Seismicity at the Castor gas reservoir driven by pore pressure diffusion and asperities loading

ARTICLE Seismici y a he Cas o gas ese oi d i en by po e p essu e di usion and aspe i ies loading Simone Cesca 1✉, Daniel S ich2,3, F ancesco G igoli 4,5, Alessand o Vuan 6, José Ángel López-Comino 2,3,7, Pe e Niemz 1,7, Es e anía Blanch 8,9, To s en Dahm 1& William L. Ellswo h 10 The 2013 seismic sequence a he Cas o injec ion pla o m o sho e Spain, including h ee ea hquakes o magni ude 4.1, occu ed du ing he ini ial filling o a planned Unde g ound Gas S o age acili y. The Cas o sequence is one o he mos impo an cases o induced seis- mici y in Eu ope and a a e example o seismici y induced by gas injec ion in o a deple ed oil field. He e we use ad anced seismological echniques applied o an enhanced wa e o m da ase , o esol e he geome y o he aul s, de elop a g ea ly enla ged seismici y ca alog and eco d de ails o he up u e kinema ics. The sequence occu ed by p og essi e aul ailu e and unlocking, wi h seismici y ini ially mig a ing away om he injec ion poin s, igge ed by po e p essu e di usion, and hen back again, b eaking la ge aspe i ies loaded o highe s ess and p oducing he la ges ea hquakes. Seismici y occu ed almos exclusi ely on a seconda y aul , loca ed below he ese oi , dipping opposi e om he ese oi bounding aul . h ps://doi.o g/10.1038/s41467-021-24949-1 OPEN 1GFZ Ge man Resea ch Cen e o Geosciences Po sdam, Po sdam, Ge many. 2Ins i u o Andaluz de Geo ísica, Uni e sidad de G anada, G anada, Spain. 3Depa amen o de Física Teó ica y del Cosmos, Uni e sidad de G anada, G anada, Spain. 4Depa men o Ea h Sciences, Uni e si y o Pisa, Pisa, I aly. 5Swiss Seismological Se ice, ETH Zu ich, Zu ich, Swi ze land. 6Na ional Ins i u e o Oceanog aphy and Applied Geophysics - OGS, T ies e, I aly. 7Ins i u e o Geosciences, Uni e si y o Po sdam, Po sdam-Golm, Ge many. 8Depa amen de Física-EPSEB, UPC Ba celona Tech, Ba celona, Spain. 9Obse a o i de l’Eb e (OE), CSIC—Uni e si a Ramon Llull, Roque es, Spain. 10 Depa men o Geophysics, S an o d Uni e si y, S an o d, CA, USA. ✉email: simone.cesca@g z-po sdam.de NATURE COMMUNICATIONS | (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1|www.na u e.com/na u ecommunica ions 1 1234567890():,; An h opogenic seismici y is s imula ed by s ess pe u ba- ions, ac u ing p ocesses, o po e p essu e changes in he subsu ace accompanying fluid and mass mo emen s d i en by indus ial ac i i ies1,2. Since hese human ac ions only a ec he uppe mos se e al kilome e s o c us , induced ea h- quakes a e expec ed o occu a simila dep hs. As a ma e o ac , almos all majo cases o injec ion- ela ed induced seismici y wi h an accu a e hypocen al loca ion occu ed a ~4–8 km dep h (Supplemen a y Table 1). Due o i s shallow dep h, induced seismici y poses an impo an seismic haza d, as e en mode a e magni ude ea hquakes can p oduce s ong shaking and localized damages3–6. Fo hese easons, an h opogenic seismici y has become a ma e o g ea socie al conce n and ep esen s a new ocus o seismic haza d analysis1,7–9. While induced seismici y accompanied he exploi a ion o di e en na u al esou ces o e he pas cen u y7,10–14, so-called injec ion-induced seismici y (IIS) has gained scien ific a en ion in he pas ew decades1,15,16. The in e es in IIS has been p omp ed by he ecen occu ence o a numbe o impo an ea hquakes, which ha e been associa ed wi h fluid injec ion in he ame o hyd aulic ac u ing17–19, was ewa e disposal4,20, con en ional hyd oca bon ex ac ion7,14, and geo he mal exploi a ion ope a ions5,6,21. Unde g ound gas s o age (UGS) acili ies pe mi s o age and ex ac ion o la ge olumes o na u al gas o manage fluc ua ions in demand o p o ide a s a egic ese e. They a e ope a ed ou inely a hund eds o si es globally, ypically using deple ed oil/gas ese oi s, aqui e s, o ca e ns in ock/sal o ma ions. This ype o injec ion/ex ac ion ope a ions ypically has a weak seismic impac , and only a ew cases o seismici y po en ially ela ed o gas s o age ha e been epo ed. Gas s o age-induced seismici y may ha e occu ed a he Gazli gas s o age, Uzbekis an, a e he p e ious deple ion o a gas field, which had al eady expe ienced la ge damaging ea hquakes in 1976 and 198422. When he ese oi was used o gas s o age, cycling injec ion/ ex ac ion ope a ions we e accompanied by ea hquakes up o M 523. Ano he case o gas s o age-induced seismici y has been p oposed o he Hu ubi UGS, Xinjiang, China24, whe e he la - ges ea hquake eached Mw 3.0 and was loca ed a ~4 km dep h, sligh ly deepe han he ese oi o ma ion. Mo e ecen cases o induced mic oseismici y in Eu ope, epo ed a di e en gas s o age acili ies2,25, did no exceed Mw 1.0. The gene al lack o induced seismici y obse a ions a UGSs may be pa ially a ibu ed o poo seismic moni o ing24, bu e en well-moni o ed gas ese oi s, such as he Collal o ese oi , NE I aly, showed an almos comple e absence o induced seismici y, wi h no mic o- seismici y abo e M 0.0 wi hin 3 km om he ese oi 26. Ano he eason o he low seismogenic po en ial o gas s o age ope a ions could be ha gas injec ion schedules a e ypically designed and enginee ed no o exceed he s ess condi ions exis ing p io o o du ing he o iginal ese oi p oduc ion. To da e, he mos impo an case o seismici y induced by gas s o age in Eu ope is he Cas o p ojec , o sho e Spain, which has been conside ed o ha e igge ed a seismic sequence o >1,000 ea hquakes, peaking wi h h ee M 4+ea hquakes in Sep embe –Oc obe 201327,28. A he Cas o p ojec , cushion gas was injec ed om a pla - o m, loca ed ~22 km o sho e he coas o Spain, in o a o me ly deple ed hyd oca bon ese oi (Fig. 1). The ese oi i sel ex ends along he NNE–SSW di ec ion and deepens eas wa d. The geology is cha ac e ized by a ka s ed, ac u ed limes one. The ese oi is sealed on he NW by he Ampos a aul , a no mal aul ac ing as he ese oi oo . The o he sides o he ese oi a e sealed by an aqui e 29. When gas is injec ed, i emains apped a he op o he ese oi , and as he injec ion con inues, his will displace he gas–oil–wa e in e aces downwa ds. A se ies o injec ion es s we e pe o med a Cas o s a ing in 201329. Seismici y was fi s obse ed close o he Cas o p ojec pla o m on Sep embe 5, 2013, 3 days a e he s a o a la ge- scale gas injec ion in he deple ed Ampos a oil ese oi 27,29. The close p oximi y o he hypocen e s o he injec ion wells, he empo al co ela ion be ween injec ion ope a ions and seismic un es , and he low na u al backg ound seismici y in he epi- cen al egion we e a s ong a gumen o ea ly e idence o igge ed o induced seismici y27–32. The seismici y inc eased in a e and maximum magni ude, eaching a magni ude MbLg 3.0 on Sep embe 13, 2010 (Ins i u o Geog afico Nacional, IGN, ca alog). Injec ion ope a ions we e s opped on Sep embe 17, 201329. Ne e heless, as obse ed o o he induced seismici y cases4,5,33, seismici y con inued in he ollowing days. La ge ea hquakes occu ed, peaking in h ee ea hquakes wi h mag- ni ude Mw >4 in ea ly Oc obe 27, be o e finally ading away. The Cas o p ojec was pe manen ly closed in 2019. Seismici y pa e ns and hei ela ionship o gas injec ion ope a ions ha e been discussed in many ea ly publica ions and open epo s on he Cas o p ojec , assessing epicen al loca ions27–29, hypocen al dep hs27–29,31,32, ocal mechanisms and momen enso s27,29,31, s a is ical seismici y pa ame e s27,30, he local eloci y s uc u e28, and po en ial mechanisms o aul eac i a ion27,31,32,34. Un o una ely, he g owing numbe o scien ific publica ions has no been accompanied by a clea e unde s anding o he seismogenic p ocesses a Cas o , as some seismological esul s emain deba ed, and ul ima ely he e is no common ag eemen on which aul (s) was ac i a ed. The fi s majo ques ion conce ns he o e all spa ial dis ibu ion o he epicen al loca ions, po en ially p o iding in o ma ion on he o ien a ion o ac i a ed aul (s). In one s udy, he o ien a ion was ound o be oughly pa allel o he coas 27 and pe pendicula o i in ano he 28. As o he hypocen al dep h, he e is some ough ag eemen on he iden ifica ion o shallow c us al sou ces bu wi h a b oad ange o dep h es ima es, mos ly a ying be ween 1–527 and 6–8km 28,32 bu including e en deepe es ima es28,29. Such la ge unce ain y p e en s de elopmen o eliable models o aul ac i a ion. Finally, he e is consensus on he ocal mechanism, showing s ike-slip mechanisms wi h a s eep NW–SE plane and a SE dipping NE–SW plane, wi h some a iabili y on he dip angles27,29,31. Un o una ely, each o he possible wo plane o ien a ions fi s one o he p oposed seismici y dis ibu- ions, so ha he aul geome y emains un esol ed. B inging oge he p e ious esul s is challenging, gi en he a ie y o da a, eloci y models, and seismological me hods and he lack, in mos p e ious s udies, o eliable unce ain y es ima ions. He e we ecollec a b oad da ase om 31 seismic s a ions (Supplemen a y Fig. 1), including one ocean bo om seismome e (OBS), subs an ially inc easing he da ase and/o he azimu hal co e age, in compa ison o p e ious wo ks, ei he using <10 s a ions27,29 o <20 wi h a s ongly asymme ic ne wo k geome y31. We model seismic da a by a combina ion o mode n seismological echniques, including a p obabilis ic momen enso in e sion, empi ical G een’s Func ion, a - egional and ele- seismic a ay analysis, and empla e ma ching, which a e used o he fi s ime o analyze his seismic sequence. While he assess- men o pa ame e unce ain ies allows explaining p e ious pa - ially con adic o y esul s, ou new esul s significan ly imp o e he loca ion accu acy and sou ce pa ame e esolu ion, which allows us o econs uc he complex, mul i-s age e olu ion o seismici y and p opose a sel -consis en up u e scena io o he seismic un es . Resul s We adop a ange o mode n echniques (see de ails in he “Me hods”sec ion), which p o ide subs an ial new esul s. Templa e ma ching is used o enhance he ca alog size, ela i e ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 2NATURE COMMUNICATIONS| (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1 | www.na u e.com/na u ecommunica ions loca ion me hods o es ablish he con o e sial dis ibu ion o epicen e s, a ay echniques o in e accu a e dep h es ima es based on independen da a, momen enso and up u e di ec- i i y analysis o cons ain he up u e plane o ien a ions, and finally an accu a e assessmen o wa e o m simila i y o confi m he o e all simila i y in loca ion and ocal mechanism. Resul s s abili y is assessed using he local eloci y model G28, es ing al e na i e models (Supplemen a y Fig. 2), and p o iding sou ce pa ame e s unce ain ies, which help o explain some incon- sis encies among p e ious s udies. The new esul s esol e de ails o he e olu ion o seismici y wi h ongoing injec ion and shed ligh on he physical causes o induced seismici y and i s magni ude. Templa e ma ching de ec ion. We ep ocess con inuous eco dings o enhance he seismic ca alog wi h he aim o be e acking he seismici y e olu ion in space and ime. Using wa e o m empla e ma ching (examples o h ee de ec ions a di e en magni ude le els a e shown in Supplemen a y Figs. 3–5), we a e able o conside ably augmen he de ec ed ea hquakes om 536 (IGN ca alog) and 982 (Eb o ca alog) o 3437 ea h- quakes (Supplemen a y Fig. 6 and Supplemen a y Da ase 1). The augmen ed ca alog includes he 148 empla es used. The mini- mum magni ude in ou ex ended ca alog is Mw −0.2. The magni ude o comple eness imp o ed om Mc ~2.0 (IGN)35 and Mc 1.3 (Eb o)27 o Mc 0.6 and 0.3 in he injec ion and pos - injec ion pe iods, espec i ely (Supplemen a y Fig. 7), using magni udes o he Eb o ca alog as e e ence. The analysis o he ex ended ca alogs confi ms a subs an ial change in he seismici y occu ed on Sep embe 17, 2013, when injec ion ope a ions we e in e up ed (Supplemen a y Fig. 7). The b- alue d opped om 0.99 ± 0.04 o 0.77 ± 0.01 be ween he injec ion and pos -injec ion seismici y phases, suppo ing p e ious findings based on ca alog alone27. Ea hquake eloca ion. We use wo independen , ad anced ela i e ea hquake localiza ion me hods, based on wa e o m c oss-co ela ion and S - P di e en ial imes combined wi h dis- ance geome y echniques. Bo h me hods ake ad an age o he P and S wa e o ms and ha e a highe esolu ion han he con- en ional, absolu e localiza ion me hods used in p e ious s udies o he Cas o sequence27–29. These echniques a e especially sui ed o he analysis o spa ially clus e ed seismici y, as obse ed a Cas o , and less sensi i e han absolu e loca ion echniques o s uc u al he e ogenei ies ou side he seismogenic olume and o ne wo k asymme y. Wa e o m-based ela i e hypocen e loca- ions ob ained o a subse o 51 e en s wi h magni udes >2.0 (Supplemen a y Da ase 2) demons a e ha he sequence was es ic ed o a single, ~4 km long, N42°E ending lineamen (Fig. 2a), e en a e accoun ing o unce ain ies (Supplemen a y Fig. 8). P ecise loca ions esul om c oss-co ela ion ela i e iming, and achie ed an a e age oo mean squa e a el ime e o s o 0.02 s. Hypocen e s concen a e wi hin an ~1 km wide dep h in e al (Fig. 2c, d), al hough ela i e dep hs a e di ficul o esol e due o he asymme ic ne wo k geome y (Supplemen a y Fig. 1). As discussed below, he absolu e dep h o seismici y is less well esol ed and likely a a dep h o ~3 km. Loca ions de eloped using an independen and al e na i e app oach36 confi m he o ien a ion o he seismici y. The me hod elies on he solu ion o a seismological dis ance geome y Fig. 1 O e iew map o he s udy egion o sho e Spain. The map shows he loca ion o he Cas o pla o m (whi e squa e), closes seismic s a ions ALCN and ALCX (whi e iangles), and he app oxima e ace o he Ampos a aul ( ed do ed lines)47,48. The uppe igh inse shows he geog aphical loca ion o he s udy a ea, he lowe igh inse a pho o o he Cas o pla o m (cou esy Ál a o González). The map has been plo ed wi h GeoMapApp (www.geomapapp.o g), using he Global Mul i-Resolu ion Topog aphy (GMRT)70 and egional ba hyme y a e he EMODNET 500 m compila ion. NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1| www.na u e.com/na u ecommunica ions 3 p oblem36 using di e en ial S-P a el imes a he closes seismic s a ions, ALCN and ALCX. These wo s a ions, he closes o he Cas o pla o m, a e op imally si ua ed o his me hod, ha ing almos pe pendicula azimu hs wi h espec o he seismici y (Fig. 2b and Supplemen a y Fig. 9). While mo e sca e ed, loca ions we e ob ained o a la ge da ase o 408 ea hquakes (Supplemen a y Da ase 3). The ea hquakes loca ed wi h he wa e o m co ela ion app oach ha e a la ge ange o S-P di e en ial imes a ALCX (SW), han a ALCN (NW), which confi ms he NE–SW end o he seismici y (Supplemen a y Fig. 9). Loca ion esul s a e s able and no subs an ially biased by he assumed eloci y model (Supplemen a y Fig. 10). Bo h ca alogs a e included in he Supplemen a y Ma e ial. Momen enso in e sion. Momen enso solu ions o 11 ea hquakes wi h Mw >3.0 (Supplemen a y Table 2) a e based on a p obabilis ic in e sion app oach37 and conside s di e en one- dimensional (1D) eloci y models and wa efield a ibu es (Supplemen a y Figs. 11–13). The esul s a e s able and inde- penden o he eloci y models used. All s udied e en s display a simila , p edominan ly s ike-slip aul ing s yle (Fig. 3). The le - la e al nodal plane ends NE–SW, oughly pa allel o he coas , and dips owa d SE (s ike 42° ± 3°, dip 48° ± 11°, ake −1° ± 8°, es ima ed om he dis ibu ion o bes quali y A and B solu ions o 9 ea hquakes assuming model G, Supplemen a y Table 2). The igh -la e al nodal plane, ending NW–SE (s ike 313° ± 4°), is sub- e ical. The enhanced ca alog o momen enso s esem- bles he gene al ocal mechanism p oposed in p e ious wo ks27,28,31. Cen oid loca ions depend on he chosen model, wi h he a e age loca ions closes o he injec ion pla o m o model G. Cen oid dep hs a e age 3.4 ± 1.6 km. The la ges magni udes o Mw 4.09 ± 0.04 a e ound o model G. Simila esul s a e ob ained o o he eloci y models, wi h a ew cases ha ing sligh ly g ea e dep hs and magni udes (Supplemen a y Table 2). Focal dep h es ima ion om dep h phases a egional and eleseismic dis ances. We conduc ed an independen analysis o he ocal dep h o he h ee la ges e en s using dep h phases obse ed a egional and eleseismic dis ances on di e en a ays, using he abede o ool (h ps://gi hub.com/He Muelle luedenscheid/abede o). G een’s unc ions compu ed using local eloci y models a bo h he sou ce and a ay loca ions (de ails in he Supplemen a y No e 8) o m syn he ic beams ha also accoun o he momen enso s es ima ed in his s udy. Measu ed ime delays be ween di ec P and he nea -su ace eflec ed pP phase a e o he o de o 1.5–1.8 s, as es ima ed a he GERES a ay38, Ge many (~1400 km dis ance). We ob ained simila esul s using he ILAR, US a ay a eleseismic Fig. 2 Loca ions based on wa e o m co ela ions and S - P me hod. a,c,dLoca ions based on wa e o m co ela ions in map iew and c oss-sec ion. bLoca ions based on S - P me hod in c oss-sec ion. Ea hquake loca ions (ci cle, colo s deno ing o igin ime in a,c,d, and S-P di e en ial ime a s a ion ALCX in b, acco ding o colo ba s in aand b, espec i ely) a e plo ed in map iew (a,b) and along AB (c) and CD (d) c oss-sec ions (p ofiles AB and CD a e shown in a). Black ci cles in bdeno e hose e en s used in he wa e o m co ela ion loca ion (a), o which S-P ime es ima es a e a ailable. A pen agon deno es he loca ion o he Cas o pla o m, double g ay solid lines ske ch he injec ion wells29. ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 4NATURE COMMUNICATIONS| (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1 | www.na u e.com/na u ecommunica ions dis ances (~8000 km). These delays a e compa ed o he mo eou and wa e o m pa e ns syn hesized om heo e ical seismog ams a each a ay loca ion. Syn he ic wa e o ms and delaysa econsis en wi h cen oid dep hs o 3–4km(Fig.4and Supplemen a y Fig. 14). Rup u e di ec i i y. The up u e p ocess o he wo la ges ea hquakes o he sequence, he Oc obe 1, 2013, 03:32 Mw 4.1 (EQ1) and Oc obe 2, 2013, 23:06 Mw 4.1 (EQ2) ea hquakes, e ie ed by applying empi ical G een’s unc ion (EGF) echniques39, show simila appa en sou ce ime unc ions (ASTFs), wi h du a ions o 0.25–0.72 s (Fig. 5and Supplemen a y Fig. 15). A clea azimu hal pa e n o he ASTFs is iden ified o bo h e en s, om fi ing S wa e-based appa en du a ions (Fig. 5). The esul ing sou ce pa ame e s (Supplemen a y Table 3) show ha bo h ea hquakes sha e simila up u e cha ac e is ics. We es ima e up u e leng hs o ~1.1 km, in ag eemen wi h he size app op ia e o a c ack model (see “Discussion”), wi h a up u e du a ion o ~0.5 s. The in insic ade-o be ween ise ime and up u e eloci y40,41 leads o high unce ain ies; assuming a ise ime o 0.15–0.25 s, he es ima ed up u e eloci y would be ~2.7 km/s, simila o he shea wa e eloci y a 3 km dep h in model G. Rup u es om bo h ea hquakes a e ound o be asymme ic bila e al, wi h mos o he up u e (66 and 75% o he o e all up u e leng h o he wo ea hquakes, espec i ely) p opaga ing owa d NNE om he ini ial hypocen e . These esul s be e co ela e wi h he N42°E-o ien ed aul , dipping SW, which is esol ed independen ly om he band o seismici y and he momen enso solu ions, and hey a e inconsis en wi h a up u e along he al e na e plane o he momen enso , which is sub- e ical and s ikes o NW. Wa e o m-based classifica ion. We employ ne wo k-based e en simila i y clus e ing42 o iden i y e en s wi h high wa e o m simila i y a mul iple s a ions, which implies high simila i y in hei loca ions and ocal mechanisms. The subse o 51 eloca ed e en s, wi h magni udes >2.0, shows e y simila wa e o ms ac oss he s a ion ne wo k. Su ace wa es a e highly simila bu only isible o la ge magni ude e en s, so we analyzed P and S Fig. 3 O e iew o he momen enso in e sion esul s. a O e lay o de ia o ic momen enso solu ions ob ained o he Oc obe 2, 2013, 23:06 UTC ea hquake (Supplemen a y Table 2) ou o he boo s ap app oach (g ay ocal sphe es) and bes solu ion using all da a ( ed hick line ocal sphe e). bP obabili y densi y unc ions o momen magni ude (Mw) and dep h o he same e en ( ed do ed and solid line indica e bes and mean solu ions, da k o ligh ed backg ound confidence le els o 68, 90, and 95%). cExamples o wa e o m fi ( ed and black lines deno e syn he ic and obse ed displacemen wa e o ms, espec i ely) a selec ed s a ions o he same e en (s a ion name, azimu h, dis ance, weigh , and maximal displacemen a e epo ed). dO e lay o momen enso solu ions o 9 s udied e en s (quali y A and B, Mw 3.4–4.1), assuming h ee di e en eloci y models: model I (blue), G (pu ple), and V ( ed ocal sphe e); hick black lines deno e he ocal sphe e o he cumula i e momen enso . NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1| www.na u e.com/na u ecommunica ions 5 wa e o ms (Fig. 6). The ne wo k-based clus e ing iden ifies six compac , spa ially sepa a ed clus e s (Fig. 6and Supplemen a y Fig. 16) wi h wa e o ms changing g adually along he lineamen . A supplemen a y analysis o P/S and Rayleigh/Lo e (R/L) ampli ude a ios e eals smoo h ends along he NE s iking lineamen (Supplemen a y Fig. 17), which can be modeled using syn he ic seismog ams. The esul s indica e ha he loca ion change is he eason o he obse ed wa e o m a ia ions among neighbo ing clus e s, no a ying ocal mechanisms. All o hese esul s imply ha he aul geome y emains s able o e he cou se o he sequence, wi h seismici y occu ing in neighbo ing clus e s a di e en imes, suppo ing he p og essi e ailu e o di e en pa ches along one common, plana s uc u e. Only clus e s a he NE and SW bounda ies o he cloud show some di e ences, which may be a ibu ed o he la ge sepa a ion o hese e en s, o possibly changes in aul geome y and local he e ogenei y a he e mina ions o he ac i e aul segmen . Discussion The cu en deba e abou seismological esul s a Cas o p inci- pally e ol es a ound he hypocen al loca ions, hei spa ial dis ibu ion, and he dep h ange o he seismici y. Absolu e loca ions a e indeed poo ly esol ed, as demons a ed by la ge changes in loca ion when using di e en eloci y models28. Absolu e loca ion unce ain ies o a single e en 28 show a simila pa e n, wi h poo esolu ion in he NW–SE di ec ion and a ade-o wi h he hypocen al dep h. Such loca ion unce ain ies a e a ibu ed o he s ongly une en ne wo k geome y, in combina ion wi h la e al s uc u al he e ogenei ies in he Valen- cia T ough43. Mul i-e en s ela i e loca ion me hods, based on di e en ial phase a i al imes om phase picks and/o wa e o m c oss-co ela ions, can imp o e he loca ion esolu ion and sha pen seismici y pa e ns44–46. Ou esul s, and specifically hose o he c oss-co ela ion-based eloca ion, esol e how seismici y is dis ibu ed along an elonga ed olume, ex ending sub-pa allel o he coas along a N42°E di ec ion (Fig. 2and Supplemen a y Fig. 8). The o ien a ion, shape, and ex en o he imaged ac i e egion ma ch he ese oi ’s ou line29,47. The hickness o he laye a ec ed by seismici y is confined o 2 km only, wi h good esolu ion. As ea hquakes did no occu abo e o below his laye , and epicen al unce ain ies a e well below 1 km, he dip o he aul ac i a ed ac oss he laye canno be esol ed om seismici y. Shallow hypocen al dep hs, wi hin he uppe c us , a e ound in o he s udies27,28,32, bu hei absolu e alues a e inconsis en . The delay imes be ween di ec and su ace eflec ed sub- e ical wa es is only 1.5–1.8 s. We model bo h he delay and he wa e o ms (model G) o pP e sus P phases o he h ee la ges ea hquakes o he Cas o sequence using indepen- den da a and can cons ain he absolu e dep hs a 3–4 km. The cen oid dep hs om ou p obabilis ic momen enso in e sions a e also consis en be ween 0 and 4.7 km (Supplemen a y Table 2). Deepe sou ces a 6–8 km ha e been ecen ly sugges ed, based on he modeling o sho pe iod c us al e e be a ion in a shallow, low eloci y laye 32. Such an app oach also has he po en ial o accu a e dep h es ima e; howe e , dep h Fig. 4 Summa y o seismic dep h es ima ions o he Oc obe 1, 2013, 03:32:44 UTC, ea hquake based on seismic a ay beam modeling. a–cCompa ison o he obse ed ( hick blue line) and syn he ic ( hin black lines) beams a he GERES a ays (Ge many). Syn he ics a e compu ed o h ee eloci y models (models G, I, and V in a–c, espec i ely), he p e e ed ocal mechanism de i ed using model G (s ike 41°, dip 56°, ake 4°), and dep hs be ween 1 and 7 km. Obse ed beams a e plo ed a he dep hs o 3–4 km, o which wa e o ms and P and pP pulses a e bes modeled. Bo h obse ed and syn he ic beams ep esen no malized displacemen s, bandpass fil e ed in he ange 0.25–1.80 Hz. dTheo e ical sou ce dep h as a unc ion o he pP delay o he h ee models a he sou ce (model I in blue, model G in pu ple, model V in ed): gi en he pP-P delay, and assumed a eloci y model, we can es ima e he dep h. We conside delays o 1.5–1.8 s, as picked om beams a GERES, Ge many and ILAR, US a ays, and an a e age o 1.65 s. Th ee scena ios a e shown (dashed lines), leading o a dep h o 2.4 ( ed line), 3.8 (pu ple line), and 5.4 (blue line) km. ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 6NATURE COMMUNICATIONS| (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1 | www.na u e.com/na u ecommunica ions unce ain ies we e no epo ed he e, and i is unclea o which ex en his ype o modeling is a ec ed by he adop ion o a simplified 1D model. Hypocen e s a such dep hs would imply a delay o he pP phase o 2.6–3.3 s, when assuming he local model G, which a e oo la ge compa ed o he delays o 1.5–1.8 s obse ed o he la ges e en s om egional and eleseismic a ay analysis. Thus, we conclude ha ea hquakes we e shallowe , mos likely in he ange 3.0–4.0 km bu no exceeding 5.3 km (Fig. 4d). Ou new Bayesian dep h es ima es imply ha ea h- quake oci would be only ~1–2 km deepe han he injec ion poin a ~2 km dep h a he op o he gas s o age laye . As o he momen enso s, ou esul s ag ee wi h p e ious findings27,31, which a e cha ac e ized by p edominan ly s ike-slip mechan- isms, wi h aul planes o ien ed N42°E and N47°W, espec i ely. We p o ide explici unce ain y es ima es o all esol ed sou ce pa ame e s, which accoun o bo h da a- and model- ela ed unce ain ies. This allows a obus join in e p e a ion o he esul s om di e en echniques, aking in o accoun he eso- lu ion o each me hod. Ou esul s ha e impo an implica ions o he iden ifica ion o he ac i a ed aul (s). P e iously, i has been hypo hesized ha he seismici y a Cas o migh ha e occu ed along mapped aul s, such as he la ge N25°E-o ien ed48,49 Ampos a aul 29 and he Mon sia aul sys em a he ese oi si e, which has mul iple sub- pa allel aul s o ien ed NW27. Al e na i ely, unknown aul s wi h he wo possible o ien a ions iden ified by he aul planes o he s ike-slip ocal mechanisms ha e been in oked: ei he a NE s iking aul , gen ly s eeping owa d SE27, o NW ending aul (s)28. Ou esul s, bo h conside ing he spa ial elonga ion o ela i e loca ions, simila i y o momen enso s, he high co e- la ion o wa e o ms and wa e o m a ibu es along he end, and he up u e di ec i i y o he la ges e en s, sugges he ac i a ion o a single NE–SW-o ien ed aul . Ou esul s also exclude he ac i a ion o he shallow Mon sia aul sys em o an unknown deepe aul wi h simila NW–SE o ien a ion28, below he Ampos a aul . Simul aneous ac i a ion o he whole Mon sia sys em, which could po en ially explain he spa ial dis ibu ion o seismici y27, can be ejec ed as incompa ible wi h he up u e (a.1) (b.1) (a.2) (b.2) Fig. 5 Rup u e di ec i i y in e ed o he wo la ges ea hquakes o he sequence. a Mw 4.1 Oc obe 1, 2013, 03:32 UTC and bMw 4.1 Oc obe 2, 2013, 23:06 UTC ea hquakes a he Cas o pla o m. a.1, b.1. The du a ion o appa en sou ce ime unc ions a e plo ed so ed by azimu h, showing appa en du a ions (ci cles) and momen a es (g ay a ea) a each s a ion (see labels), and he asymme ic bila e al up u e model om he di ec i i y analysis (dashed lines). a.2, b.2. Appa en du a ions as in a.1, b.1 (ci cles) along wi h he syn he ic p edic ions o he in e ed model (dashed lines) show he p edominan up u e di ec ion in a pola plo , whe e appa en sou ce du a ions a e plo ed along he adial axis (axis icks co espond o 0.2 s). The unce ain y in he di ec ion o he up u e di ec ion ( ange o angles deno ed wi h blue and ed a eas) a e es ima ed om he esiduals. NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1| www.na u e.com/na u ecommunica ions 7 di ec i i y esul . The spa ial dis ibu ion o epicen al loca ion esembles qui e well he o ien a ion o he Ampos a aul , which o ms he oo o he ese oi , and, as so, o he ese oi geome y. Howe e , he Ampos a aul dips o he NW, wi h a ~60° dip angle a shallow dep hs48,49, ~40° a he ese oi le el, and ~30° below i 29, which is inconsis en wi h p e iously p o- posed ocal mechanisms27,29,31, as well as wi h all ou bes quali y (A and B) momen enso solu ions. We conclude ha he Ampos a aul did no pa icipa e in he seismic sequence, con- fi ming p e ious esul s27,28. Only one up u e scena io emains consis en wi h all esul s: ac i a ion o a p e iously unde ec ed aul an i he ic o he Ampos a aul , wi h a simila s ike, bu dipping owa d he SE. This hypo hesis has been p e iously p oposed27 and can now be efined. The la e al dis ibu ion o he seismici y, which esembles he end o he Ampos a aul , and he shallow dep h o he hypocen e s, jus 1–2 km below he injec ion, poin s o a aul wi hin o a he below he ese oi , dipping SE, and bounded o he NW by he Ampos a aul ; such aul o ien a ion and mechanism ype a e consis en wi h he local s ess o ien a ion27. The spa ial associa ion o he hypocen e s and injec ion poin s and he clea empo al co ela ion among injec ion ope a ions and seismici y make a clea case o induced seismici y50, which indeed has ne e been deba ed. Joining he in o ma ion o ela i e loca ions and empla e ma ching, we can econs uc he spa io empo al e olu ion o seismici y (Fig. 7). No e ha in Fig. 7we cons ain absolu e loca ions in he way ha ea lies e en s ma ch he injec ion loca ion. We can iden i y h ee phases. The fi s begins on Sep- embe 2, oge he wi h he s a o he injec ion, and con inues un il he injec ion s ops. The second phase begins on Sep embe 17 and con inues un il he end o Sep embe and is associa ed wi h he e olu ion o po e p essu e and seismici y. These wo phases a e cha ac e ized by di e en b- alues, which d ops om ~1.0 (la ge p edominance o small ea hquakes) o 0.8 (highe a e o la ge magni ude e en s). The hi d phase, om he end o Sep embe o ea ly Oc obe , includes all o he la ges e en s o he sequence. O e all, wa e o ms o ea hquakes occu ing in di e en phases a e e y simila (Fig. 6); mino wa e o m changes can be a ibu ed o sligh ly changing loca ions. Du ing bo h phases 1 and 2, seismici y mig a es unila e ally owa d he SW (Fig. 7c), a an a e age eloci y o ~180 m/day, consis en wi h a unila e al p essu e di usion model51. In he hypo hesis ha he di usion is confined o a linea s uc u e, he empo al mig a ion o he seismici y on can be econs uc ed (Fig. 7b) assuming a di usi i y o app oxima ely 0.5 m2/s51.I is no ewo hy ha he gas is injec ed close o he pla o m loca ion and o he SW o i h ough 4 wells spanning abou 600 m, whe e he ese oi oo is shallowes 29,47. The ho izon al dis ibu ion o epicen e s esembles he ese oi shape29,47. The empo al mig a ion o he seismici y can be explained as ollows (Fig. 8). The gas was injec ed in o he highes a ea o a pen oo -shaped ese oi , elonga ed NNE–SSW, bounded on he NW by he Ampos a Faul and becoming p og essi ely deepe owa d SSW and NNE47. I has been in e p e ed ha he no he n end o he Ampos a aul e mina es a ~2 km dep h in o Te ia y laye s, sugges ing ha his po ion o he aul is ec onically inac i e29. Due o he sligh inclina ion o he idge egion o he sealed ese oi in he SW and NE di ec ion, he gas sa u a ed olume would slowly ex end o dep h du ing phase 1, o occupy a na ow olume elonga ed NNE–SSW29,47. As he injec ion p oceeded, mode a e magni ude seismici y g adually mig a ed o he SW o he injec ion poin bu no owa d he NE (Fig. 7). The lack o seismici y o he NE sugges s di e ences in he pe meabili y o in aul (s) p ope ies, such as a aul bending o o se . The seismi- ci y mig a ion con inued du ing phase 2, ollowing he end o injec ion, p obably eflec ing ongoing po e p essu e edis ibu ion. The hi d phase began ab up ly by Sep embe 28–29, 2013, cha ac e ized by a as (~1 km/day) backwa d p opaga ion o he seismici y om he ex eme SW end owa d he injec ion poin , accompanied by all o he la ges e en s o he sequence, including he h ee M 4+ea hquakes (Fig. 7). Phase 3 does no ma k a subs an ial change in he b- alue (0.8). In de ail, he ini ial ac i i y appea ed a a ew spo s loca ed on he edges o o be ween he up u e a eas o he la e , la ge e en s (Fig. 8). Es ima ed up u e a eas o la ge e en s, excep he wo la ges ones, a e gene ally non-o e lapping. The assump ions he e a e a cons an s ess d op (3 MPa), which leads o up u e sizes consis en wi h leng hs up o 1.0–1.2 km in e ed om ASTFs, and ha hey a e on a single plane. Howe e , as mig a ion and simila wa e o ms and ocal mechanisms o ea hquakes sugges a single aul ailing p og essi ely, wi h homogeneous geome y, poin ing o a SE dipping, shallow (~3 km) aul plane. The NE unila e al di ec i i y esol ed o he la ges ea hquakes in ea ly Oc obe suppo s he spa ial e olu ion o he up u e p ocess om SW o NE. The las e en s o he sequence, loca ed a he NE ip o he ac i a ed aul , a e mos likely a e shocks o he la ges ea ly Oc obe e en s and a e loca ed in he o wa d di ec i i y di ec- ion. O e all, he seismici y e olu ion ma ks a kind o ound- ip, whe e a slow di usion-d i en mig a ion away om he injec ion wells is ollowed by a sudden e u n by he seismic igge ing o unb oken aspe i ies. Based on he space– ime pa e ns, se e al scena ios a e possible o explain he complex up u e mig a ion. One is simply po e p essu e di usion, as he ea lies ac i i y occu s a he No h, whe e he pla o m is loca ed, and close o he injec ion well bo om. Seismici y mig a es o he SW a a eloci y o abou 180 m/day. This is a ypical ange o po e p essu e di usion51 and may ep esen he mo emen o he p essu e on , causing c i- ically s essed a eas o up u e fi s , as he p essu e inc eased. A e he injec ion s ops, he SW mig a ion con inues as po e Fig. 6 Wa e o m simila i y o eloca ed e en s. P and S wa e o ms (he e shown o s a ion EMOS) a e e y simila , e en when fil e ed wi h a passband o 1–8 Hz. The signals change g adually along he SW–NE ending lineamen (equi alen o p ofile CD in Fig. 2a). By applying a wa e o m-based e en simila i y clus e ing algo i hm, we find six spa ially sepa a ed clus e s o e en s (colo -coded, clus e numbe ing based on o igin ime o he fi s clus e ed e en ); no e ha he clus e ing is no only based on he wa e o ms o he exempla y s a ion EMOS, shown he e, bu also on se e al o he s a ions. Wa e o ms a e scaled by he maximum ampli ude wi hin he ime window (bold) and aligned by he c oss- co ela ion ime shi o emphasize wa e o m simila i y. ARTICLE NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 8NATURE COMMUNICATIONS| (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1 | www.na u e.com/na u ecommunica ions p essu e g adien s a e s ill p esen o u he d i e he po e p essu e on in he ese oi . The e e se pa e n in phase 3 is di e en . The occu ence o he la ges e en s p og essi ely up u ing p e iously unb oken pa ches om SW o NE. The mig a ion eloci y is abou fi e imes as e , in he ange o 1 km/ day. Ou analysis indica es ha la ge aspe i ies successi ely up u e in la ge magni ude ea hquakes. The aspe i ies may ha e been loaded o mo e c i ical s ess du ing phases 1 and 2, whe e s ess in be ween neighbo ing aspe i ies was ans e ed. A pos- sible cause o he apid backwa d mig a ion o he ea hquakes could be he g adien o e ical s ess de eloped by he con- inuous exchange o wa e by gas in he uppe mos pa o he ese oi . Because o he inclina ion o he cap ock, which dips gen ly o he SW, he e ical gas column a he injec ion poin is highe han a he sou hwes e n end o he injec ed gas laye . No mal s ess is smalle a he injec ion poin han a he sou hwes e n ip. A up u e p og ession owa d highe Coulomb s ess change has been sugges ed as he inducing mechanism o he seismici y52. The sequen ial up u ing o aspe i ies could be addi ionally a o ed by a p og essi ely educed aul ic ion, as wa e displaced by he injec ed gas would flow away in o ac u es and pe meable pa hs, such as he ac i a ed aul . Faul al ing, in ol ing p essu e changes and uns eady fluid mig a ion along ac u ed egions and aul s53, can also pa ially explain he al e na ion o seismici y bu s s and quiescence pe iods. This could also imply he occu ence o aseismic slip, o which we ha e no di ec e idence so a . The clus e ing analysis shows ha weak ea hquakes in phases 1 and 2 ha e he same mechanism as he la ge ones in phase 3 (Fig. 6), and occu ed along he same SE dipping s uc u e. I aseismic slip on he Ampos a aul accompanied he ini ial phases o seismici y, as has been ecen ly sugges ed34, i did so comple ely aseismically. Induced seismici y is a complex issue o public in e es , whe e he exploi a ion o na u al esou ces needs o be pe o med in a sus ainable way, while also ca e ully assessing he isk. Wi hin he esea ch field o induced seismici y, scien ific ou comes a e di ec ly ele an o egula o s, indus y, and socie y. In his amewo k, scien ific esul s need o be explained wi h a igo ous, non- echnical and b oadly unde s andable language6. A scien ific discussion is desi able and in insic o science, bu he commu- nica ion o incomple e and con adic o y esul s may impede explana ion o scien ificfindings o a b oad communi y. A pa - icula p oblem is a lack o communica ion abou unce ain ies, bo h epis emic and alea o y, which limi ou scien ific answe s o socie al ques ions. This was p obably he case o he Cas o p ojec , whe e di e en scien ific publica ions claimed appa en ly obus answe s o ques ions ha emained deba ed. Besides p o iding new analysis and esul s on he p ocesses con olling he Cas o seismic sequence, he p esen wo k aims o ha monize p e ious esul s and hei appa en disc epancies. Using ad anced, wa e o m-based seismological me hods and combining local and egional seismic da a, plus seismic a ays a la ge dis ances, allows us o econs uc he up u e p ocess o mode a e ea hquakes o sho e and hus o shed ligh on he Cas o seismic sequence, p o ing ha his is possible despi e he lack o a dense local ne wo k. Seismic moni o ing egula ions Fig. 7 Timeline o injec ion and seismici y. Tempo al e olu ion o amagni udes, bepicen al dis ances om he injec ion well, cepicen al loca ion p ojec ions along p ofile AB, as in Fig. 2, and dcumula i e momen elease and cumula i e numbe o ea hquakes o he ime pe iod om 2013-09-02 00:00 un il 2013-10-06 00:00 UTC. The onse o h ee main phases a e ma ked by indigo e ical ba s, deno ing injec ion s a ( 1 , 2013-09-02), injec ion s op ( 2 , 2013-09-17), and app oxima e fi s igge ing o he aul hos ing la ges ea hquakes ( 3 , 2013-09-28, 16:00). The colo scale used in a–c e e s o e en la i udes ( op igh ). A dashed line deno es he di usion cu e om he injec ion poin a ime 1 , assuming a di usi i y o 0.5 m2/s, which explains he mig a ion o seismici y du ing phases 1–2, and a dashed-do ed line he as e ea hquake igge ing backwa d p opaga ion wi h a eloci y o ~ 1 km/ day du ing phase 3. No e ha plo ed seismici y co esponds o 964 e en s: 51 accu a ely eloca ed e en s (Fig. 2a) and hose de ec ed by empla e ma ching using he 51 e en s as empla es; we a ibu ed o de ec ed e en s he same loca ions as hei empla es. NATURE COMMUNICATIONS | h ps://doi.o g/10.1038/s41467-021-24949-1 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4783 | h ps://doi.o g/10.1038/s41467-021-24949-1| www.na u e.com/na u ecommunica ions 9