The effects of thermal stress and fluid pressure on inducedseismicity during stimulation to production within fracturedreservoirs

Ghazal Izadi and Derek ElsworthDepartment of Energy and Mineral Engineering, EMS Energy Institute and G3 Center, Pennsylvania State University, University Park

Pennsylvania, 16802, USA

ABSTRACT

We use a continuum model of reservoir evolution to explore

the interaction of coupled thermal, hydraulic and chemical

processes that influence the evolution of seismicity within a

fractured reservoir from stimulation to production. Events

occur from energy release on seeded fractures enabling

moment magnitude, frequency and spatial distribution to be

determined with time. Event magnitudes vary in the range

�2 to +2 with the largest event size (~2) corresponding to

the largest fracture size (~500 m) and a prescribed stress

drop of 9 MPa. Modelled b-values (~0.6–0.7) also correspond

to observations (~0.7–0.8) for response in the Cooper Basin

(Australia). We track the hydrodynamic and thermal fronts to

define causality in the triggering of seismicity. The hydro-

dynamic front moves twice as fast as the thermal front and

envelops the triggered seismicity at early time (days to

month) – with higher flow rates correlating with larger mag-

nitude events. For later time (month to years), thermal draw-

down and potentially chemical influences principally trigger

the seismicity, but result in a reduction in both the number of

events and their magnitudes.

Terra Nova, 25, 374–380, 2013

Introduction

Fluid injection at pressures interme-diate between the minimum principalstress and the Coulomb stress willinduce shear failure within enhancedgeothermal reservoirs (EGS) (Hub-bert and Rubey, 1959; Segall, 1989;Majer and Peterson, 2007; Majeret al., 2007) and may trigger seismic-ity (Yerkes and Castle, 1976; Grassoand Wittlinger, 1990). Influence offluid pressures on failure is exacer-bated by the significant changes intotal stress that may result from ther-mal drawdown and potentially fromchemical effects within the reservoir.Shear failure may occur aseismically,but is also manifest as microseismicactivity in many cases.In these cases, the induced seismic-

ity results from fluid injection and isexpected to migrate within the reser-voir with time as driven by thevarious interactions of thermal,hydraulic, mechanical and chemical(THMC) processes, which alsomigrate through the reservoir on dif-

ferent length-scales and time-scales(Walsh, 1965; Goodman, 1976; Bar-ton et al., 1985; Taron and Elsworth,2009; Elsworth and Yasuhara, 2010).Defining the potential causes ofinduced seismic activity due to pro-duction from engineered geothermalsystems is a significant concern –both to understand mechanisms andto mitigate damaging consequences.The size of the resulting seismicevent is defined by the total energyrelease that in turn relates to thestress drop and how fast it fails(Brune and Thatcher, 2002).In the following, we use a contin-

uum model of reservoir evolutionsubject to coupled THMC processes(Taron and Elsworth, 2009) toexplore the evolution of production-induced seismicity in a prototypicalEGS reservoir. In this study, wedefine the relationship between themagnitudes of induced seismic eventsand the applied fluid injection ratesas well as the evolution of thermalstress. Our focus is to determinedominant behaviours controlling thetriggering of induced seismicity thatis unique in comparison with previ-ous studies. We show that inducedseismicity is modulated by hydraulic,thermal and chemical processes,which also migrate through the reser-voir (Taron and Elsworth, 2009;Elsworth and Yasuhara, 2010). We

then explore how pore pressure andthermal stress can be linked to theseismic frequency–magnitude distri-bution, which is described by itsslope, the b-value.

Model dimension and description

Reservoir and fracture networkcharacterization

We assume a doublet geometry(500 m spacing) and the dimensionof the reservoir volume in the modelis 2000 9 1000 9 100 m3 for thehalf-symmetry. Fig. 1 is the represen-tative of the Cooper Basin geother-mal field. The THMC modelevaluates the evolution of flow rate,pressure and temperature distributionduring stimulation. The appliedinjection pressure and temperatureare assumed to be equal to 39.8 MPaand 70 °C. Reservoir pressure andtemperature are assumed to be29.8 MPa and 250 °C respectively.Boundary stresses and the values ofthe solid medium properties utilizedin the simulation for this case aredefined in Table 1. Reactive compo-sition of the host reservoir rock ispresented in Table 2.Calcite and amorphous silica are

expected to be the minerals primarilyresponsible for permeability changedue to precipitation and dissolution.

Correspondence: Ghazal Izadi, Depart-

ment of Energy and Mineral Engineering,

EMS Energy Institute and G3 Center,

Pennsylvania State University, University

Park, PA 16802, USA. Tel.:

001 814 777 4099; e-mail: ghazal.izadi@

gmail.com and gui2@psu.edu

doi: 10.1111/ter.12046

374 © 2013 John Wiley & Sons Ltd

Other likely minerals are also fol-lowed, as listed in Table 2. Rate con-stants for precipitation/dissolutionand mineral reactive surface areas ofthese common minerals are availablein the literature (Kovac et al., 2006;Xu and Pruess, 2001), and were uti-lized as in Xu and Pruess (2004).The present model uses the vertical

stress and two horizontal principalstresses, which are designated as the

maximum and minimum principalstresses. In the process of stimula-tion, fluid pressure is increasedwithin the near-wellbore rock volumewith low permeability, causinghydro-shear displacement on some ofthe pre-existing fractures (fracturepropagation is not considered). Dis-crete penny-shaped fractures areseeded within the reservoir volumeand fractures are distributed withinthe volume with multiple orientationsand a Gaussian distribution oflengths (1–500 m). As the large frac-tures (200–500 m), which representthe weaker planes in the system, arelimited (large spacing) and have thelargest probability to generate theseismic events during change of porepressure as well as temperature alongthese planes, we can say that thegreatest impacts are caused only bylarge fractures and large fractures/faults capable of generating the lar-ger magnitude events. Thus, the

Gaussian distribution is used for thesmall infill fractures, which whenadded to the large fractures give anextended tail representing a log-nor-mal-like distribution. Then, we candescribe the combinations ofdistributions for these variables usingthe mathematical probability func-tion, as we have different indepen-dent variables. The maximum lengthfractures will transit multiple blocksand the smallest fractures will becontained within a single finite differ-ence block. An average fracture den-sity of ~0.1 m�1 is applied to thereservoir and fracture spacing isdefined based on fracture size (10–500 m), so that larger fractures aremore widely spaced – to recover theprescribed fracture density (Fig. 2).

THMC model

Here, we focus on the complex inter-action of coupled THMC processesthat influence the evolution of EGSreservoirs and describe the dominantbehaviours that progress with theevolution of the reservoir. We applya coupled THMC model (Taronet al., 2009) with static–dynamic fric-tional strength-drop to determineenergy release for fractures of differ-ent size embedded within an elasticmedium.Change of stress state is calculated

from the pore pressures, thermaldrawdown and chemical effectswithin FLAC3D-TOUGHREACT(Taron and Elsworth, 2009). Shearfailure calculations are handled withFLAC3D utilizing a Mohr–Coulombfailure criterion (fracture propaga-tion is not considered). The frictionangle on fractures is assumed to beconstant (35°) and during failure, themaximum shear stress drop(~9.0 MPa) is prescribed to representthe residual strength. This model cal-culates the flow rate, pressure andtemperature distribution during stim-ulation. The changes in pressure andtemperature induce displacementsthat consequently lead to a newchange in pressure distribution.These include short-term response

where effective stresses and thermalquenching are expected to dominatethe behaviour of the reservoir andare influenced by the local structurein the rock and orientation of pre-existing fractures. Typical behaviours

σh = 57 MPa

σv = 65 MPa

σhσH = 129 MPa

Tinj = 70 °CPinj = 39.8 MPa

Pprod = 19.8 MPa2000 m

1000 m

Fig. 1 Geometric layout and boundary conditions of enhanced geothermal reser-voirs reservoir as used in the simulation.

Table 1 Parameters utilized in the simulation.

Parameter Unit Cooper Basin

Szz MPa 65

Syy MPa 57

Sxx MPa 129

Fluid pressure (Pinj) MPa 39.8

Injection temperature °C 70

Reservoir temperature °C 250

Bulk modulus of intact rock (Km) GPa 17

Cohesion MPA 10

Poisson’s ratio (υ) – 0.27

Bulk modulus of fluid (kf) GPa 8

Bulk modulus of solid grain (Ks) GPa 54.5

Internal friction angle (φ) 35

Residual friction angle (ß) 11

Coefficient of thermal expansion (aT) 1/°C 1.2 9 10�5

Thermal conductivity (k) W/mK 2.9

Heat capacity (cp) J/kgK 918

Porosity within fractures (/) – 0.3

Table 2 Initial volume fraction of reac-

tive minerals in host reservoir.

Mineral

Volume fraction of solid rock

Granodiorite Fractured vein

Anorthite 0.33 –

Calcite 0.02 0.31

Chlorite – 0.23

K-Feldspar 0.17 –

Quartz 0.34 0.17

Amorphous

Silica

– –

Terra Nova, Vol 25, No. 5, 374–380 G. Izadi And D. Elsworth • Thermal effect on induced seismicity

............................................................................................................................................................

© 2013 John Wiley & Sons Ltd 375

include the reduction in local meanstresses and the development of shearfracturing principally on pre-existingfractures. The very short-termresponse (days to months) (Fig. 3) iscontrolled by fluid pressure andeffective stress effects (HM).Throughout the evolution of the

reservoir, these coupled effects con-

trol the development of permeability,of heat-transfer area, and therebythermal output of the reservoir,together with the evolution ofinduced seismicity. We applied differ-ent models to illustrate the timing ofTHMC effects separately (Fig. 3).Thus, we note the sequencing of fluidpressure effects as fastest-acting fol-

lowed by thermal stresses, thenchemical effects influence the timingand migration of changes in effectivestress within reservoirs as illustratedin Fig. 3. As seismicity is indexed tochanges in effective stresses, theserates of change will influence trigger-ing within the reservoir.

Observation and calibration

We examine the performance of ourmodels against the observed responseof the Cooper Basin geothermal field(Australia) and specifically the pro-gress of seismicity as the reservoir isdeveloped in terms of rates, magni-tudes and locations (Baisch et al.,2006).

Evolution of moment magnitude

The potential energy released withinreservoirs for different fracture spac-ing is defined based on the evalua-tion of the elastic energy releasedfrom the failure of large penny-shaped fractures. Shear stress drop iscalculated in FLAC3D as a functionof normal stress and pore pressurefor a variety of uniform fracturespacings at different locations withinthe reservoir.Shear stress builds in the early

time (<1 month) and reaches a maxi-mum magnitudes in the range~28 MPa as defined by prescribedpeak strengths. This peak strength isdefined by a cohesion of 10 MPa,peak friction angle of 35 and duringfailure, the maximum shear stressdrop (~9.0 MPa) is prescribed to rep-resent the residual strength. Then, we

0

0.2

0.4

0.6

0.8

1

1.2

1 2 3 4 5 6 7 8 9 10

Nor

mal

ized

effe

ctiv

e st

ress

Time (year)

HydraulicHydraulic+ThermalHydraulic+ChemicalHydraulic

Hydraulic+Thermal

Hydraulic+Chemical

THMC Effect

Fig. 3 Normalized effective stress vs. time at 50 m from injection point. Hydraulic,thermal and chemical effects are separately illustrated during 10 years productionand their influence on triggering seismicity. Red solid line, injection pressureassumed 39.8 MPa and reservoir pressure is 29.8 MPa and temperature of injectionis assumed the same as rock temperature (250 °C) and no chemical reaction occurs.Black dashed line, both pressure and temperature changed; injection pressure is39.8 MPa and reservoir pressure is 29.8 MPa, injection temperature is 70 °C andreservoir temperature is 250 °C, again no chemical reaction. Grey dashed line,injection pressure is 39.8 MPa and reservoir pressure is 29.8 MPa, injection tem-perature is the same as reservoir temperature (250 °C), chemical reaction is consid-ered (injection fluid component illustrated in Table 2).

0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

0 5000

50

100

150

200

Fracture size (m)

Num

ber

of f

ract

ures

Fig. 2 Fracture data utilized in the simulation. Fracture density of 0.1 m�1 and these vary in length from 1 to 500 m. Fracturedistribution by location (left) and fracture distribution by size (right).

376 © 2013 John Wiley & Sons Ltd

Thermal effect on induced seismicity • G. Izadi And D. Elsworth Terra Nova, Vol 25, No. 5, 374–380

.............................................................................................................................................................

calculate the potential energy release,Ep, from failure of penny-shapedfractures due to a stress drop Dτ atdifferent locations within the reser-voir. This relation is defined as Ep ¼2Ds2a3=3G, where a is the radius ofthe fracture in the plane and G isshear modulus.Energy release from fractures is

most conveniently represented as amoment magnitude (Keylis-Borok,1959; Aki, 1967; Kanamori, 1977).The moment magnitude relation isdefined as (Purcaru and Berckemer,1978): logM0 = 1.5Ms + 9.1, whereM0 is seismic moment and Ms ismoment magnitude. In this model,M0 is seismic energy, which is derivedfrom the elastic energy released byshear on pre-existing fractures. Thisrelation allows us to determine boththe spatial and temporal evolution ofmoment magnitude in EGS reservoirsand such moment magnitude (Ms)varies from �2 to +2 for the largestfracture size (~500 m).We then use a stress–strain frac-

ture criterion to determine the totalstrain energy available for release –

this energy is assumed to be shedseismically. During the rupture pro-cess, the shear stress drops anamount Dτ from an initial value of τito final value τf, then we can definean expression for total energyET = ∫ DτTDɛdV, where strainchanges from an initial value of ei toa final value ef where V is volume ofthe matrix. Here, we introduce arelation to determine the number ofevents, which occur during the fail-ure process based on potential andtotal energy as Nevent ¼ ET=Ep,where Nevent is the number of seismicevents, ET is the total energy of thematrix block and Ep is the potentialenergy released from the fractures.Here, the energy release of frac-

tures within the reservoir determinesthe evolution of seismic event magni-tude. The release of seismic energy(stress drop = 9 MPa), which is gen-erated in the Cooper Basin EGS res-ervoir due to thermal, mechanicaland chemical effects when seededwith fractures (1–500 m), is illus-trated in Fig. 4. This outcome indi-cates that during stimulation, the

potential energy within the reservoircontaining fracture networks isreleased gradually and extends farfrom injection. The characteristicevent magnitude distribution thatoccurs has a stress drop of ~9 MPa,which implies the occurrence oflarger magnitude events (~1–2) nearinjection at early times (days tomonth) during the stimulation. Thisprocess is controlled by the popula-tion of fracture sizes and the stressdrop.

Empirical relation for triggeringseismicity

The b-value recovered from the anal-ysis is used to calibrate the modelagainst field data. The b-value is anobservable parameter from accumu-lated seismic observations and maybe used to calibrate models – as isattempted here. Aki (1981) specu-lated that the b-value relates to thefractal geometry of the fault planeassemblage and argued that thelog N–M relation for earthquakes isequivalent to a fractal distribution.

1 < M < 20 < M < 1

InjectionWithdrawal

200 600 1600 2000 200 600 1000 1400 1600 2000 200 6001000 1400 1000 1400 1600 2000

1 Month 5 years 10 years

100

400

600

800

1000

Fig. 4 The evolution of moment magnitude in the Cooper Basin enhanced geothermal reservoir over 10 years of simulation.Solid lines illustrate the failure in each location of the reservoir. Blue region illustrates smallest event magnitude and red regionillustrates the largest potential energy that is released in different locations due to combined thermal, mechanical and chemicaleffects.

0

400

600

800

1000

InjectionWithdrawal

1 Month

200

b Value = 0.7

00

0.5 1 1.5 2 2.5 3Moment magnitude

0.5

1

1.5

2

2.5

Log(

Nev

ent)

Fig. 5 Evolution of moment magnitude in the Cooper Basin enhanced geothermal reservoir after 1 month stimulation (left).Number of events as a function of magnitude indicating b-value of 0.7 and moment magnitude range 0–2 (right).

© 2013 John Wiley & Sons Ltd 377

Terra Nova, Vol 25, No. 5, 374–380 G. Izadi And D. Elsworth • Thermal effect on induced seismicity

............................................................................................................................................................

A

BC

ABC

(a)A

(b)

A

B

C

A

B

C

(c)

(d)

(e)

(f)(h)

(g)

Fig. 6 (a) Radius vs. time plot for short term of the progress of fluid and thermal fronts in the reservoir and of induced seis-micity after 10 years production (injection flow rate is 15 l/s). Seismic moment of individual events shown green (smallerevent): 0 < Ms < 1 and red (larger event): 1 < Ms < 2. The hydrodynamic front illustrated with blue and thermal front withpink solid line. Fracture density is ~0.1 m�1 (fractures are in the ranges of 1–500 m, fracture spacings are in the ranges of 10–500 m). We separately zoom in on three windows (considering 2 weeks period at each window) A, B and C at three differenttimes (14 days after stimulation, after 5 years and after 10 years). (b) Seismic activity at early time (2 weeks period). (c) Seismicactivity after 5 years production (2 weeks period). (d) Seismic activity after 10 years production (2 weeks period). (e, f, g) num-ber of events as a function of magnitude indicating b-value of 0.72 at early time, 0.7 after 5 years and 0.68 after 10 yearsproduction (moment magnitude range 0–2). (h) Number of events as a function of moment magnitude for three different times(first month, 5 and 10 years). The number of events with small (0 < Ms < 1) and large (1 < Ms < 2) magnitude shows in threedifferent colours; red, blue and green. Red, number of events at the stimulation period; blue, number of events after 5 yearsproduction; green, number of events after 10 years production.

378 © 2013 John Wiley & Sons Ltd

Thermal effect on induced seismicity • G. Izadi And D. Elsworth Terra Nova, Vol 25, No. 5, 374–380

.............................................................................................................................................................

The b-value is an important parame-ter for the estimation of earthquakehazard. The b-value is related to therock type and the state of stress andincreases with the ductility of therock (Scholz, 1986).The modelled b-value represents

the cumulative number of seismicevents at each location within thereservoir with the local magnitudeevaluated from its seismic moment.Gutenberg & Richter (1944)expressed the magnitude distributionin the form log Nevent = a � bM,where Nevent is the number of seismicevents within a magnitude intervalM � DM. Here, we discuss the scal-ing of the frequency of events Nevent

as a function of fluid pressureand temperature. The magnitude–frequency relation evaluated at earlytime (first month) is illustrated inFig. 5 – left. The approximate b-valueis ~0.7 and this is constrained only byan assumed fracture density. Theshape of the resulting frequency–mag-nitude distribution is similar to thatreported for the Cooper Basin (Baischet al., 2009; Shapiro and Dinske,2009b) and the evaluated b-value inparticular matches the Habanero-1well data. The histogram of eventmagnitudes, which was determinedfor the 2005 data catalogue, indicatesa b-value of 0.8 in the magnituderange �0.8 to +1.5 (Baisch et al.,2009) and the approximate b-valuefrom Shapiro and Dinske (2009b) is+0.75. Events migrate during the sim-ulated evolution due to the progressof fluid pressure diffusion. For thisparticular reservoir, the microseismic-ity begins at the injector and migratesupwards with time. From the Guten-berg–Richter relation, from Aki’s(1981) supposition and from ourobservations, we conclude that theb-value describes fracturing processesin the seismic region and is related tothe size, location, distribution andspacing of fractures.

Effect of hydrodynamic and thermalfront on triggered seismicity

The numerical model is used to fol-low the hydrodynamic and thermalfronts as they propagate through thereservoir. The data used in this simu-lation are related to the Cooper BasinEGS, which is also discussed else-where (Shapiro and Dinske, 2009a).

The location of these fronts isdefined by the propagation of the fluidflow rate at any given time within theTOUGHREACT module (Xu et al.,2006). The evolution of seismicity for10 years reservoir production isshown in a radius-time plot relative tothe propagation of these fluid andthermal fronts in Fig. 6a (injectionflow rate is set at 15 l/s in the simula-tion). This shows the progress of thefluid and thermal fronts in the reser-voir together with the progress ofinduced seismicity. We observe thatthe rate of propagation of the hydro-dynamic front is approximately twiceas rapid as the thermal front.This illustrates that most of the

seismic activity is triggered byhydraulic effects at early times (daysto month) relative to the initiation ofstimulation. At later time (month toyears), thermal effects specifically, orchemical effects possibly, may con-tribute to the seismicity when theseismicity front lags behind thehydrodynamic front due to smallchanges in pressure. By following thepropagation of both fluid pressureand thermal fronts through the reser-voir with time, we associate largeearly-time events with the fluid frontand the lower seismic magnitudelater-time events with the transit ofthe thermal/chemical front.Then, we confirm the form of

observations in our model againstthose reported in Shapiro and Dinske(2009a), which in turn has been cali-brated against field data. We observegood agreement between our modelobservations and those in the CooperBasin (Shapiro and Dinske, 2009a,b).We then compare the magnitude–fre-quency relations evaluated for theCooper Basin stimulation at threedifferent times, Fig. 4 (first month fol-lowed by after 5 and 10 years). Weconsidered a 14-day window at eachtime frame to normalize the numberand magnitude of events at any loca-tion within the reservoir Fig. 4. Thenumber of events as a function of timeis controlled only by the cumulativemass of fluid injected, which thenresults in failure and energy release.The larger this energy release, the lar-ger number of seismic events inducedat a given location and time.Also, we characterize the induced

seismicity by the b-value for thesedifferent time periods. The approxi-

mate b-value is illustrated in Fig. 4and corresponds closely with obser-vations. The plot of event magni-tudes determined here indicates a b-value of 0.72 in the magnitude range0 to +2 at early time, +0.7 after5 years and +0.68 after 10 years.This behaviour shows that the b-value is highest at early time andmeans that the larger events occurearliest (days to month) and due tothe passage of the fluid pressurefront. At later times (month toyears), the principal factor triggeringthe seismicity is thermal effects.Finally, in Fig. 4, we illustrate thenumber of event at each time. Itshows that the largest number ofseismic events exists for 0<Ms<2induced near injection and decreasesaway from the point of injection.

Conclusion

We explore the coupling betweenTHMC behaviours on fractured res-ervoirs to understand the critical pro-cesses during the thermal recoveryfrom EGS reservoirs. Large fractures(faults) with infill fractures (small–large) of various spacing, orientationand fracture placement represent thedistributed fracture network withinthe reservoir. We use this distributionfor a fractured geothermal reservoirin the Cooper Basin to understandthe significant behaviours of THMCeffects on fractured reservoirs. Theactivated fractures due to pore pres-sure propagation and induced ther-mal and chemical strains can generatedifferent failure regimes in the reser-voir and the incremental developmentof energy release of fractures caninduce seismic events with varyingranges of magnitude (�3 to 5) atshort and long time-scales.We capture the energy release of

fractures and then the magnitude isutilized to obtain the magnitude–moment relation and to compare ourmodel with the measured b-valuethat observed at early-time reservoirstimulation (2 weeks) to understandthe induced seismicity at the CooperBasin. This is then extended toexplore the evolution of seismicityexpected over an extended period ofproduction (10 years).Events occur from energy release

on seeded fractures enabling momentmagnitude, frequency and spatial

© 2013 John Wiley & Sons Ltd 379

Terra Nova, Vol 25, No. 5, 374–380 G. Izadi And D. Elsworth • Thermal effect on induced seismicity

............................................................................................................................................................

distribution to be determined withtime. We evaluated the magnitude ofevents and it varies from �2 to +2and the largest event size (~2) corre-sponds to the largest fracture size(~500 m), which represents the weak-est planes within reservoir and a pre-scribed stress drop of 9 MPa.We separately examine the impact

of thermal stresses, pore pressure andchemical effects on the evolution ofseismicity during this long-term pro-duction. The most important mecha-nism, which triggers slip and promotesseismic events along the weak planesand also may change the moment, isthe augmentation of fluid pressure atearly time. However, with migrationof fluid in the fractures at later time,thermal quenching can cause a reduc-tion in effective stress and additionallycontribute to triggered seismicity.Modelled b-values (~ 0.6–0.7) also

correspond to observations (~ 0.7–0.8) and this approach is verifiedusing data for the Cooper Basin(Australia) geothermal field. Wetracked the hydrodynamic and ther-mal fronts to define causality in thetriggering of seismicity.To describe the reservoir from valid

data, we tried to find a good correlationamong spacing, length, orientation andplacement of fractures in the modeland then calibrate the model by repli-cating observed b-values during reser-voir production. The b-value describesthe fracture process within reservoirsand is related to size, location, distribu-tion and spacing of fractures.Finally, we illustrate the penetration

of hydrodynamic and thermal frontsthrough the reservoir with time. Thehydrodynamic frontmoves twice as fastas the thermal front and envelops thetriggered seismicity at early time (daysto month) – with higher flow rates cor-relating with larger magnitude events.For later time (>1 year), thermal draw-down and potentially chemical influ-ences principally trigger the seismicity,but result in a reduction in both thenumber of events and theirmagnitude.

Acknowledgements

This study is the partial result of supportfrom the Department of Energy Office ofEnergy Efficiency and Renewable Energyand Geothermal Technology Programunder contract EE-10EE0002761. Thissupport is gratefully acknowledged.

References

Aki, K., 1967. Scaling law of seismicspectrums. J. Geophys. Res., 72, 1217–1231.

Aki, K. 1981. A probabilistic synthesis ofprecursory phenomena. In: EarthquakePrediction, (D.W. Simpson andP.G. Richards, eds), pp. 566–574. Am.Geophys. Union, Washington D.C.

Baisch, S., Weidler, R., V€or€os, R.,Wyborn, D. and Graaf, L., 2006.Induced seismicity during thestimulation of a geothermal HFRreservoir in the Cooper Basin, Australia.Bull. Seismol. Soc. Am., 96, 2242–2256.

Baisch, S., V€or€os, R., Weidler, R. andWyborn, D., 2009. Investigation offault mechanisms during geothermalreservoir stimulation experiments in theCooper Basin, Australia. Bull. Seismol.Soc. Am., 99, 148–158.

Barton, N., Bandis, S. and Bakhtar, K.,1985. Strength, deformation, andconductivity coupling of rock joints. Int.J. Rock Mech. Min. Sci., 22, 121–140.

Brune, J. and Thatcher, W., 2002.International Handbook of Earthquakeand Engineering Seismology, vol. 81A.International Association of Seismologyand Physics of Earth’s Interior,Committee on Education, pp. 569–588.

Elsworth, D. and Yasuhara, H., 2010.Mechanical and transport constitutivemodels for fractures subject todissolution and precipitation. Int. J.Num. Meth. Geomechs., 34, 533–549.

Goodman, R.E.. 1976. Methods ofGeological Engineering in DiscontinuousRocks. West Publishing, New York,472 pp..

Grasso, J.R. and Wittlinger, G., 1990. 10years of seismic monitoring over a gasfield area. Bull. Seismol. Soc. Am., 80,450–473.

Gutenberg, R. and Richter, C.F., 1944.Frequency of earthquakes in California.Bull. Seismol. Soc. Am., 34, 185–188.

Hubbert, M.K. and Rubey, W.W., 1959.Role of fluid pressure in mechanics ofoverthrust faulting 1. Mechanics offluid-filled porous solids and itsapplication to overthrust faulting. Geol.Soc. Am. Bull., 70, 115–166.

Kanamori, H., 1977. The energy releasein great earthquakes. J. Geophys. Res.,82, 2981–2987.

Keylis-Borok, V.I., 1959. On estimationof the displacement in an earthquakesource and of source dimensions. Ann.Geofisic., 12, 205–214.

Kovac, K.M., Xu, T., Pruess, K. andAdams, M.C., 2006. Reactive chemicalflow modeling applied to injection inthe Coso EGS experiment. In:Proceedings of the 31st workshop ongeothermal Reservoir engineering,Stanford University.

Majer, E.L. and Peterson, J.E., 2007. Theimpact of injection on seismicity atThe Geysers, California GeothermalField. Int. J. Rock Mech. Min., 44,1079–1090.

Majer, E.L., Baria, R., Stark, M., Oates,S., Bommer, J., Smith, B. andAsanuma, H., 2007. Induced seismicityassociated with enhanced geothermalsystems. Geothermics, 36, 185–222.

Purcaru, G. and Berckemer, H., 1978. Amagnitude scale for very largeearthquakes.Tectonophysics, 49, 189–198.

Scholz, C.H., 1986. The frequency-magnitude relation of microfracturingin rock and its relation to earthquakes.Bull. Seism. Soc. Am., 58, 399–415.

Segall, P., 1989. Earthquakes triggered byfluid extraction. Geology, 17, 942–946.

Shapiro, S.A. and Dinske, C., 2009a.Fluid-induced seismicity: pressurediffusion and hydraulic fracturing.Geophys. Prospect., 57, 301–310.

Shapiro, S.A. and Dinske, C., 2009b.Scaling of seismicity induced bynonlinear fluid-rock interaction.J. Geophys. Res., 114, B09307.

Taron, J. and Elsworth, D., 2009.Thermal-hydrologic-mechanical-chemical processes in the evolution ofengineered geothermal reservoirs. Int. J.Rock Mech. Min. Sci., 46, 855–864.

Taron, J., Elsworth, D. and Min, K.B.,2009. Numerical simulation of thermal-hydrologic-mechanical-chemicalprocesses in deformable, fracturedporous media. Int. J. Rock Mech. Min.Sci., 46, 842–854.

Walsh, J.B., 1965. Effect of cracks onuniaxial elastic compression of rocks.J. Geophys. Res., 70, 399–411.

Xu, T. and Pruess, K., 2001. Modelingmultiphase non-isothermal fluid flowand reactive geochemical transport invariably saturated fractured rocks: 1.Methodology. Am. J. Sci., 301, 16–33.

Xu, T. and Pruess, K., 2004. Numericalsimulation of injectivity effects ofmineral scaling and clay swelling in afractured geothermal reservoir. ReportLBNL-55113, Lawrence BerkeleyNational Lab, Berkeley.

Xu, T., Sonnenthal, E., Spycher, N. andPruess, K., 2006. TOUGHREACT—asimulation program for non-isothermalmultiphase reactive geochemical transportin variably saturated geologic media:applications to geothermal injectivity andCO2 geological sequestration.Comput.Geosci., 32, 145–465.

Yerkes, R.F. and Castle, R.O., 1976.Seismicity and faulting attributableto fluid extraction. Eng. Geol., 10,151–167.

Received 14 March 2012; revised version

accepted 13 March 2013

380 © 2013 John Wiley & Sons Ltd

Thermal effect on induced seismicity • G. Izadi And D. Elsworth Terra Nova, Vol 25, No. 5, 374–380

.............................................................................................................................................................