, A., 1993. Monitoring induced seis-micity around geothermal fields and reservoirs. In: Presented at First EgyptianConference on Earthquake Engineering, Hurgada – Egypt, 1993, pp. 1–10.

Mortensen, A.K., Egilson, T., Franzson, H., Richter, B., Ásmundsson, K., Danielsen, P.E.,Steingrímsson, B., Thórisson, S., 2006. Hverahlíð - Hola HE-21 (Borehole HE-21at Hverahlíð). Report no. ISOR-2006/018, ÍSOR (Iceland Geosurvey), Reykjavik,Iceland, 79 pp.

Mukuhira, H., Asanuma, H., Niitsuma, H., Schanz, U., Häring, M., 2008. Charac-terization of microseismic events with larger magnitude collected at Basel,Switzerland in 2006. Transactions Geothermal Resources Council 32, 87–94.

Nagy, S., 2007. The simplified estimation of wellhead pressure in reinjection geother-mal processes. Wiertnictwo 24 (1), 341–346.

Nicholson, C., Wesson, R.L., 1990. Earthquake Hazard Associated with Deep WellInjection-A. Report to the U.S. Environmental Protection Agency. USGS bulletinno. 1951. US Geological Survey, Alexandria, VA, USA, 74 pp.

Norden, B., Förster, A., Vu-Huang, D.M.F., Springer, N., Le Nir, I., 2010. Lithologi-cal and petrophysical core-log interpretation in CO2 SINK, the European CO2

onshore research storage and verification project. SPE Reservoir Evaluation andEngineering 13 (2), 179–192.

Onacha, S., Kahn, D., Malin, P., Shalev, E., 2005. Resistivity and microearthquakeimaging of Krafla Geothermal Field, NE Iceland. Transactions GeothermalResources Council 29, 497–502.

Orzol, J., Jung, R., Jatho, R., Tischner, T., Kehrer, P., 2005. The GeneSys-Project: extrac-tion of geothermal heat from tight sediments. In: Proceedings World GeothermalCongress. Paper 1621, Antalya, Turkey, April 24–29, p. 9.

Parker, R.H., 1989. Hot Dry Rock Geothermal Energy: Phase 2B Final Report of theCamborne School of Mines Project. Pergamon Press, Oxford, U.K, 1391 pp.

Pernecker, G., 1996. Altheim geothermal plant for electricity production by organicrankine cycle turbogenerator. In: Proceedings 21st Workshop on GeothermalReservoir Engineering, Stanford University, Stanford, CA, USA, January 22–24,pp. 73–77.

Pernecker, G., Uhlig, S., 2002. Low-enthalpy power generation with ORC turbogen-erator: the Altheim project, Upper Austria. Geo Heat Center Bulletin 23 (1),26–30.

Pine, R.J., Batchelor, A.S., 1984. Downward migration of shearing in jointed rock dur-ing hydraulic injections. International Journal Rock Mechanics Mining Sciences& Geomechnics Abstracts 21, 249–263.

Pine, R.J., Baria, R., Pearson, R.A., Kwakwa, K., McCartney, R., 1987. A technical sum-mary of phase 2B of the Camborne School of Mines HDR Project: 1983-86.Geothermics 16, 341–353.

Pletl, C., Angerer, J., Graf, R., Stoyke, R., Toll, H., 2010. Bohrerfahrungen bei Deutsch-lands grösstem Geothermieprojekt, (Drilling experience from Germany’s largestgeothermal project). BBR (Fachmagazin für Brunnen- und Leitungsbau), Sonder-heft -Tiefe Geothermie 38–47 (June), 84 pp.

Poppei, J., Wilfried, R., Wenderoth, F., 2000. Characterisation of aquifers for theutilisation of low-enthalpy geothermal energy during the individual stages ofsite planning and operation. In: Proceedings 25th Workshop on Geothermal

Reservoir Engineering, Stanford University, Stanford, CA, USA, January 24–26, p.8.

Prevedel, B., Wohlgemuth, L., Henninges, J., Krüger, K., Norden, B., Förster, A., 2008.The CO2-SINK boreholes for geological storage testing. Scientific Drilling 6,32–37, doi:10.2204/iodp.sd.6.04.2008.

therm

R

R

R

R

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

T

T

T

T

K.F. Evans et al. / Geo

aleigh, C.B., Healy, J.H., Bredehoeft, J.D., 1976. An experiment in earthquake controlat Rangely, Colorado. Science 191, 1230–1237.

einecker, J., Tingay, M., Müller, B., Heidbach, O., 2010. Present-daystress orientation in the Molasse Basin. Tectonophysics 482, 129–138,doi:10.1016/j.tecto.2009.07.021.

öckel, T., Lempp, C., 2003. The state of stress in the North German Basin. Erdöl,Erdgas. Kohle 119 (2), 73–80.

ummel, F., Konietski, H., Schäfer, T., 1991. Aussagen zum Spannungsfeld in derForschungsbohrung Urach 3 – Stand: 12/1991 (Evaluation of the stress state ofthe reserach borehole Urach 3 – Status 12/1991). Report no. BMFT 0326598DBundesministerium für Forschung und Technologie, Bochum, Germany,31 pp.

apínska-Slíwa, A., 2003. The effects of developing the geothermal energy in Uniejówtown, Poland. In: Proceedings of International Geothermal Conference, Reyk-javik, Iceland, 14–17 September, pp. 27–33.

apínska-Slíwa, A., Gonet, A., 2010. Geothermal heat and water management inthe view of improving economic conditions of geothermal district heating inUniejów. In: Proceedings World Geothermal Congress, Bali, Indonesia, April25–29, p. 9.

apínska-Slíwa, A., Wojnarowski, P., Gonet, A., 2010. Analysis of Uniejów geothermalwater reservoirs modelling in view of various aspects of geothermal water andheat management. In: Proceedings World Geothermal Congress, Bali, Indonesia,April 25–29, p. 6.

chädel, K., Dietrich, H.-G., 1982. Results of the fracture experiments at the Geother-mal Research Borehole, Urach 3. In: Haenel, R. (Ed.), The Urach GeothermalProject. Schweizerbart’sche Verlagsbuchhandlungen, Stuttgart, Germany, pp.323–343.

chanz, U., Stang, H., Tenzer, H., Homeier, G., Hase, M., Baisch, S., Weidler, R., Macek,A., Uhlig, S., 2003. Hot dry rock project URACH – A general overview. In: Proceed-ings of European Geothermal Conference, Szeged, Hungary, 25–30 May 2003, p.9.

chindler, M., Baumgärtner, J., Gandy, T., Hauffe, P., Hettkamp, T., Menzel, H., Pen-zkofer, P., Teza, D., Tischner, T., Wahl, G., 2010. Successful hydraulic stimulationtechniques for electric power production in the Upper Rhine Graben, CentralEurope. In: Proceedings World Geothermal Congress 2010, Bali, Indonesia, April25–29, p. 7.

chubert, A., Höferle, R., Böhm, F., 2007. Geothermal doublets Munich Riem andPullach i. Isartal in comparison: geological and drilling concepts for minimiz-ing the geological risk in the Deep Malm Aquifer. In: Proceedings of EuropeanGeothermal Congress, Unterhaching, Germany, 30 May–1 June, p. 5.

chulz, R., Jung, R., Pester, S., Schellschmidt, R., 2007. Quantification of explo-ration risks for hydrogeothermal wells. In: Proceedings of European GeothermalCongress, Unterhaching, Germany, 30 May–1 June, p. 6.

egall, P., 1989. Earthquakes triggered by fluid extraction. Geology 17 (10), 942–946,doi:10.1130/0091-7613(1989)017, <0942:ETBFE>2.3.CO; 2.

eibt, P., Kabus, F., Hoth, P., 2005. The Neustadt-Glewe geothermal power plant –practical expeience in the reinjection of cooled thermal waters into sandstoneaquifers. In: Proceedings World Geothermal Conference, Antalya, Turkey, April24–29, p. 4.

eibt, P., Kabus, F., Wolfgramm, M., Bartels, J., Seibt, A., 2010. Monitoring ofhydrothermal plants in Germany – an overview. In: Proceedings World Geother-mal Conference, Bali, Indonesia, April 25–29, p. 7.

eibt, P., Kellner, T., 2003. Practical experience in the reinjection of cooled thermalwaters back into sandstone reservoirs. Geothermics 32, 733–741.

eibt, P., Wolfgramm, M., 2008. Practical experience in the reinjection of thermalwaters into sandstone. In: Proceedings of Workshop for Decision Makers onDirect Heating Use of Geothermal Resources in Asia, Tianjin, China, May 11–18,p. 18.

hapiro, S.A., Kummerow, J., Dinske, C., Asch, G., Rothert, E., Erzinger, J., Kümpel, H.-J.,Kind, R., 2006. Fluid induced seismicity guided by a continental fault: injectionexperiment of 2004/2005 at the German Deep Drilling Site (KTB). GeophysicalResearch Letters 33 (L01309), 4, doi:10.1029/2005GL024659.

tober, I., 1986. Strömungsverhalten in Festgesteinaquiferen mit Hilfe von Pump-und Injektionsversuchen (Fluid flow in crystalline aquifers from pump and injec-tion tests). Geologisches Jahrbuch C42, 3–204.

tober, I., 2005. Hydrogeologishe Untersuchungen im kristallinen Grundgebirge imZuge der Planung für die zweite Tiefbohrung für das HDR-Projekt in Bad Urach(Hydrogeological investigations for the second borehole of the HDR project atBad Urach). Bericht Naturforschende Gesellschaft zu Freiburg im Briesgau 95(1), 237–253.

tober, I., Bucher, K., 2000. Hydraulic properties of the upper continental crust. In:Stober, I., Bucher, K. (Eds.), Hydrogeology of Crystalline Rocks. Kluwer AcademicPublishers, Dordrecht, Netherlands, pp. 53–78.

tober, I., 2011. Depth- and pressure-dependent permeability in the upper conti-nental crust: data from the Urach 3 geothermal borehole, southwest Germany.Hydrogeology Journal 19, 685–699, doi:10.1007/s10040-011-0704-7.

alebi, S., Cornet, F.H., 1987. Analysis of the microseismicity induced by a fluidinjection in a granitic rock. Geophysical Research Letters 14, 227–230.

ang, C., Rial, J.A., Lees, J.M., 2005a. Shear-wave splitting: a diagnostic tool to monitorfluid pressure in geothermal fields. Geophysical Research Letters 32 (L21317),3, doi:10.1029/2005GL023551.

ang, C., Rial, J.A., Lees, J.M., 2008. Seismic imaging of the geothermal field at Krafla,Iceland using shear-wave splitting. Journal Volcanology Geothermal Research176, 315–324, doi:10.1016/j.volgeores.2008.04.017.

ang, C., Rial, J.A., Lees, J.M., Elkibbi, M., 2006. Shear-wave splitting observations andmeasurements at the geothermal field at Hengill, Iceland. In: Proceedings 31st

ics 41 (2012) 30– 54 53

Workshop on Geothermal Reservoir Engineering, Stanford University, Stanford,CA, USA, January 30–February 1, pp. 512–517.

Tang, C., Rial, J.A., Lees, J.M., Thompson, E., 2005b. Seismic imaging of the geothermalfield at Krafla, Iceland. In: Proceedings 30th Workshop on Geothermal ReservoirEngineering, Stanford University, Stanford, CA, USA, January 31–February 2, p.8.

Tenzer, H., Schanz, U., Homeier, G., 2000. HDR reserach programme and results ofdrill hole Urach 3 to depth of 4440 m – the key for realization of a HDR pro-gramme in southern Germany and northern Switzerland. In: World GeothermalCongress, Morioka, Japan May 28th–June 10th, pp. 3927–3932.

Tenzer, H., Schanz, U., Stang, U., Baisch, S., Weidler, R., 2004. Potential for geother-mal power production from deep hot crystalline rock at Bad Urach (SouthernGermany). Zeitschrift für Angewandte Geologie 2, 43–49.

Thouvenot, F., Fréchet, J., Tapponier, P., Thomas, J.-C., Lebrun, B., Ménard, G., Lacassin,R., Jenatton, L., Grasso, J.-R., Coutant, O., Paul, A., Hatzfeld, D., 1998. The Ml-5.3Épagny (French Alps) earthquake of 15 July 1996: a long awaited event on theVuache fault. Geophysical Journal International 135 (3), 876–892.

Thouvenot, F., Jenatton, L., Gratier, J.-P., 2009. 200-m-deep earthquake swarm inTricastin (lower Rhône Valley, France) accounts for noisy seismicity over pastcenturies. Terra Nova 21 (3), 203–210, doi:10.1111/j.1365-3121.2009.00875.x.

Torp, T.A., Brown, K.R., 2005. Underground storage costs as experienced at Sleip-ner and Weyburn. In: Proceedings of 7th Int. Conf. on Greenhouse Gas ControlTechnologies (GHGT-7), Vol. 1, Vancouver, Canada, 5–9 September, pp. 531–540.

Torp, T.A., Gale, J., 2004. Demonstrating storage of CO2 in geologicalreservoirs: the Sleipner and SACS projects. Energy 29, 1361–1369,doi:10.1016/j.energy.2004.03.104.

Turbitt, T., Walker, A.B., Browitt, C.W.A., 1987. Perceptible hydrofracture seismicevents caused by the Hot-Dry-Rock Geothermal Project in Cornwall. GlobalSeismology Report Number 339. British Geological Survey, Edinburgh, U.K., 23pp.

Ungemach, P., Antics, M., 2006. Geothermal reservoir management: a thirty yearpractice in the Paris basin. In: Presented at ENGINE Launching Conference,Orléans, France, February12–15, p. 1.

Ungemach, P., Antics, M., Papachristou, M., 2005. Sustainable geothermal reser-voir management. In: World Geothermal Congress. Paper 0517, Antalya, Turkey,April 24–29, p. 12.

Unger, H.J., Risch, H., 2001. Die Bohrung Simbach-Braunau Thermal 1: Geologie,Mikrofaunen, paläogeographische Einbindung (Palegeographic interpretationof geology and microfauna observations from the Simbach-Brunau Thermal-1borehole). Geologica Bavarica 106, 33–58.

Valley, B., 2007. The relation between natural fracturing and stress heterogeneitiesin deep-seated crystalline rocks at Soultz-sous-Forêts (France). DoctoralThesis. Swiss Federal Institute of Technology (ETH), Zurich, Switzerland,241 pp.

Valley, B., Evans, K.F., 2007. Stress state at Soultz-sous-Forêts to 5 km depth fromwellbore failure and hydraulic observations. In: Proceedings 32nd Workshopon Geothermal Reservoir Engineering, Stanford University, Stanford, CA, USA,January 22–24, pp. 329–338.

Valley, B., Evans, K.F., 2009. Stress orientation to 5 km depth in the basement belowBasel (Switzerland) from borehole failure analysis. Swiss Journal Earth Sciences102, 467–480, doi:10.1007/s00015-009-1335-z.

Van Eck, T., Goutbeek, F., Haak, H., Dost, B., 2006. Seismic hazard due tosmall-magnitude, shallow-source, induced earthquakes in The Netherlands.Engineering Geology 87, 105–121.

Vidal-Gilbert, S., Nauroy, J.-F., Brosse, E., 2009. 3D geomechanical modelling for CO2

geologic storage in the Dogger carbonates of the Paris Basin. International Jour-nal of Greenhouse Gas Control 3 (3), 288–299, doi:10.1016/j.ijggc.2008.10.004.

Vogfjörd, K.S., Hjaltadóttir, S., 2007. Kortlagning skjálfftavirkni við Hverahlið áHellisheiði í febrúar 2006 (Mapping microseismicity at Hverahlíd on Hellisheidiin February, 2006) during pumping of well H-21. Report no. VÍ-ES-04. IcelandicMetereological Office, Reykjavík, Iceland, 20 pp.

Walker, A.B., 1989. SW England seismic monitoring for HDR Geothermal Programme1987–1988. Report no. WL/89/02. British Geological Survey, Survey, Edinburgh,UK, 23 pp.

Wallroth, T., 1992. Hydromechanical breakdown of crystalline rock. A study basedon experiments on hot dry rock geothermal energy. Doctoral Thesis. ChalmersUniversity of Technology, Gothenburg, Sweden, 195 pp.

Wallroth, T., Eliasson, T., Sundquist, U., 1999. Hot dry rock research expeiments atFjällbacka, Sweden. Geothermics 28, 617–625.

Wessling, S., Junker, R., Rutqvist, J., Dmitriy, S., Sultzbacher, H., Tischner, T., Tsang,C.F., 2009. Pressure analysis of the hydromechanical fracture behaviour in stim-ulated tight sedimentary geothermal reservoirs. Geothermics 38, 211–226.

Wieczorek, J., 1999. Geological framework of the Tatra Mountains – Podhale geother-mal system. In: Proceedings of European Geothermal Conference, Vol. 1, Basel,Switzerland, 28–30 September, pp. 261–269.

Wiejacz, P., Debski, W., 2009. Podhale, Poland, earthquake of November 20, 2004.Acta Geophysica Polonica 57 (2), 346–366, doi:10.2478/s11600-009-0007-8.

Wiese, B., Böhner, J., Enachescu, C., Würdemann, H., Zimmermann, G., 2010.Hydraulic characterisation of the Stuttgart formation at the pilot test site forCO2 storage, Ketzin, Germany. International Journal Greenhouse Gas Control 4,960–971, doi:10.1016/j.ijggc.2010.06.013.

Wolfgramm, M., Bartels, J., Hoffmann, F., Kittl, G., Lenz, G., Seibt, P., Schulz, R.,Thomas, R., Unger, H.J., 2007. Unterhaching geothermal well doublet: structuraland hydrodynamic reservoir characteristic: Bavaria, Germany. In: Proceedingsof European Geothermal Congress. Paper 47, Unterhaching, Germany, 30 May–1June, p. 6.

5 therm

W

Z

4 K.F. Evans et al. / Geo

ürdemann, H., Möller, F., Kühn, M., Heidug, W., Christensen, N.-P., Borm,G., Schilling, F.R., 2010. CO2SINK – from site characterisation andrisk assessment to monitoring and verification: one year of opera-tional experience with the field laboratory for CO2 storage at Ketzin,Germany. International Journal Greenhouse Gas Control 4, 938–951,doi:10.1016/j.ijggc.2010.08.010.

immermann, G., Reinicke, A., Brandt, W., Blöcker, G., Milsch, H., Holl, H.-G., Moeck,I., Schulte, T., Saadat, A., Huenges, E., 2008. Results of stimulation treatmentsat the geothermal reserach wells in Gross Schönebeck. In: Proceedings 33rdWorkshop on Geothermal Reservoir Engineering, Stanford University, Stanford,CA, USA, January 28–30, pp. 96–100.

ics 41 (2012) 30– 54

Zinivicius, F., Rasteniene, V., Bickus, A., 2003. Geothermal development in Lithuania.In: Proceedings of European Geothermal Conference, Szeged, Hungary, 25–30May, p. 9.

Zinivicius, F., Sliupa, S., 2010. Lithuania – geothermal energy country update. In:Proceedings World Geothermal Congress Bali, Indonesia, 25–29 April, p. 8.

Zoback, M.D., Harjes, H.P., 1997. Injection-induced earthquakes and crustal stress at

9 km depth at the KTB deep drilling site, Germany. Journal Geophysical Research102 (B8), 18477–18491.

Zumberge, M., Nooner, S., Eiken, O., 2006. A seafloor survey of the Sleipner Fieldto monitor CO2 injection. Final report on contract DE-FC26-02NT41587. ScrippsIntitute of Oceanography, San Diego, California, 15 pp.