HAL Id: hal-02117944
https://hal.archives-ouvertes.fr/hal-02117944
Submitted on 28 Aug 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
Seismicity and fault aseismic deformation caused by
fluid injection in decametric in-situ experiments
Louis de Barros, Yves Guglielmi, Diane Rivet, Frédéric Cappa, Laure Duboeuf
To cite this version:
Louis de Barros, Yves Guglielmi, Diane Rivet, Frédéric Cappa, Laure Duboeuf. Seismicity and fault
aseismic deformation caused by fluid injection in decametric in-situ experiments. Comptes Rendus
Géoscience, Elsevier Masson, 2018, 350 (8), pp.464-475. �10.1016/j.crte.2018.08.002�. �hal-02117944�
Internal
Geophysics
(Seismology)
Seismicity
and
fault
aseismic
deformation
caused
by
fluid
injection
in
decametric
in-situ
experiments
Louis
De
Barros
a,*
,
Yves
Guglielmi
b,
Diane
Rivet
a,
Fre´de´ric
Cappa
a,c,
Laure
Duboeuf
aa
Universite´ Coˆted’Azur,CNRS,ObservatoiredelaCoˆted’Azur,IRD,Ge´oazur,SophiaAntipolis,06560Valbonne,France
b
EarthandEnvironmentalScienceArea,LawrenceBerkeleyNationalLaboratory,Berkeley,CA,USA
c
InstitutuniversitairedeFrance,75231Pariscedex05,France
1. Introduction
Fluidsarepresentintheuppercrustandareusually relatedtoeitherunderground resources(water,oil,and
gas)ornaturalhazards.Onemajorprobetoinferwhere fluidsarepresent andhow theyare movingatdepth is throughseismicmonitoringandimaging.Forinstance,in natural hazard assessment,seismicity is indeedusedto monitor volcanic eruptions and magma upwelling (De Barrosetal.,2013;McNutt,2005).Anthropogenic activi-ties in reservoir operations such as fluid storage or extraction alsoinduceseismicity.Forexamples,induced seismicity hasbeenobserved in deepgeothermalfields
ARTICLE INFO
Articlehistory: Received1stJune2018
Acceptedafterrevision20June2018 Availableonline1September2018 HandledbyVincentCourtillot Keywords:
Fluid-inducedseismicity In-situdecametricexperiments Aseismicdeformation Stresstransfer
ABSTRACT
Seismicityinducedbyfluidperturbationsbecameanimportantsocietalconcernsincefelt earthquakes(Mwupto6)occurredafteranthropogenicactivities.Inordertomitigatethe
risksassociatedwithundesiredseismicity,aswellastobeabletousethemicro-seismicity asaprobeforin-depthinvestigationoffluid-drivenprocesses,itisofcrucialimportanceto understandthelinksbetweenseismicity,fluidpressureandflow.Wehavedevelopeda seriesofin-situ,decameter-scaleexperimentsoffaultzonereactivationbycontrolledfluid injection, in order to improve the near-source geophysical and hydromechanical observations.Thedeployedgeophysicalmonitoringclosetotheinjectionallowsoneto coverthefullfrequencyrangeofthefaultresponsesfromthestaticdeformationtothe very high-frequency seismic emissions (up to 4kHz). Here, we focus on the microseismicity (Mw–4 to –3) recorded during two fluidinjection experiments in
low-permeableshaleandhighly-fracturedlimestoneformations.Inbothexperiments,the spatio-temporaldistributionoftheseismicevents,theenergybalance,andtheseismic velocitychangesofthefracturedmediumshowthatmostofthedeformationdoesnot actuallyemitseismicsignals.Theinduceddeformationismainlyaseismic.Basedonthese high-resolutionmultiparametricobservationsinthenear-field,wethereforeproposeda newmodelforinjection-inducedseismicity:theseismicityisnotdirectlyinducedbythe increasingfluidpressure,butitisrathertriggeredbythestressperturbationstransferred fromtheaseismicmotioncausedbytheinjection.
C 2018Acade´miedessciences.PublishedbyElsevierMassonSAS.Thisisanopenaccess
articleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/ 4.0/).
* Correspondingauthor.Norsar,GunnarRandersvei15,POBox52, N2027Kjeller,Norway.
E-mailaddress:debarros@geoazur.unice.fr(L.DeBarros).
ContentslistsavailableatScienceDirect
Comptes
Rendus
Geoscience
ww w . sci e nc e di r e ct . com
https://doi.org/10.1016/j.crte.2018.08.002
1631-0713/ C 2018Acade´miedessciences.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http:// creativecommons.org/licenses/by-nc-nd/4.0/).
(e.g.,Lengline´ etal., 2017;Wei etal., 2015), duringgas storageintodepletedreservoir(e.g.,Cescaetal.,2014),and inducedbyCO2geologicalstorage(Payreetal.,2014)or
gas and hydrocarbon extraction (Albano et al., 2017; Bardainne et al., 2008). In recent years, unconventional shale gas exploitation strongly increases the seismicity rate around the reservoirs, in particular in the central UnitedStates and WesternCanada(e.g.,Baoand Eaton, 2016).WhileinOklahoma(USA)theseismicityismainly inducedbythelargevolumeofwastewaterinjectedbelow theshalelayers(e.g.,Keranenetal.,2014;Schoenballand Ellsworth,2017),theseismicityinWesternCanadamight beproducedbythefrackingprocessitself(Atkinsonetal., 2016). Earthquakes are also observed during and after hydraulic dam fillings (e.g., Gupta et al., 2017). Recent observationsatfluidinjectionsitesindicatedthatinduced earthquakes canreach magnitudesuptoMw6,asthe
Prague and Pawnee earthquakes in Oklahoma (USA). Therefore, seismichazard due toanthropogenicactivity isamajorconcernandshouldbemitigated.
It is now well known that fluid pressure induces failures,bydecreasingtheeffectivenormalstressactingon faultsandfractures,whichbringsthestressstatecloseto thefailure envelope.Therefore,whenthefluid pressure reaches a failure threshold, an earthquake occurs if sufficient frictionalweakening occurs. Within this idea, theseismicityshouldfollowthefluidpressurediffusion, andthemeasureoftheseismicityfrontcouldleadtoan estimation of the diffusivity of the medium (Shapiro, 2015).Consequently,asthesizeofthepressurizedvolume increaseswiththevolumeofinjectedfluids,thelengthof the faults which get close to failure also increases. Therefore, the maximum magnitude of the induced seismicityshouldscalewiththeinjectedvolume(McGarr, 2014).
However, the response of a faulted mediumto fluid pressureperturbationcanbemuchmorecomplex.Firstly, the seismicitymayoccurfarand muchdeeperfromthe injection, as observed in Oklahoma, where wastewater disposalsinducedseismicitymorethan40kmawayfrom theinjection(SchoenballandEllsworth,2017;Yecketal., 2017).Theseismicitymayalsobedelayedintime,withthe largesteventoccurringaftertheshut-inoftheinjection,as observed at the Basel geothermalfield (Mukuhira etal., 2013).Furthermore,onlyafractionoftheenergyinjected intothereservoirsisconvertedtoseismicity,asalargepart of the deformation may not express seismic signatures (Guglielmi et al., 2015a; Schmittbuhl et al., 2014). For instance, a large aseismic motion was induced by the geothermal field of Brawley in southernCalifornia(Wei etal.,2015),which,inturn,triggeredtwoM5 earthqua-kes.Atthelaboratoryscale,Goodfellowetal.(2015)showed thattheseismicenergyrepresentslessthan1.2510 4%of
the injection energy during hydraulic fracture tests, highlightingthataseismicdeformationrepresentsalarge partofrupturemechanisms.
Therefore,therelationshipsbetweenfluids, earthqua-kesandaseismicdeformationarestillunderdebate,and detailedobservationsnearinjectionwellsarerare,evenif the understanding of induced seismicity is crucial for reservoirengineeringand hazardmitigation.Inorderto
get well-constrained data close to injection and to investigatetheseismologicalandhydromechanical behav-iorof the pressurized rockvolume, we have developed aseriesofinnovativein-situexperimentsatadecameter scale. The main idea is to reactivate well-identified geologicalstructuresbelongingtomaturefaultzonesin shaleandlimestoneatabout300mdepth.Todoso, high-pressurewaterisdirectlyinjectedintothem.Aroundthe injection zone, a dense network of geophysical sensors allowsthesimultaneousmonitoring ofthehydrological, mechanical,andseismologicalresponsesoftheruptured faults overa broadfrequency band, fromstatic tohigh frequency (10kHz). Such experiments therefore benefit fromrealisticin-situconditions,well-characterized geo-logicalstructures,controlledhydraulicperturbations,and amulti-parametricmonitoringnetworkatclosedistance (metertodecameter)fromtheinjection.
In this paper, we first describe the two injection experimentsperformedinfaultzonesnestedinlow-and high-permeability rocks, together with the seismicand hydromechanical responses of the tested geological structures. Importantly,theseexperiments providenew completedatasetforseismologicalandhydromechanical characterizationandshowunequivocallythatmostofthe deformationinducedbythefluidinjectionisaseismic.In thelightofthecoupledanalysisofthedata,wetherefore suggestthattheseismicityisonlyasecondaryresponseto thefluidpressure,andthatseismicityisrathertriggeredby aseismicdeformationcausedbytheinjection.
2. Dataandexperiments
2.1. Geologicalcontextandinjectionprocedure
Theexperimentsaimatinjectinghigh-pressurewater (1to5MPa)intoselectedexistinggeologicalfeatures(e.g., fault,fracture,beddingplane),inordertoinducedslipson the tested structures. Those experiments took place in undergroundresearchlaboratories,atabout300mdepth. Thisdepthallowsastressstatethatfavorsshearmotionsof thetestedstructures.Fromthegalleries,boreholeswere drilled,tointersectthedifferentgeologicalstructuresofa maturefaultzone.Insidetheinjectionborehole,theSIMFIP probe (Guglielmi et al., 2013) isolated a 2.5-m-long chamber witha straddlepacker system.The probeis a newborehole tool that allows water injection together withsimultaneousmeasurementsofthefluidpressureand of3Dmechanicaldisplacements.Withthis probe,water wasinjectedintothegeologicalstructuresthatcrosscut thesealedsectionoftheborehole.Theinjectionsequences weredesign to(1) test thepore-elasticresponseof the mediumwithlow-pressureinjections(e.g.,pulsetestsof short duration, some minutes) and (2) induce the mechanical shearing and opening of existing fractures with a step-by-step increasing pressure. The pressure increase however stays below the fracturing pressure, sincetheaimofsuchexperimentsistoreactivateexisting faultsandisnottoinducefluiddrivenhydro-fractures.
The first experiment of this type was performed in 2010 into the fractured limestone of the Low Noise
UndergroundLaboratoryofRustrel(southernFrance).The experiment consisted in injecting water in a high-permeablefaultzone.Thehydromechanicaland seismo-logicalmeasurementsshowedunprecedentedanddirect evidenceofanaseismicmotionprecedingseismicfailures (Guglielmi et al., 2015a). However, this experiment sufferedfroma verylimitednumberofseismicsensors, which prevented the in-depth analysis of the seismic responses(Derodeetal.,2015).Toresolvethislimitation, similarexperimentstookplaceagaininthesamefacility,in 2015, with a denser monitoring network made up of 31seismicsensorsaroundtheinjection(from1to20m) (Duboeuf et al., 2017). Injections were performed at 11 differentlocations(Fig.1A),totest theresponsesof
either(1)thesub-horizontalbeddingplanesthatseparate layers withdifferent facies,and (2) theminorfaults or fractures,belongingtotheextended damagezonesof a kilometer-long fault (Jeanne et al., 2012). Results con-firmed that most deformation is aseismic in these experiments,withseismicitymainlyoccurringatdistance (1to12m)fromtheinjection.
In order to test another lithology and investigate seismicityinlow-permeableformations, asetof experi-ments(DeBarrosetal.,2016;Guglielmietal.,2015b)was conducted in 2014 in the Toarcian shale of the IRSN underground experimental platform of Tournemire (France). Here, the injection borehole fully crossed a kilometer-longfaultzone,and4injectionswereperformed
Fig.1. Experimentalsetupoftheexperiments.A.Mapviewofthegalleryandverticalcross-section,showingthelocationsofthemonitoringnetworkand theinjectiontests,intheRustrellimestone(seeDuboeufetal.(2017)fordetails).B.MapviewoftheexperimentintheTournemireshalewiththeschematic structuresofthefaultzoneandthelocationoftheinjectiontestsandtheseismicsensors(seeDeBarrosetal.(2016)fordetails).C.Schematicviewofthe SIMFIPinjectionprobe,anddetailsontheoptic-fiberstrainmeteranchoredintheinjectionchamber.
indifferentgeologicalstructures:(1)minorfaultsinthe damagedzones,(2)theunfracturedmedium,(3)fractures inthedamagedzone,and(4)thefaultcore(Fig.1B).Intop of the responses of those isolated structures, the full thicknessofthefaultzone,includingthefaultcoreandthe western damages zone, was tested using a 15-m-long portionoftheboreholeasinjection.
2.2. Monitoringnetworkanddata
Both injection experiments (limestone in Rustrel, 2015andshaleinTournemire,2014)wereinstrumented withadensemonitoringnetworkinordertorecordinhigh detail the hydrological, mechanical, and seismological responses(Fig.2)ofthehydraulically-stimulatedstructures. Insidetheinjectionchamber,thenewlydevelopedSIMFIP probe allowed the simultaneous measurements of the injectedpressure,theflowrate,andthe3Dmotionsofthe injected structures. This borehole strainmeter (Fig. 1c), based on optical-fiberBragg gratingmeasurements, was anchoredon theboreholewalls,independentlyfromthe injection system (Guglielmi et al., 2013). It captured deformations assmallas1
m
m/m.Atafewmeters from theinjection,deformationwasalsorecordedbythree two-componentstiltmeterssetonthegalleryfloorintheRustrel experiment.AtTournemire,aboreholewasequippedwith 40extensometricgauges,distributedona12-m-longprobe.Surrounding theinjection, theseismicitywasmainly recordedbyadensenetworkof22and16accelerometers, fortheRustrel(Fig.1A)andTournemire(Fig.1B) experi-ments,respectively. These sensors,witha flat response between 2Hz and 4kHz, were preferred to classical geophones, in order to have a much wider and higher frequency response. Fig. 3 shows the time and the frequencysignatureofaseismiceventrecordedatRustrel. Below500Hz, accelerometers and seismometers have similarresponses. Abovethis frequency,while the geo-phoneislosingsensitivity,theaccelerometersfullycatch the event waveform, including its frequency corner at around1.2kHz.Tocomplementthisbackbonenetworkat lowerandhigherfrequency,afewgeophones(10–800Hz) andacousticsensors(10Hz–10kHz)werealsodeployed (Fig.1).
All these data were synchronously recorded, at samplingfrequencies goingfrom 1kHz for hydrological andmechanicaldatato2,10or20kHzforseismological data.Fig.2summarizesthismulti-parameterdataset,by showingdatarecordedbydifferentsensorsduringpartof aninjectioninRustrel(2015).Inadditiontothepressure andflowrate,whicharemorecommonlyrecordedduring reservoir-scalefluidinjection,thevarietyofsensorsused hereallowonetospanthefullfrequencyrangeofthefault responsesfromthestaticdeformationtothevery high-frequencyseismicemissions.
Fig.2.Exampleofdatarecordedduringstep-rateinjectionsintest11,Rustrel.Toppanel:hydraulicdata,withpressureandflowrateattheinjection; middlepanel:quasi-staticdata,withonecomponentofthedeformation(INJ)attheinjectionpointandofatiltmeter(TILT)setonthegalleryfloor;bottom panel:high-frequencydata,withhigh-passfiltereddeformation(Inj,10–500Hz)attheinjectionpoint,geophone(Geo,10–800Hz),accelerometric(Acc, 0.1–4kHz)andacoustic(AE,0.2–10kHz)data.Thelightblueverticallinesshowthedetectedseismicevents.
Finally, in both experiments, active seismic sources were also installed. The repetitive signals from these sources(hammershotsinTournemireandvibratingdevice inRustrel)arethenusedtomonitorsmallvelocitychanges associated with medium perturbations during fluid injectionsandfaultreactivations.
3. Experimentalresults 3.1. Seismicsignatures
Inbothexperiments,seismiceventsweredetectedby amplitudethreshold methodsand then checked by eye-screening. In the Tournemire dataset, to identify small amplitudesignals,thealreadydetectedeventswereused as templates and cross-correlated with the continuous seismicrecords(template-matching,GibbonsandRingdal, 2006). At theend, only34 events weredetected in the Tournemiredatasetduringthreeinjectiontests(DeBarros etal.,2016).Theothertwotestsshownoorsingleseismic emission.Outofthe11testsperformedinRustrel,sixtests showseismicity,withatotalof215events(Duboeufetal., 2017).AnexampleofseismiceventisshowninFig.3.
P-andS-wavearrivaltimesandP-polarizationswere used to get an absolute location for 24 events in the Tournemire experiment. One hundred and thirty-seven events from Rustrel were located both absolutely and relatively using arrival times and inter-event delays
measuredbetweensimilarevents.Thefocalmechanisms, assumingapuredouble-couple mechanism,were deter-minedusingeitherthefirstpeakamplitudesfor16events intheTournemiredataset,orthefirst-motionpolarityfor 59eventsinRustrel.
The event magnitudes lie between Mw=–3.2 and
Mw=–4.2.Forbothexperiments,thedetectionthreshold
is around Mw–4, at least around the injections.
Frequency corners are found to be greater than 1kHz, whichleadstoruptureareawithanestimatedradiusof about0.1mto0.3mforacircularcrack.Suchseismicityis quitesmallcomparedtoreservoir-scaleseismicity,usually greater than Mw–3 or tectonic seismicity (Mw>0).
However,suchmagnitudeandsourcesizeareconsistent with classical scaling laws for shearing events (e.g., Madariaga, 1976), witha stress dropof about 0.1MPa. Nomajordifferencesinmagnitudesorfrequencycorners wereobservedbetweenexperiments.
Unconventionalseismicity,likelow-frequencyevents, tremorsignalsorlong-period/long-durationsignalshave been observedaround fluid injection in reservoirs(e.g., Kumaretal.,2017). Theyareinterpretedastheseismic signature of slow-slip failures, but their existence and interpretationarestillunderdebate(Zecevicetal.,2016). Inthedatasetsrecordedduringthe2010Rustrel experi-ments, Derode et al. (2015) identified suchevents. The sourceprocessbehindthemwas,however,notclear,asthe numberofsensorsdidnotallowadetailedanalysis.Some low frequency events,witha peaked frequencycontent
Fig.3.ExampleofaMw–3.6seismiceventrecordedintheRustrelexperiment.(Left)Accelerometricdata,sortedaccordinglytothehypocentraldistance.
(Right)Waveformsandspectrum,recordedbyageophoneandanaccelerometer(aftertimeintegration)setatthesamelocation.Thecornerfrequencyof thiseventisaround1.2kHz.
at400Hz,werealsodetectedintheTournemiredataset. However, thelocationofthose eventsshowedthat they occurred in or around boreholes. They are likely to be producedbyboreholewallresonanceinducedbypressure change (Tary et al., 2014). Therefore, unconventional seismic signals cannot be clearly linked to the fault reactivationinourdatasets.
3.2. Deformationattheinjectionpoints
Hydromechanical responses strongly differ between thetwoexperimentalsites.IntheshaleofTournemire,at low pressure,theflowratewasverysmall,indicating a very low initial fracture permeability. Once a pressure thresholdwasreached,thetestedfracturesslippedwith somedilatancy,leadingtoastrong increaseinflowrate andpermeability.Thispressurethreshold,calledFracture Opening Pressure(FOP;Zoback, 2010),clearlyseparated theporo-elasticbehaviorneartheinjectionborehole,from theplasticbehaviorononeorafewfractureswhichcould extend several meters from the injection. Logically, seismicity only occurred once the FOP was reached. Maximum displacement measured between 0.08 and 0.55mm with the SIMFIP probe was observed at the injections.At theend of thetests,the0.05 to0.52mm residual displacements confirmedthat a plastic motion wasinducedforalltests,eveniftwooutofthefivetestsdid notshowanymeasurableseismicevents,highlightingthat aseismicmotionsoccurduringinjections.
The fractured limestone of Rustrel, belonging tothe unsaturatedzoneofthemassif,behavesdifferently.Before anyinjectionswereperformed,somefracturesalreadyhad amillimeteropening.Therefore,evenatlowpressure,the flow rate was high: for four tests, injecting in highly permeable fractures led to a 70L/min flow rate, while injectionpressureremainedbelow2MPa.Noclearfailures wereobservedinthosetests,andnoseismicityoccurred. Onthecontrary,forsixothertests,thepressurereached high value (5.5MPa), with a flow rate increasing with pressure. Within these six tests, we observed a strong increase of permeability for two of them. A 20-fold permeability increase was also measured between the beginningandtheendofthetestperformedinRustrelin 2010(Guglielmietal.,2015a).Mechanicalfailuresatthe injectionpointsareconfirmedbyresidualdisplacements of0.01to0.05mm.However,contrarytotheTournemire tests,thepressurethresholdbetweentheporo-elasticand theplasticresponsescannotbeclearlydefinedusingeither the hydraulic or the deformation data. It suggests a transitional behavior, rather than thebinary open/close process observed in the Tournemire experiments. For thosesixtests,aswellasforthe2010test,atleastafew seismicevents,andupto120events,havebeenobserved. 3.3. Partitioningbetweenseismicanddeformationenergy
Ourobservationsshowthattheseismicityisunevenly distributed among tests. Some tests are fully aseismic, even if residual displacements are measured at the injectionpoints,indicatingaplasticmotionontheinjected structures.Moreover,thereisaclearlackofseismicitynear
the injections, and the seismicity is also unevenly distributedinspace,withsomeareaswithoutanyseismic sources. As instance, the seismicity in the Tournemire experimentsoccurredonlyintheeasterndamagezoneof themainfault.Wecanthereforewonderwhatfractionof thedeformationisactuallyemittingseismicsignals.
Toevaluatethecontributionof theseismicitytothe deformation, seismic moments are compared to an equivalent moment computed from the deformation (Fig.4).ThelattercanbecomputedasM0def=
m
DS,withm
theshearmodulus.ThedisplacementD,measuredatthe injections,canbeafractionofeitherthemaximumorthe residual displacement, by considering either the total motion or only the shearing part.The surface S of the shearing zone can be estimated by assuming that the seismiceventsarealllocatedeitherinsideoraroundit. Therefore, both D and S are rough estimations. They, however,showthatlessof0.1%ofthedeformationenergy is actually emitting seismicity in theTournemire shale. This ratio is higher, but still very small, in the Rustrel limestone,asseismicityrepresentslessthan6%inaverage. Therefore, in both cases, most of the deformation is aseismicduringfluidinjections.Based on our observations, we suggest that not consideringaseismic deformation induced by injections canleadtoanoverestimationofthemaximummagnitude prediction as proposed by McGarr (2014). The authors linkedthemaximumseismicmomentM0maxofinduced
earthquakestotheinjectedvolumeVbyM0max=
m
V,withV theinjected volume. The maximum seismic moment observed here is indeed linearly related to the volume (Fig.4).However, theMcGarr(2014)theoretical predic-tionsaresixordersofmagnitudesaboveourobservations. Thisstrongdiscrepancymaybeexplainedbythestrong aseismic component of the deformations, as McGarr (2014) considered that all deformations are emitting seismicity.
3.4. Locationofthedeformation
Repetitive,activeseismicsourceswereshotduringthe experimentstoimageseismicvelocityperturbationsthat testifytochangesinthemechanicalpropertiesofthefault zone.Monitoringseismicvelocitychangeshasprovenits efficiencyingivingvaluableinformationaboutactivefault behavior with either seismicslip (e.g.,Brenguier et al., 2008),aseismicmotion(e.g.,Rivetetal.,2011),or pore-pressurechanges(e.g.,Hillersetal.,2015).Insuchinjection experiments,withadensemonitoringnetworkaroundthe fault, seismic velocity change measure aims to detect aseismic deformation processes distributed around the reactivatedstructures.
The active seismic signals recorded during injection periodswerecompared withreferencesignals recorded priortothetestsandconsideredasthebaselineresponses ofthemedium.Thedelaybetweensignalsaretheninferred fromthephasedifferencesbyinterferometrictechniques. AsP-andS-wavesdonothavethesamesensitivitywith respectto deformationand fluid content,we measured travel-time changes for both phases separately. Using tomography methods, these P- and S- wave velocity
changeswerelocatedinthemedium,byconverting travel-timedifferencestovelocitychanges(Rivetetal.,2016).
Fig.5presentsimagesoftheseismicvelocitychanges measured during the first injection test of Tournemire experimentinasetofminor,secondaryfaultsbelongingto thewesterndamagedzoneandlocatedatthecenterofthe seismicsensorarray.Thistestdoesnotshowclearseismic emissions,despite plasticdeformations observed at the injection point. First, below the FOP (i.e. before plastic deformationon the injectedstructures), both P- and S-wavesshowaslightincreasearoundtheinjectionchamber (Fig.5a,c).Thepressureinthechambermakesitinflate, which increasesthe stress in thesurroundingmedium. AbovetheFOP,velocityperturbationsaremuchstronger (upto5%,Fig.5b,d).AstrongdecreaseofP-andS-wavesis associatedwiththedilatancyoftheminorfaultsthatare aseismically slipping in response to the injection. The plurimetric extension of the perturbations (i.e. of the slipping area), is confirmed by numerical modeling of fluid-assistedstressperturbations(Rivetetal.,2016).The dilatantshearonthesefaultsinducesastresstransfer,with astressincreaseinthesurroundingareas,associatedwith apositiveperturbationofthevelocity.
The seismic velocity changes are therefore related either to the dilatancy of the slipping faults or to the opening/closingof micro-cracks due tostress perturba-tions. As no seismicity was recorded during this test, velocitychangesappear asa goodprobe tomonitor (1) aseismicdeformationand (2) stresstransfer.Across the
testedfault,averagingthepositiveandnegative perturba-tionsleadstoasmalldecreaseofvelocityofabout0.1%. This agrees with the perturbations measured across tectonicfaultzones,andinducedby earthquakes( Bren-guier et al., 2008). Therefore, the small negative per-turbationsobservedafterearthquakesmightbeaspatial averageofstrongerandlocalizedperturbations.
4. Discussion
Our experiments only induced a sparse seismicity, despite plastic behaviors at the injections. Particularly, thereisalackofseismiceventsneartheinjectionpoints andanunevenspatialdistributionaround.Thenumberof seismiceventsstronglydiffersfromonetesttoanother, with some tests without any seismicity. Therefore, the spatio-temporal distribution of the seismicity is very heterogeneous. Assuming that thefluidpressure, which reducestheeffectivestress,directlytriggerstheseismicity, theearthquakesshouldmoveawayfromtheinjectionwith the square (or cubic) root of time for a poro-elastic diffusionofthefluid(Shapiro,2015).Suchdistance-time plotisgiveninFig.6,withanormalizeddiffusivitytomake the different tests comparable. The actual diffusivity is foundtorangebetween0.0015and0.13m/s2amongtests,
whichagreeswithvaluesusuallymeasuredingeothermal areas. However, this diffusive pattern is not clear, as seismicitydoesnotgatherbehindaseismicfront.Forsome tests(e.g.,test2,Rustrel),seismicityoccurredclusteredin
Fig.4.Cumulatedandmaximumseismicmoment(redandgreensymbols),deformationmomentandpredictedvolumemoment(blueandblacksymbols) versusinjectedvolumeVforthedifferenttestsofRustrel(square)andTournemire(circle).Thevolumemomentisthemaximumseismicmomentfollowing
McGarr(2014)relationships,asM0max=mV.TheshearmodulimaredifferentforTournemireandRustrel.Thedeformationmomentiscomputedby
Fig.5. MapviewoftheP-(leftpanels,aandb)andS-wave(rightpanel,candd)velocityvariationsobservedduringtest1,Tournemire(seeFig.1A).Theblue rectangleshowstheinjectionchamber.aandc:velocitychangesforaninjectionpressurebelowtheFOP;bandd:velocitychangesforaninjectionpressure abovetheFOP,i.e.whenaseismicfailureswereobserved.
Fig.6.Spatio-temporaldistributionoftheseismicity.DistanceRbetweenhypocentersandinjectionpointsversustimetwherethepressureisabove
s3.Thetimehasbeennormalized,inordertomakethedifferenttestscomparable,withadiffusivityD=1m/s2.Thisnormalizationcoefficientisthemean
forthethreeeventswiththelargestdiffusivity.ThedottedlineisthereforethetheoreticalpredictionR=sqrt(4pDt),withD=1m/s2
.Thecolor-codestands forthetestnumber,ineitherTournemire(crosssymbols)orRustrel(filledsymbols).
time but spatially scattered. Test11 seismicity (Rustrel) seemstofollowtwo differentdiffusive patterns,witha higherdiffusivityatearlytimethanlaterinthetest,which isinconsistentwiththeobservedincreaseofpermeability. Therefore,aporo-elasticdiffusionistoosimpletomodel thefullcomplexityofthemediumresponse.Ontopofthat, themediumischangingwithfluidinjection,asnewfluid pathsarecreated.Theintertwinedrelationshipsbetween injectionparameters,geologicalsettings,aseismic defor-mationandseismicityarethereforecomplex.
Basedonouroriginalexperiments,therespectiverole oflithological, hydraulicand frictionalpropertiesof the faultzonecanbebetterconstrained.Firstly,weobserve that the lithological properties influence the seismic behavior. For similar fluid perturbations, seismicity production is smaller in the Tournemire shale than in the Rustrel limestone. Similar experiments, also at decametricscale,but incrystallinerocks reveala much numerousseismicity(Jalalietal.,2017;Zangetal.,2016). Besides,intheTournemireshale,11outof16eventswith computedmechanisms occurred on thesame familyof fractures(Fig.7),whicharetheonlyonesintheareawitha calcite filling. While the shale hasa rate-strengthening behavior, the calcite fillings may be rate-weakening. Therefore,seismicityoccurredonlyongeological structu-reswithanadequatemineralogyandfrictionalbehaviors. InRustrel, morenumerous seismicityis observedinthe layerswiththehighestdensityof fractures.Thedensity and frictional properties of the fractures around the injection are then one controlling factor of seismicity production.
Fracture permeability also has a strong effect on seismicity. For both experiments, injections into areas with higher permeability fractures induce very little seismicity.Asthefluidflowsrapidlyouttheinjectedarea,
thevolumeinvadedbythefluidsis toolarge toallowa pressure higher than theFOP away from the injection. Therefore, no failures and no seismicity could happen. Finally, at the experiment scale, the fault zone is very heterogeneous, such as the seismicity distribution. The faultcoreseemsindeedtoactasabarrierforthefluidand for the stress field. Itleads to heterogeneous hydraulic responsesandstressconcentrationandrotation(Faulkner et al., 2006). Those heterogeneities, together with the permeability differences, may explain the asymmetric distribution of seismicitybetween thetwo sides ofthe faultcoreintheTournemireshale(DeBarrosetal.,2016). Thestressfieldmayalsoimpacttheseismicbehavior. Theseismiceventsshowveryscatteredmechanismsfor both experiments. Particularly, for Tournemire test 2, 7mechanismswerecomputedforeventsthatsharesimilar horizontal location. Those events show similar nodal planes,butoppositepolarities.Thefourleft-handedstrike slipmechanisms(Fig.7a)agreewiththelocalstressfield measured byCornet(2000)andconfirmedbyGuglielmi et al. (2015b). The three remaining events show an opposite mechanism(right-handedmechanism,Fig.7b). As thefluid pressureequivalently reduces theeffective stressofallnormalcomponents,sucheventswithopposite slip-direction cannot be directly induced by a fluid-pressureincrease.Anadditionalforcing,likelytheaseismic deformation,shouldinterfereinordertostronglyperturb thelocalstressfield.
Astrongaseismicmotionwasobserved,when compar-ingdeformationattheinjection,inducedseismicity,and medium perturbations imaged throughseismic velocity changes.Thestrongdiscrepancybetweenseismicmoment and either deformation energy or McGarr (2014) rela-tionshipsclearlyshowsthatmostofthedeformationisnot emittingseismicity.Suchsmallseismic-to-aseismicratio
Fig.7.ExampleofmechanismscomputedfortwoseismiceventsinTournemire.Theboldlinesarethefaultplanes,whichareN150–180,30–608W structureswithcalcitefillings.Thelocalstressfield(S1=42MPa,S2=3.80.4MPaandS3=2.11MPa)wasmeasuredbyCornet(2000)andconfirmedby
Guglielmietal.(2015b).OutofthecomputedmechanismsintheTournemiretest2,foureventshaveamechanismsimilarastheleft-handedstrike-slipgivenin (a),andthreeeventsfollowaright-handedstrike-slipmechanism(b).
has been observed at the laboratory scale (Goodfellow etal.(2015),duringreservoirmonitoring(e.g.,Calo` etal., 2011;Schmittbuhletal.,2014)orinsubductionareas(e.g., Valle´eetal.,2013).
In our experiments, deformation is particularly ob-served aseismic at the injection points. Therefore, it is firstlyinducedbeforeanyseismicfailures(Guglielmietal., 2015a).Bothvelocitychangemeasurementsand numeri-cal modeling show that this aseismic motion may propagateover a distance oftens of metersaway from theinjectionpoints(Guglielmietal.,2015b;Rivetetal., 2016).Aroundit,thestressfieldshouldbeperturbed,by Coulomb failure stress change, in the same way as for earthquakes (Stein,1999). Moreover,this stresstransfer aroundmainshockstriggeredaftershocks,whosenumber decaywithtimefollowsanOmori’slaw.Here,stackingthe event distributions for all tests, wealso observed a 1/t decayofthenumberofseismicevents(Fig.8).Therefore, theobservedseismicityseemstobe‘‘aftershocks’’ofthe aseismic motion. In other words, themain slow failure
inducedattheinjectionmodifiesthestressfield,strongly enoughtotriggeropposite-slipseismicevents.
Wecanthereforeproposeanewmodeltoexplainhow fluidperturbationsinduceseismicity(Fig.9).Theincrease offluidpressureandtheassociatedreductionineffective normal stress induce large, aseismic failures. These aseismic deformations modifythefracture permeability and open new pathsfor the fluid flow. The volume to pressurizebecomes larger, withnewaseismic deforma-tionsoncethepressurelevelbecomeshighenough.The fluidpropagationcouldfollowafault-valve-likeprocessas proposed by Sibson (1990), except that failures do not generateseismicity.Itcouldalsobemodifiedbysudden shiftsfromhighlychanneledtolargepressurizedpatches. Inthatcase,theevolutionofpermeabilitywithstressand strain could be an important mechanism driving the growthofaseismicfaultrupture(Jeanneetal.,2018).The seismicityislikelyaconsequenceofthemainfailures:the stresstransferredfromtheaseismicdeformationmodifies the local stress field, which generates seismicity on
Fig.8.Stackedtimedistributionoftheseismiceventsforalltests.Thetimeisdefinedasthetimewhentheinjectionpressureisaboves3.Theredlineisthe
cumulativenumberofevents(dividedby5forthesakeofclarity).TheblacklineistheOmori’slaw(N(t)1/t),whichbestfitsthedistribution.
structureswithadequaterate-weakeningfrictional beha-viors.Itmeansthattheseismicitymaysurroundthemain deformation area, and can eventually be outside the pressurized area. Moreover, as the pressurized zone increases with injected volume, the seismic cloud size alsoincreaseswithtime,leadingtoanapparentdiffusivity oftheseismicity.
However,fromthesaturatedarea,fluiddiffusionmay also occur, and may directly induce seismicity by decreasingeffectivestress.Therefore,a dualprocess,by eitherstressperturbations orincrease in fluid pressure, mayleadtoseismicity.The balancebetween thosetwo processes mainly depends on the pressure level, the variations of the medium permeability with stress and strain,andtheinjectiontime(Cornet,2016;Jeanneetal., 2018).Inourexperience,seismicityseemsmostly domi-nated by stress perturbation processes, but a lower-pressure fluid, injected during a longer time might generateaseismicfrontthatfollowsfluiddiffusion.
Aseismic deformation is therefore dominating the hydraulicresponsesofreservoirs(i.e.thefluid-flowpaths). Thus,itis ofcrucialimportancetomonitor deformation during reservoir stimulation, by using either borehole instruments(strainmeters,extensometers,etc.),orsurface measurementswhenpossible.Monitoringvelocity chan-gesprovetobeapromisingtooltoinferin-situaseismic deformationandstresschanges(Calo` etal.,2011;Hillers etal.,2015;Rivetetal.,2016).Onthecontrary,theuseof inducedseismicitymaynotbeadirectprobeforthefluid, asitdepends,amongotherparameters,on theinjection properties,onthelocalstressfieldandonthedensity,the frictionalandstrengthproperties,andthepermeabilityof the fractures. Therefore, great care should be taken in interpreting fluid extension from the spatio-temporal distributionofseismicity.
5. Conclusion
Twoexperimentsoffaultreactivationbyfluidpressure injectionswereperformedinshale(Tournemiresite)and limestone(Rustrel site).Themulti-parameteranddense monitoringnetworkallowedthecloseobservationsofthe hydro-mechanical and seismological responses of the injected structures, from the static deformation to the high-frequency(10-kHz)acousticemissions.Whileplastic deformation wasobserved after most of the injections, onlyasparseseismicitywasobserved.Itsuneven spatio-temporaldistribution,theenergybudgetandtheseismic velocity changes show that more than 95% of the deformationisindeedaseismic. Seismicityratherseems tobearesponsetothestresstransferredfromthislarge aseismicdeformation,thantothefluidpressure.Therefore, adualprocessofstresstransferfromaseismicdeformation andporo-elasticdiffusionofpressureleadstoacomplex seismicbehavior.Moreover,theseismicemissionsdepend, notonlyontheinjectedvolume,butalsoonthelocalstress field, on the geologicalheterogeneities, on the fracture permeabilityandfrictionalproperties,andontheirstress– straindependency.Therefore,thespatio-temporal distri-butionoftheseismicityin,e.g.,reservoirsmaymainlymap the aseismic deformation, together with lithological
heterogeneities. Therefore,inordertoimprovereservoir monitoringandseismichazardsmitigation,itisofcrucial importancetobettermonitor andunderstandthe defor-mation that are not recorded by conventional seismic instruments.
Acknowledgments
Tournemire experiment was funded by TOTAL SA through ‘‘Fluids and Faults’’ project (PIs: Claude Gout, RaymiCastillaandPierreHenry).Rustrelexperimentwas supported by the ‘‘Agence nationale de la recherche’’ (HYDROSEISproject,PI.F.Cappa,ANR-13-JS06-0004-01) andbyTotalSA(HPMS-Caproject,Albion,PI.G.Massonat). We thank theIRSN (French Instituteof Radioprotection andNuclearSafety)fortheirdedicatedhelpandaccessto theIRSNTournemireplatform,andthelow-noise under-ground laboratory (LSBB) of Rustrel for logistical help during the experiment. SITES Company (J. Durand, H. Caron,and Y.Zouhair)is acknowledgedfor installing andmaintainingtheSIMFIPprobe,sensors,andacquisition duringtheexperiment.WethanktheMagnitudeCompany (Sainte-Tulle,Franceoffice)fortheTournemire microseis-mic data processing.Louis DeBarros deeplythanksthe French Academie des Sciences for awarding him the Gouilloud–Schlumbergerprize.
This paper is invited in the frame of Acade´mie des Sciences2017Prizes(GrandprixGouilloud-Schlumberger 2017).Ithasbeenreviewed/approvedbyMichelCampillo andVincentCourtillot.
References
Albano,M.,Barba,S.,Tarabusi,G.,Saroli,M.,Stramondo,S.,2017. Dis-criminating between natural and anthropogenic earthquakes: insightsfromtheEmiliaRomagna(Italy)2012seismicsequence. ScientificReports7(1),282.
Atkinson,G.M.,Eaton,D.W.,Ghofrani,H.,Walker,D.,Cheadle,B.,Schultz, R.,Liu,Y.,2016.HydraulicfracturingandseismicityintheWestern CanadaSedimentaryBasin.Seismol.Res.Lett.87(3),631–647.
Bao,X.,Eaton,D.W.,2016.Faultactivationbyhydraulicfracturingin westernCanada.Scienceaag2583.
Bardainne,T.,Dubos-Salle´e,N.,Se´ne´chal,G.,Gaillot,P.,Perroud,H.,2008.
AnalysisoftheinducedseismicityoftheLacqgasfield(Southwestern France)andmodelofdeformation.Geophys.J.Int.172(3),1151– 1162.
Brenguier,F.,Campillo,M.,Hadziioannou,C.,Shapiro,N.M.,Nadeau,R.M., Larose,E.,2008.PostseismicrelaxationalongtheSanAndreasfaultat Parkfieldfromcontinuousseismologicalobservations.Science321 (5895),1478–1481.
Calo`,M.,Dorbath,C.,Cornet,F.H.,Cuenot,N.,2011.Large-scaleaseismic motionidentifiedthrough4-DP-wavetomography.Geophys.J.Int. 186(3),1295–1314.
Cesca,S.,Grigoli,F.,Heimann,S.,Gonza´lez,A.,Buforn,E.,Maghsoudi,S., Dahm, T., 2014. The 2013September–October seismicsequence offshore Spain:a caseof seismicity triggered by gas injection? Geophys.J.Int.198(2),941–953.
Cornet,F.H.,2000.De´terminationduchampdecontrainteauvoisinagedu laboratoiresouterraindeTournemire,Rapp.N898N33/0073.Rapport dulaboratoiredeme´caniquedesroches,Institutdephysiqueduglobe deParis.
Cornet,F.H.,2016. Seismicandaseismic motionsgenerated byfluid injections.Geomechanics forEnergyandtheEnvironment,5,pp. 42–54.
DeBarros,L.,Bean,C.J.,Zecevic,M.,Brenguier,F.,Peltier,A.,2013.Eruptive fracturelocationforecastsfromhigh-frequencyeventsonPitondela FournaiseVolcano.Geophys.Res.Lett.40(17),4599–4603.
De Barros,L.,Daniel,G.,Guglielmi,Y.,Rivet,D.,Caron,H.,Payre,X., Barbieri,E.,2016.Faultstructure,stress,orpressurecontrolofthe
seismicityinshale?Insightsfromacontrolledexperimentof fluid-inducedfaultreactivation.J.Geophys.Res.:SolidEarth121(6),4506– 4522.
Derode,B.,Guglielmi,Y.,DeBarros,L.,Cappa,F.,2015.Seismicresponses tofluidpressureperturbationsinaslippingfault.Geophys.Res.Lett. 42(9),3197–3203.
Duboeuf,L.,DeBarros,L.,Cappa,F.,Guglielmi,Y.,Deschamps,A.,Seguy,S., 2017.Aseismicmotionsdriveasparseseismicityduringfluid injec-tionsintoafracturedzoneinacarbonatereservoir.J.Geophys.Res: SolidEarth122(10),8285–8304.
Faulkner,D.R.,Mitchell,T.M.,Healy,D.,Heap,M.J.,2006.Sliponweak faultsbytherotationofregionalstressinthefracturedamagezone. Nature444(7121),922–925.
Gibbons,S.J.,Ringdal,F.,2006.Thedetectionoflowmagnitudeseismic eventsusingarray-basedwaveformcorrelation.Geophys.J.Int.165 (1),149–166.
Goodfellow,S.D.,Nasseri,M.H.B.,Maxwell,S.C.,Young,R.P.,2015. Hy-draulicfractureenergybudget:Insightsfromthelaboratory. Geo-phys.Res.Lett.42(9),3179–3187.
Guglielmi,Y.,Cappa,F.,Lanc¸on,H.,Janowczyk,J.B.,Rutqvist,J.,Tsang,C.F., Wang,J.S.Y.,2013.ISRMsuggestedmethodforstep-rateinjection methodforfracturein-situproperties(SIMFIP):Usinga 3-compo-nentsboreholedeformationsensor.In:TheISRMSuggestedMethods forRockCharacterization,TestingandMonitoring:2007–2014.179– 186.
Guglielmi, Y.,Cappa, F., Avouac,J.P.,Henry, P., Elsworth, D.,2015a.
Seismicitytriggeredbyfluidinjection–inducedaseismicslip.Science 348(6240),1224–1226.
Guglielmi,Y.,Elsworth,D.,Cappa,F.,Henry,P.,Gout,C.,Dick,P.,Durand,J., 2015b.Insituobservationsonthecouplingbetweenhydraulic dif-fusivity and displacementsduring fault reactivation inshales.J. Geophys.Res.:SolidEarth120(11),7729–7748.
Gupta,H.K.,Arora,K.,Rao,N.P.,Roy,S.,Tiwari,V.M.,Patro,P.K.,Srihari,M., 2017.Investigations ofcontinuedreservoirtriggered seismicityat Koyna,India.GeologicalSociety,London.Spec.Publ.445(1),151–188.
Hillers,G.,Ben-Zion,Y.,Campillo,M.,Zigone,D.,2015.Seasonalvariations ofseismic velocitiesinthe SanJacinto faultareaobserved with ambientseismicnoise.Geophys.J.Int.202(2),920–932.
Jalali,M.R.,Gischig,V.,Doetsch,J.,Krietsch,H.,Amann,F.,Klepikova,M., 2017.MechanicalHydraulicandSeismologicalBehaviorofCrystalline RockasaResponsetoHydraulicFracturingattheGrimselTestSite.In: 51stUSRockMechanics/GeomechanicsSymposium.AmericanRock MechanicsAssociation.
Jeanne,P.,Guglielmi,Y.,Cappa,F.,2012.Multiscaleseismicsignatureofa smallfaultzoneinacarbonatereservoir:RelationshipsbetweenVP imaging,faultzonearchitectureandcohesion.Tectonophysics554, 185–201.
Jeanne,P.,Guglielmi,Y.,Rutqvist,J.,Nussbaum,C.,Birkholzer,J.,2018. Permeability Variations Associated With Fault Reactivation in a ClaystoneFormationInvestigatedbyFieldExperimentsand Numeri-calSimulations.J.Geophys.Res.:SolidEarth123,http://dx.doi.org/ 10.1002/2017JB015149.
Keranen,K.M.,Weingarten,M.,Abers,G.A.,Bekins,B.A.,Ge,S.,2014.Sharp increase in central Oklahoma seismicity since 2008 induced by massivewastewaterinjection.Science345(6195),448–451.
Kumar,A.,Zorn,E.,Hammack,R.,Harbert,W.,2017.Long-period, long-durationseismicityobservedduringhydraulicfracturingofthe Marcel-lusShaleinGreeneCounty,Pennsylvania.LeadingEdge36(7),580–587.
Lengline´,O.,Boubacar,M.,Schmittbuhl,J.,2017.Seismicityrelatedtothe hydraulicstimulationofGRT1,Rittershoffen,France.Geophys.J.Int. 208(3),1704–1715.
Madariaga,R., 1976.Dynamics ofan expanding circular fault.Bull. Seismol.Soc.America66(3),639–666.
McGarr,A.,2014.Maximummagnitudeearthquakesinducedbyfluid injection.J.Geophys.Res.:SolidEarth119(2),1008–1019.
McNutt,S.R.,2005.Volcanicseismology.Annu.Rev.EarthPlanet.Sci.32, 461–491.
Mukuhira,Y.,Asanuma,H.,Niitsuma,H.,Ha¨ring,M.O.,2013. Characte-risticsoflarge-magnitudemicroseismiceventsrecordedduringand afterstimulationofageothermal reservoiratBasel, Switzerland. Geothermics45,1–17.
Payre,X.,Maisons,C.,Marble´,A.,Thibeau,S.,2014.Analysisofthepassive seismicmonitoringperformanceattheRousseCO2storage demon-strationpilot.EnergyProcedia63,4339–4357.
Rivet,D.,Campillo,M.,Shapiro,N.M.,Cruz-Atienza,V.,Radiguet,M.,Cotte, N., Kostoglodov,V., 2011. Seismicevidence of nonlinearcrustal deformationduring alarge slowslip event inMexico. Geophys. Res.Lett.38(8.).
Rivet,D.,DeBarros,L.,Guglielmi,Y.,Cappa,F.,Castilla,R.,Henry,P.,2016.
Seismicvelocitychangesassociatedwithaseismicdeformationsofa faultstimulatedbyfluidinjection.Geophys.Res.Lett.43(18),9563– 9572.
Schmittbuhl,J.,Lengline´,O.,Cornet,F., Cuenot,N.,Genter, A.,2014.
InducedseismicityinEGSreservoir:thecreeproute.Geothermal Energy2(1),14.
Schoenball,M.,Ellsworth,W.L.,2017.ASystematicAssessmentofthe SpatiotemporalEvolutionofFaultActivationThroughInduced Seis-micityinOklahomaandSouthernKansas.J.Geophys.Res.:Solid Earth.
Shapiro,S.A.,2015.Fluid-inducedseismicity.CambridgeUniversityPress.
Sibson,R.H.,1990.Conditionsforfault-valvebehaviour.Geological Soci-etyLondon.Spec.Publ.54(1),15–28.
Stein,R.S.,1999.Theroleofstresstransferinearthquakeoccurrence. Nature402(6762),605.
Tary,J.B.,Baan,M.,Eaton,D.W.,2014.Interpretationofresonance fre-quenciesrecordedduringhydraulicfracturingtreatments.J.Geophys. Res.:SolidEarth119(2),1295–1315.
Valle´e,M.,etal.,2013.Intenseinterfaceseismicitytriggeredbyashallow slowslipeventintheCentralEcuadorsubductionzone.J.Geophys. Res.:SolidEarth118,2965–2981.
Wei,S.,Avouac,J.P.,Hudnut,K.W.,Donnellan,A.,Parker,J.W.,Graves, R.W.,Eneva,M.,2015.The2012Brawleyswarmtriggeredby injec-tion-inducedaseismicslip.EarthPlanet.Sci.Lett.422,115–125.
Yeck,W.L.,Hayes,G.P.,McNamara,D.E.,Rubinstein,J.L.,Barnhart,W.D., Earle,P.S.,Benz,H.M.,2017.Oklahomaexperienceslargest earth-quakeduringongoingregionalwastewaterinjectionhazard mitiga-tionefforts.Geophys.Res.Lett.44(2),711–717.
Zang,A.,Stephansson,O.,Stenberg,L.,Plenkers,K.,Specht,S.,Milkereit,C., Dahm,T.,2016.Hydraulicfracturemonitoringinhardrockat410m depthwithanadvancedfluid-injectionprotocolandextensivesensor array.Geophys.J.Int.208(2),790–813.
Zecevic,M.,Daniel,G.,Jurick,D.,2016.Onthenatureoflong-period long-durationseismiceventsdetectedduringhydraulicfracturing. Geophysics81,KS109–KS117.
Zoback,M.D.,2010.ReservoirGeomechanics.CambridgeUniversityPress, Cambridge,UK,459p.
