an induced seismicity experiment across a creeping...

JOURNAL OF GEOPHYSICAI. RESEARCH, VOL. 105, NO. 86, PAGES 13.595-13.612, JUNE 10, 2000

An iilduced se i s~~~ ic i ty experiment across a creeping segment of the Philippine Fault

R. Pr iou l . ' F.H. C o r n c t , ' C. D o r b a t h , ' L. D o l b a t l ~ , ~ ICI O g e r ~ a . ~ a n d E. Rsmos4

Abstract. T h e location of seismicity induced by forced fluid flow provides informat ion a b o u t domains of pore pressure variat ion: while changes in fluid c o n t e n t a r e identified t h r o u g h changes in seismic velocity. T h e s e effects have been investigated in t h e geo the rmal field of rrongona.n. which lies o n a creeping por t ion of t h e Phil ippine Fall l t o n Lev te Island. T,ocally, t h e left-lateral strike-slip Phil ippine Fa,ult branches o u t in to th ree subparallel segments (Eas te rn , Cen t ra l a n d Western Fau l t Lines). In June-Jnly 1997, a w a t e r st inlulation was under taken in a well t h a t in tersects t h e Cen t ra l Fau l t Line l 9 8 0 m below grour~ t l surface; 36,000 m3 were iriject,ed between t h e casing shoe a t 1308 m and t h e well bo t tom at 2177 m. T h e seismicity w a s moni tored wi th a surface s t a t ion nctwork of 18 sta.iions. M o r e t h a n 400 events , induced by t h e injection cxpcrimcrlt as well a s by rout ine injections associated wi th t h e gcothcrrnal field exploi ta t ion , were recnrded in t h e vicinity of t h e well. T h e y have been located througlr 3, s im~l l t , aneo l~s t , h r ~ e - d i m e n s i o r ~ d (3-D) velocity-hypocent,er invervior~ procedure. None of t h e microear thquakes a re locateti a long t h e Cen t ra l Fau l t Line, they all occurred below t h e casing shoe t o t h e eas t of t h r farllt line; i.e.. within t h c gcothcrmal reservoir and most ly below t h r b o t t o m of t h e well. Resul ts f rom t h e injcct,ion exper iment a n d t h e 18 rnonths of seismic moni tor ing a long t h c C>erlt.ral a n d West Fa.uIt Lirics suggest a n aseismic behavior of th i s m a j o r cont inenta l fa ,~l l t a t t h i s localion. T h e 3-D velocily n ~ o d e l , determirled l ro r r~ t h c lravel lirrle ir~versioll fur seismic e r e n t s observed dur ing irUecLiu~~s, is c u n ~ p a r e d l o l h a t ub1ai11t.d rrom seismic moni tor ing conducted prior to a n y injection activit ies. A n increase of P wave velocity is observed dur ing t h e wa te r injection. 'This vclocitj- irlcrease is localized a-ilhin t h e seisrr~icily cloud a n d is in terpre ted as an increase in liquid content. wi th in t h e initial liquid-vapor mul t iphase p a r t of t h e reservoir.

1. Introduction rzcr cl al.: 19901. Geological obser\,ations and kine-

In June 1.997, a large water inject,ion rxperinient was undertaker] across a segment of the Philippine Tlal~lt, on Leyt,r Island, a t the T o n g o ~ ~ a n geot,heril~ai field (Fig- ures l a and lh ) . The Philippine Fault is a 111ajor left-lateral strike-slip Cault 1ocal.ed hehiuil a aubclui:- tion zone. I1 cxl.cnds uvt:r. 1200 krrl t l~rough the whole Philippine archipelago [Alien: 1062; F ~ t c i ~ . 1972, Bar.-

r l~alic al~alysis yield a shear displaccmcnt rat,c of 2 to 2.5 cm yr ' for the fault [Ba7-i-iei- c t al.; 19911 Repeated Global Positioning System (GPS) measurement,s con- ducl.cd in briwccn 1991 and 1995 have shown that the fault creeps a l a rate of 3.5 cm y r r ' [Duquesnoy, lii9iI on the northern Leyte seglni3rlt of ILhr Philippine Fault (11'05'-11°30'V). The gcothrrnial field is located where the fault intersects a volcanic arc (Figure i h ) . On site.

thc fault zonc is divided int,o three main parallel vertical 'D-Spartement de Sismologie, irlstitul dc Phgsiqile d ~ i segment,s [i:ast, do>vll to 2.5 km, as stlown seismic

Globe de Pill-IS, France. profiles); they are referred to the East,ern, Central, and ZDCpart~ment dr Sismologii.. lnstitut de Physiqrle dn (Clubc clr Str-ssburng, France. Wrsl.<>l-n Falllt, l,incs (r)ignrp l c ) S i r <:PS ranlpaigns

3philipplne yalional oil c ~ ~ ~ ~ , ~ , E~~~~~ ~ ~ \ - ~ l ~ ~ , ~ ~ ~ ~ cor~duct,ed ;l.r:ross I ) I P West a n d Central 1;;1.1111 1 . in~s he- Curuorai.iai~, hlilliaii Citv. Philinnincc-. !,wee11 1991 and 1097 have shown that locallv the fault . ,

'Philippine Institute of \ i o l ~ ~ ~ o l o ~ y am1 Seismology. Que- crccps a t a rate of'2.4 cm yr-' [Duquesnoy et al.: 1994: zon City, Philippines. L7uqt~csnoy, 1997; Bacolcoi, IgSS]. Although no large

historical srisrllic event lias beer, observed i r ~ the ilortll- Copyright 2000 by the i\rnerical Geophysical Uriion.

er11 part of the island (1l0O5'11"30'N), one event with Papcr number 200OJB900052 0148-0227/00/2000JBY0005L~09 00

magnitude 5.4 occurred on May Ii. 1,993, in the vicinity of the central section of the fault (Figure l b ) . A

13,596 PRIOUL ET .4L: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

11 "09'

l U O 1l006'

G o 11'03' l Z U D L Z 4 O 124'36' 121°3Y' 124'42' 124'45'

Figure 1. (a) The lnajvr left-lateral strike-slip Pliilippi~~e Fault is located behind a subduction zone and extends over 1200 km through the whole Phili pine archipelago. Repeated GPS measurements yield a shear displacement rate of 3.5 cm yr-'on the northern Leyte 50 km segment and close to 3.3 cm yr-' on the Masbate Island segment (measurements from Duquesnoy [1'3'37]). (h) Structural map of Leyle Island where the fault irrtersects a volcanic arc (modificd from Aurclio [1092], 1, recents volcanics; 2 and 4, Pleist,orene and mid-Mioccnr limestones; 3 and 5 , late and mid-Miocene sediments; 6, volcanoclastics; 7, magmatic rocks: 8, ophiolits; I), marine deposits). The only known earthquake in central Leyte larger than magnitude 5 is also displayed. (c) Map of the Tongonan geothermal site where the fault zone is divided into three main parallel segments, called the Eastern, Central: and Western Fault Lines. Locations of the 18 seismic stations are displayed as well as the horizontal projection of well MG2R.D and t,he power plant,s. Single and three component stations from PHIVOLCS and EOPGS Instituts have been used (see text for details). Topographic conlour lines are displayed every 100 rn.

few moderate events with magnitude close to fi have occurred since 1980 in the southern section (10'-10°30'N).

For the last 10 years, the Tongonan geothermal field has hpen ~xt,ensively developed, and large amounts of brines, generat,ed by t,he exploit,a.t,ion, have t,o he rein- jected. For that purpose, a few boreholes were drilled in br(wi!e~i the Western and Easlern Fault Lines These were f i ~ l n d 10 I I P ~ T T ~ P P P Y ~ C I L I S . 01lr CII' these wells was se-

lected to conduct a hydraulic stimulation experiment. A characteristic of the selected nonvertical well is that it intersects one branch of the central fault. The concept of tile experiment was to slowly build up tllc irljcctiori flow rate so as to stimulate, by shearing and cooling, the preexisting fractures intersected by the well [Cor.net and Juiaes, 1994; Corrhct cl ul., 19971. Prior lo l l~is imjec- tion a surface seismic netwurk was ir~stalled i r ~ order lo monitor any induced seismic activity.

.4fter a short description of the injection experiment and of the seismic monitoringsystem, this paper presents the results of the induced seismicity analysis. Local earthquake travel times recorded during the experiment period are inverted in order to simultaneously obtain the t,hree-dimensional (3-D) velocity structure and the locat,ion of hypocmt,ers. Precise relocations of seismic events are analyzed in terms of domains of pore pres-

sure variation in the vicinity of the fault. Results of the 3-D velocity structure variation with time are discussed in terms of fluid content changes and temperature per- t,urhations caused by inject,ions. Finally, a mechanism for interpreting the observed seismicity is discussed.

2. Illjectioll Experiment

A large hydraulic injection experiment has been undertaken in well MG2H.D (Figures lc and 2). This devi- ated well was selected because it intersects the Central Fault Line -1080 m bclow ground surfacc according to geological interpretation (Figure 3; depths are here inafter referred as meters below ground surface: m bgs). Thc intcrscction with thc fault was coniirmcd by a thermal discontinuity, observed on a thermal log conducted just after drilling (Figure 4). The wellhead is located 260 m above sea level, while the botlorn of the well is located 2177 m bgs lo the east o l 111e fault [Bondocoy, 19931. The borehole is cased down to a depth of 1308 m bgs. Below this depth, the slotted liner section (per- forated pipe allowing inward and ont,ward flow) intersects the Mahanagdong Claystone (MC) and then pene- trates the hlahiao Sedimentary (:omplex (MSC) after it has crossed the central fault. 'The Mahanagdong Clay-

PRIOUL ET AL.. INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

Injection experiment

. I : . . . . . ...

. . Day (June-July)

Injection

Figure 2. Injection flow rate, wellhead pressure, and level of microseismicity monitored during June-July 1997 for the injection experiment in well MG2RD and for the injection tests in wells MGlSD, MG20D, MGZlD, MN1; MN2RD, and MN3RD. The increase in flow rate, for MGZRD, is followed by the increase in wellhead pressure and seismicity level and suggests that induced seismicity results from a pore pressure increase. The two short peaks, observed before and aft,er the MG2RD injection, correspond to pumping tests for the experiment.

experiment .'. S::: . . . . . . .

., . . . ., ....

1

stone formation is a thick sequence of predominanlly fine clastics such as claystone, siltstone: a i d sar~dstur~e and is assumed to exhibit plastic behavior. The hfahiao Sedimentary Complex is a sedimentary conglomerate cont,aining q~iart,z and monzodiorite fragments with un- metamorphosed sandstones and siltstones and was anticipated to exhibit a more brittle behavior than the Mahanagdong Claystone formation.

" ' . .. . . , , ,

A survey conducted in the well before the injection experiment has shown that the well has a diameter reduction from 12.7 to 8.Y cm at around l880 m bgs, i.e., slightly above the depth where the borehole intersects the Central Fault Line. It was hoped that below this depth, the liner was still open at its nominal diameter.

The goal of the injection experiment was to stimulate, by shearing and cooling, the preexisting fractures that

l

I

Irijcction tests

MN3RD ,' MN2RD

i

PRIOUL ET AL.: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

WEST FAULT r LINE CENTRAL FAULT

Figu re 3. Cross section of well MGZtlU (direction NNE) that crosses the Central Fault Line 1980 m below ground surface. The well intersect,^ an andesitic formation (MF), a thick sequence of predominantly fine clastics (MC), and a sedimentary conglomerate (MSC) after it has crossed the fault (modified from Delfin et al. [1995]). Stratigraphy and isotherms were deduced from borehole loggings. Vertical depths (meters) are reported with respect to sea level.

intersect the well in the MSC formation. The injected fluid was river water at a temperature of 25'C. Starting on June 19 with a flow rate of 10.6 L/s for 24 hours, the flow rate was increased by 10.6 L/s increments every other day (Figure 2) so as to reach 53 L/s after 8 days of pumping. Because of the pumping capacity and the wellhead characteristics the flow rat,e had t,o he kept, constant at 53 L/s during the last 4 days of the test. The stimulation ended on June 30 with a total injected volume estimated at 36,000 m3.

After 11 days of pumping, the well was closed for 24 hours. Then, it was opened and logged with a Kuster tool (pressure and temperature gauges) during the flow back (from 1850 m bgs just abovc thc lincr rcduction). The temperature log (Figure 4) shows that the temperature was drastically reduced by the injection over the entire length available for logging (- 75'C). The lowest temperature (71°C) is found at the maximum ac- cessible depth and demonstrates that most of the flow reached below this depth into the Mahiao Sedimentary Complex. The thermal anomaly, which occurred during injection and is identified between 1400 and 1500 m bgs, is a small inflow associated with minor losses within the Mahanagdong Claystone.

For exploitation purposes and independent of our experimentation, injection tests were conducted in six wells in the vicinity of MG2RD. In particular, two wells, MG15D and MG20D (Figure l), have been sporadically used, st,a.rt,ing on June 2,5, wit,hout any flow rate and wellhrad pressnre monitoring. Then a regular injection program started, on July 5, in the sir wells, MGlSD, MG20D, MN1, MNZRD, MN3RD, and MC21D. Ini- tially, large injection flow rates (up to 127 L/s) have been imposed. Yct they have bcen associated wit,h

only small wellhead pressures (up to 1.5 MPa), a feature which demonstrates the initial high permeability of these wells. The flow rates and wellhead pressure are summarized on Figure 2. For later use in the paper, we need an evaluation of a lower bound of the total injected volume. In this respect, it is safe to assume that during the June 25 to July 5 period, the injection flow rate for wells MG15D and MG2OD was larger than 114 of that used during the July 5-15 period. Consequent,ly, t,he total injected volume in these wells, during the entire experimental period (June 25 to July 15), has been estimated at 327,000 i 20,000 m3. It is worth mentioning that a large decline of the injection flow rate capacit,y was observed in MN3RD (79 to 20 L/s), 2 months after the experimental period, as well as a slight decrease in MN2RD (54 to 42 L/s) and MG21D (119 to 108 L/s).

3. Moriitoring Seismicity 3.1. Record ing Networks

'The microseismic activity in Tongonan has been monitored with various types of surface stations. First, a permanent telemetered network of seven stations was installed by the Philippine Institute of Volcanology and Seismology (PHIVOLCS) betn,een February 1996 and July 1997 in order to continuously monitor the local seismicity. In addition, a seven-station telemetered network and four independent stations were deployed by Ecole et Ohservat,nir~ dp Physique du Glolhe de Stras- bourg (EOPGS) for a s ~ ~ r v r y conrlilct~~rl di~ring Octoher- Novcmber 1996 and for the injection experiment during the June-July 1997 period. Location of the various st,at,ions is shown on Figurc lc. The stations of both telemetered npt,works were equipped with a singlc vcr-

PRIOUL ET .%L: IXDLICED SEISM iICITY ACROSS PHILIPPINE FAULT 13,599

0

Before injection

400

July 1-2, 1997

2200 0 50 100 150 200

Temperature (C)

Fipllre 4. Th~rnmal logs conductetl, in MG2RD, after drilling, before the beginning of the injection experi~ menl, and after t,he inject,ion (flow back of water). A thermal anomaly is clearly identified around 1980 rrr bgs where the well intersects the fault (August 19, 1993, log). Temperature has been drastically reduced by the injection experiment over the entire length of the well (July 1-2, 1997: log).

tical l-Hz seismometer. Signals were transmitted by UHF t o a central recording system and recorded simultaneously with a GPS-synchronized timesignal. Signals recorded with the PHIVOLCS network were digitized a t a rate of 75 samples per second (sps) and signals recorded with the EOPGS net,work a t a rate of 185 sps. The other four independent stations (three components, 1 Hz, 185 sps) had their own trigerri~ig and time rrcord- ing system. For these stations, a GPS-synchronized t ime was added every 4 days. Using (,his as calihra- tions, we eslimate the timing accuracy bo 0.02 S.

The geomeiry of t,hr ndwork was chosen so as t,o study seismicity along the actively creeping fault scg- i~ieiit of the Philippine Fault as well as the seisrnic activity induced by the injection experiment. Thc whole set of stations was i ~ ~ s t a l l r d iri bdcr~sc: netrvork alollg the West and Ccr~tral Fault Lines: vrientcd roughly NNW- SSE (Figure l c ) . The azimuthal coverage of the network was limited by the inaccessible mountain to the east. The area covered by all the stations was about 20 km by 10 km in the NNW and EYE directions, respectively.

I t should be noted that the network elevation reference has been taken as the rlevaliorr of the lowest s ta t io~l , some 400 m above sea level. The elevation of the highest station was 1015 m above sea level.

Depending on the deployed instruments, four periods are identified: period l (February-August 1996) and period 3 (December 1996 t o May 1997) during which only the PHIVOLCS network was operated, and period 2 (50 days, October-November 1996) and period 4 (43 days, June-July 1997) during which the complete set of 18 stations was installed. In this present study, seismicity a.nalysis i s fociised on periods 2 and 4, i . e , before and during t,hp experimmt,a,l injection period

3.2. Microse i smic A c t i v i t y

A total of 1760 events were detected in the vicinity of the network during periods 1, 2 , 3, and 4, with -500 events for perind 1, 240 for period 2 , 460 for period 3, and 560 for period 4 wit,h coda ma.gnit,ijde range between 0.5 and 3 . Events of periods 2 and 4 were first located with the HYPOINVERSE program [Kliin, 19891 i~ r ing an updated a priori l-D four-layered velocity model (model 1, Table I ) . Thc updatcd a priori 1-D model had been previously established, following Iiissling et al. [1994], with period 1 da ta and an a priori l -D model based on geologic informations. Travel times were corrected for s ta l ior~ elevaliori by assuming the same near-surface crustal velocity as layer 1. No correlation between travel time residuals and station r l eva t io~~s has been found. Events locations were considered well constrained when their RMS arrival time residuals were <0.15 S; for all selected events the mean RRIIS value is 0.05 s. Events are distributed down to 10 km. The locations of well-constrained events for period 2 and 4 are presented on Figure 5. They correspond t o the first step for the selection of a consist,cnt da ta set for further analysis of t,he seismicity.

4. Tomography Inversion

4.1. I n i t i a l S e t t i n g s and I n v e r s i o n P r o c e d u r e

Travel tinles used in the tomugraphic inversion were selected from events which met the following criteria:

T a b l e 1. One-Dimensional a priori Velocity Models

l-D Model 1" I-D Model zb 2, km V p , km/s V V,,, km/s V,,/K

*Model 1 was used for previous locatio~~s. "Model 2 was computed for the use of a priori information

for the tomugraphy inversion.

13,600 PRIOUL E T AL.: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

Figure 5. Map of the epicenters (circles), seismic stations (triangles), well MG2RD (solid dot and line), and model grid spacing used in tomographic study. The 141 events and 292 events were selected for periods 2 (October-November 1996) and 4 (June-July 1997), respectively. The 3-D velocity valnes are calculated at the intersection points (nodes) of the spaced, 1 X 1.5 km. horizontal grid. The vertical grid spacing is 1.5 km.

number of P readings >7, number of S readings 2 2 , RMS 50.15 S, horizontal standard error 50.5 km, vertical standard error 50.8 km. This gave 292 events for period 4 and a data set of 5291 travel times (3939 P, 1352 S). Although the number of Stravel times is insufficient for giving a well-resolved S velocity model, they constrain hypoccntral dcpths and hence the inversion. Events from period 2 were sclcctcd with thc samc criteria. This gave 141 events (313i travel times. l743 P, 1391 S). Despit,e t,he fbwer number of events for period 2 , t,he spat,ial coverage is of het, t ,~r qilality for bhis period since events are not concent,rat,ed close t,o inJect,ion site (see Figure 5 ) .

The selected travel times for P and S wavcs wcrc inverted in order to determine sirnultitneously earthquake locations and the 3-L) velocity structure beneath the seismic array. The tomographic inversion method (SIMULPS10) is an ileralive damped least squares tech- uiqut: wlricl~ was uriginally developed by Thurbrr [l9331 and ~nudified by Ebrrlzurt-P/irllips [1990]. The rnelhod has been applied to active faults area by Eberhart- Phzllzps [1Y86, l Y Y U , l9Y3] and Dorbalh et al. [1996]. A modified version of the computer program (SIMULPS12) that can invert both P and S-P times [Thurbe~, 1993; Evans et al., 19941 was also applied to geothermal areas by Foulqer and Miller [l9951 and Julian et ul. [1996].

As detailed by Eberhart-Phillipj 119861, the \relocity of the medium under investigation is parameterized by as- signing velocity values at the intersections (grid points) of a nonuniform, three-dimensional grid. The propaga- tion velocity for a point along a seismic ray path and the

velocity partial derivatives are computed by linearly in- terpolating between the surrounding eight grid points. The area covered by the stations, the hypocenters, and the three branches of the Philippine Fault extends some 20 km along the faults and 10 km across them. The coordinates are chosen so that the Y axis is oriented parallel to the average strike of the Philippine Fault in this region (37' west of north). The grid frarne is represented on Figure 5

Next, a precise initial 1-D velocity model has been determined. We inverted the travel times from periods 2 and 4, with the preliminary l-D model used for the first location determination as an initial modcl (model 1 , Table l ) , so as t,o determine a minim on^ l -D velocity model [I<issling et al.. 19941. This inversion led to a variancc rcductron of 27% and 21% for 1' and S waves residuals, rcspcctivcly. Thc resulting onc-dimensional model is show11 ill Table 1 (model 2).

Once the general frame for the inversion was set up, different inversions were carried out with different grid spacings, different darnping parameters, and different initial velocity models. The grid spacings were chosen so as to resolve small velocity perturbations associated with the injection experiment as well as velocity contrast,^ across the two fault segments (WFL, CFL), which are only 2 km apart from each other. We used a regular l X 1.5 km grid spacing along the horizontal X and Y axis, respectively, in order to have rea- sonable resolution at most grid points. The vertical nodes extend from 0 to 10 km (maximum depth of seismicity) with intermediate nodes at 1.5. 3 and 5 km.

PRIOUL ET AL.: INDUCED SEISMlCITY ACROSS PSILIPPINE FAULT

(a) Pariud 2

(b) Period 4

Figure 6. Diagonal elements of the resolution matrix for the final models of (a) period 2 a l ~ d (b) period 4, at Laycr node 0 , 1.5, 3 km (plan view). Resolution reflects station spacing and distribution of seismicity. The inversion grid has been rotated fro111 37' NE. Contours units are 0.2. Stations are displayed with crosses; and MG2RD welll~ead is ~rlarked with a star illark.

Thc clcvation reference is that of the lowest stat,ion, i e., 400 m above sea level. When grid point spacings were varied (1 X 1 km, l .ii X 1 km) or grid point locations were translated by 0.5 km (X or Y ) ; thc gcncral shape and location of vclocity anomalies wcre not modificd. In this tomographic inversion, one has to find the danlping pa~amete r that gives the best compromise between data variance reduction and the model variance. We performed inversions using 10 different values for the damping parameter (between 2 and 100 sZ krrlrl) and analyzed "trade-uIT curves" belweerl resolutiurl a r ~ d starldard errurs [ E u u r ~ s utzd Acl~uuer , 19931. Fro111 these

tests the damping pararnet,c-r was cliosrn so l.klat the diagonal t,rrms of t,he a posteriori covariance rrlatrix wollld be snlaller than 2% of the mcan P velocity of each considered layer for period 4. Thcn we tested the sensitivity to thc initial vclocity model for the three- dirncnsional inversion. Tests were done with various initial velocities. An alternative a priori model (model 3) with depth spaclng every kilometer was determincd as for model 2. After different three-dirl~e~lsiurlal i~~r<: l - - sions with startirlg ~ ~ l u d e l s 2 arid 3: thc rrrran P velocity and the patbern of ve1ucit.y variations in each layrr were llearly identical for the two nlodels down to a dept,h of

13,602 PRIOUL ET AL.. INDUCED SEISb AICITY ACROSS PHILIPPINE F.4ULT

5 km. They differed by a t most fi%: and this occurred a t the s~i r facr where velocity constraint is minimal because of insufficient coverage. Moreover, the differences ill the epicenter locations were generally <300 m . In summary, we find that the inversion process, for these data , is robust, in tha t the differences in i.hr rrsillls are not very sensitive t o the initial velocity model and t,he grid spacing.

4.2. I n v e r s i o n R.es111ts

After five iteral.ions, the P da ta variance was reduced by 51% froin 0.010 t o 0.005 sZ for period 4 and by 42%, from 0.018 t o 0.010 s v o r pcriod 2. The diagonal clc- rnents of t,he resolution inatrix for the three upper layers fur periods 2 and 4 are shown on Figure G . i'alnes fur the diagonal terms of the resolution niatrix railgc fro111

0 (no resolutioti) lo I (perfect resolution). The resolu- 1ior1 a l l l ~ e surface reflecls the slation distribution: il is high close t o l l ~ r s l a ~ i o i ~ s a r ~ d low elsewllere. The resolution improves with dept.h for the 1.5 and 3 km depth mode layers, where a complete zone can be defined as being well resolved. 'The highest resolutiorl is observed for period 4 a t depth 1.5 km with values close to 0.7 (Figure 6b). Deeper resolution decreases rapidly, but a zone with resolution >0.2 call still be idriibified. The resolulions for the two periods are quite silnilar in tlie central region and indicate no significant differences in sampling.

Standard errors for velocity depend on resolution and o n mean velocity. Standard errors rrmnin <2%, of (,he mean P velocity for pcriod 4 da ta (73% of da ta <l%, a t 3 km depth) and 4' for period 2 data (72% of the dais ~ 3 % ) For t , h ~ mndrl space where resol~~t ion is >0 5 ; velorit,y errors are close to 0 06 km S-' at, thr s~lrfacc. 0.07 km S-' al. 1.5 km a n d 0 09 km S-' al 3 km depth (period 4) . The mean horizontal error for evrnt relocation is -60 ni, and the vert,ical error is -100 m.

4.2.1. Earthquake loca t ions . Epicenters of the rclocat,cd cvcnts for pcriod 4 (43 days of recording) arc displayed on Figure 7. 4 cor~iparison between the initial and the final epicr:nter locations shows ir~odest differences: 59% of the 292 relocat,ed events differ by <O5 km from the initial epicenter. 87% ditfer by <l km. The differences in depth are quite similar: 64% of thc events d i rer by <0.5 krri frorn the initial depth, and X:{% differ by < l km. Uespite these srriall shift,s, the main seismicity cluster exhibits a notable differeiice in shape and shows inorc individual chisters (Figures 5 and 7).

Most microseismic events arc located to t,he east of lhc Central Fault Line, within thc geothermal reservoir, with only very few events along the Central Fault Line and t,o t.he west of it,. The recording period (Fig- ure 7) has been divided into tlirce subperiods in order t o scparatc the scismic activity induced by the MG2RD injection cxpcrimcnt from that induced by other field activities before or after. In thc vicinity of MG2RD, no

significarit activity was observed before .June 18; microseismic activity is ohserved only alter the i~~.jcction has st,arted. Once the sbimulation ends, the associated seis- mlc cloud disappears. The absence of seismicity and field activity prior t o and after the experilllent shorvs that t,he observeid seismicit,y in the direct vicinily of the well, during rhe days of pumping, is linkcd only to the MG2RD stimulation.

More prerisrly, t,he level of microseismii:it,y per day for pcriod 4 has been displayed along wilh flow ra1.t and rvellhead pressurr for t,hc varlous in~ectioll activities (Figure 2). For the MG2RD cxpcrimcnt a clcar temporal correlation can be observcd with thc incrcasc in the number of microseismic events and thc incrcasc in wellhead pressure. Moreover, just aller the end of the stimulation: the level of inicroseisl~~icity rlert.eases ill,- rnediately aiid follows thr monitored decrrasc of pressure. The incrcase in wellliearl pressure and the facl tha t most of tlie flow rraclierl irllo the Mahiao Sedi- ~ritinlary Coniplex suggest strongly that pore pressure has been increasing in the MSC formation. 'This mechanism has been ofteii described in the l~tera ture [e.g., Ptnrson, 1981; Pine and Batchelor, 1984; l-ehler; 1989: Cornet, 19921. Other mechanisms for geothermal sites such as fluid extraction [Segall, 19891 or thermoelastic effects [12Iosso~, 19981 have also becn associated with indnced seismicity but arc not considered relevant here.

F'igurc 7 also indicates tha t the stimulat.ion has not induced any cvent along the Central Fault Line. The geometry of well klG2RD is known with an accuracy of 60 m , and the wellhead location is known to wilhin 50 m . Hence the uncertairity of ihe distance between the locat,ion of the well aiid the location of an cvent is close t,o 170 m . The eartliquakc activity remains icoi~fined within a volume which extends frorn the bolloni of thc well down t o -4 km (3.6 kill below sea level). No event has been induced in the upper sectiorl or the wcll ahovc the Central Fault Line. No clear doxvnward migrat,ion with t ime can been identified from the analysis of the scismic cloud. The absencc of seisnlic cvtnl along the Cault during thc injcctlon cxpcr~ment may reflect either an absence of' cl~angc in pore pressure or thc occurrrilcc of aseismir: slip.

In addition, wells MG15D and MG20D exhibit few induced seismic events frorn June 25 to July 4. These are clearly confined to the cxtrcmit,y of each of these wells and ar r fairly distinct from the MG2RD seismic cloud (Figure 7) . The regular iu,jectioli activities in MClSD. l lG20D, ?vlNI: kIINZRD, IVIN.?RD, and R;IG21D, after July 4, have i l i d~~ced seismic activity local to these wells (Figure 7) . Although t,he lotal injected volumc is more important in t,he regular irijed.iort program t,han that of the MG2RD st,im~rlat.ion, u~ellhead pressurc a t thc other wells is iliuch lower (clos(: t,o I hIPa) than that a t MG2RD (-9 MPa a t t h r elid o l the expcriment), a feature which may explain why the correiation bct\vrcn tlie level of microscismicit~y and tlic i~icrcase of porr pres- S I I ~ P is not as clear for t,hese wells and why the lcr,cl of

PRIOIrL ET A L INDUCED SEISM[CITY ACROSS PHILIPPINE FALLT

13,6W PRIOTJI, ET AT, INDIJCED SETSMICITY ACROSS PHILIPPINE FAULT

seismicity related to MG2RD is so high. The seismicity associaled with llle ol11t.r wells extends also from the bottorn of the wells duwrl lo -3.5 kni.

The map of relocated evn11s [or period 2 (Figure 8) exhibits the same features as fbr period 4: .Most microseismic events are located to the east of the Central Fault Line, within the geothermal reservoir, with only very few events along the fault and to the west of i t . In addition, events recorded for periods 1 and 3 (14 months) have also been relocated using the a priori 3-D velocity model for period 2 (Figure 8). These observations have to be interpreted with care because of the lower source locabion accuracy due to the fewer number ofstations. In the northern part of Tongonan, whcrc thc azimutal coverage of the stations is the best, seismicity is locaded within t,he geot,hermnl site with most of it a t shallow depth (1-5 km), and very few evrnt,s are locat,ed along the West and Central Fault Lines. Southwest to the network, evenbs ncrnr at greater dept,h (mostly 4-7 km) hut are more scattered and cannot, he easily associated to any particular structure.

4.2.2. Velocity model. The P wave velocity models in the first three node layers is presented on Plates l a and l b for periods 2 and 4, respectively. They have been contoured after interpolation. On these Plates l a and l b , the zones with a resolution <0.2 for both periods were not represented. This mask allows us to keep only the most reliable information for both models. At greater depths the well-resolved zone tends to shrink.

The overall pattern of anomalies found in the two models, on Plates l a and l b , is similar, and particular features may be observed. At the surface (0 km) the well-resolved information is concentrated in the vicinity of the stations. Thc low V, vclocity zoncs (-10 % from the mean V, at the layer) are observed at layer 1.5 km direct,ly heneat,h t,he Mahiao, Malit,hog, and M3ha.na.g dong areas, which correspond t,o t,he most prod~lct,ive zones found in the geothermal field For example, siguif- icant low velocit,ies are ohserved in Mahiao area, where hot fluid upwelling occurs. The low vrlocit,y heneat,h Mahanagdong is also observed for the 3-km layer. A

SE Azimuth 145' (km) NW SE Azimuth 14:' (km) NW 0 I 2 1 1 5 6 I B 9 10 I i 12 11 I4 i5 l6 11 1 2 3 1 5 b 7 8 "10 i l 12 r3 i & I 5 Ib 17 I8

Figure 8. (left) Seismic events for periods 1 and 3 (7 stations, 14 months) have been relocated using the 3-D velocity model determined for period 2 (469 events). (right) Relocated seismic events for period 2 are also displayed, as wcll as azimuth 145' cross sections. Thc horizontal projections of the geothermal wells are also shown shaded.

PRIOUL E T .1L: IYDUCED SElSMlCITY ACKOSS PHILIPPINE F.4CLT 13.C0.5

high I/p zone is located in the Bao valley, where no flow of high-temperature fluid exist,s within the imperme- able basement rocks and the Mahanagdong Claystone [Delfin et al.. 19951. A small low-velocity zone call be observed on cross sections of both models beneat,h t,he Central Fa l~ l t 1.inr (see Y=1.5. Plat,e 2) .

The most prominent difference is observed on node X = 0. Y = 1.5 a t layer 3 km for period 4. It is a high P wave velocity zone in the vicinity of the injection wells (Plate l b ) . This anomaly, located close to Mamban area, is also slightly seen in the 1.5 km layer. This Iligll-vclucily zone is not observed for period 2 (Plate l a ) . The feature appears Inore clearly on a cross sectioli (see Plate 2, X = 0; Y = 1.5). For this node, a t the 3 kni layer clepth, the periurbation oC Lhe P wave velocity shows a 14% increase between the two periods (from 4.44 to 5.04 km S- ' , hereinafter callcd period 4 anomaly). On Plat,e l a , one car1 also observe a high-velocity zone [hereinafter called period 2 anomaly) centered 2 km north of the first anornaly. This period 2 anomaly is located close to a lower-resolution aonr and is not present in period 4 .

A question may he raised concerning the reality of these velocity anomalies. l'hc two fields experi~uents of periods 2 and 4 have provided two independent da ta sets with the same nct,wnrk e;eometrics and instrurnen- tation. Furthrr. for bot,h data sp t s the it~ve~ninli lneth- ods and paranictcrizatio~i arrt the same. In the models the difference between the rnean I-; value in cach layer is < O 03 km s-' The similarity in the overall velocity pattern for both inversions gives us some confidence irl the results. Computations of v, a ~ r d Cb/V, ratio velocity with computcr codes SInlllLPSIO arld SIMULPS12, reprctivt.ly, clu ,lot yicld well-resolved modr.ls, but the ovrrall paticru of t,he t,wo model3 are "cry sinlilar. lie- sults from SI!d,IIJLPS12 show a high b/I< ratio at the location of the period 4 anomaly.

Before discussing physical processes tha t would lead to such velocity changes, we describe nurrlerical i e s ~ s which have heen conducted in order to assess t,lre ro- bustness of these observations. First, different subsets of the two da ta sets have beci~ used in order to insnri: that the models are not hiased by tlie poor quality of some data . For each da ta subset the general pattern was found to be ider~t,ical to t,he model derived from the coniplete da ta set. Accordinglyl tlie existence of an artifact induced hy erroneous data is excluded. Sec- ond, an import,ant concrrn is to insurc that different ray pat,h distributions from periods 2 and 4 ~voiild sample the pegion or interest s ~ ~ i l i c i r n t l ~ well for resolvilig botli anomalies.

The first test consist of inverting the da ta from pcriod 2 with an a priori l-D modcl tliai iricludcs the observed perturbatinrl of period 4 (,Y = 0, Y = 1 5 , Plate 3 a ) If the anomaly is erased after computation, tha t ur-ould mean that enough ray paths have resolved the zone and corrected it from the introdilced a priori anomaly. In pract,ice: 12 nodes (layers 1.5 and 3 km) of

the observed anomaly were included in the previous a priori l -D model (model 2) . Plate 3a shows the new a priori model and Plate 3b shows the result of the new inversio~i for period 2 for the 3-km layer. Clearlyl it is observed that the a priori anomaly, a t the location of the induced seismicity, has been "erased". More- over, the velocity pattern is the same as tha t shown on Plate 1. Layer 1.5 k m exhibits similar results. Hence it is concluded that the ray paths coverage is sufficient to resolve zones of interest with period 2 da ta and that the period 4 anomaly was not present a t tha t t ime; that is, the model space a t the location of the period 4 anomaly is well constrairled. More quantitatively, a second test has beer1 desigr~ed L V observe ihe imfluerlce or ille ray paths distributiuri on the lucation and amplit.ude of two ano~nahes in t,he inversion process. We prepared an a priori model as follows: a 15% P wave velocity perturbation has been int,roduced a t t,he locat,ion of the two nodes corresponding to the period 2 and period 4 anomalies in an a priori homogeneous model (4 km S-' :

see Plate 4a). Then, the two hypocenter distributions from periods 2 and 4 have been used separately t o cal- culate synbhebic travel times for the new a priori model. Next, the synthetic travel times of both periods have been invertrd with an a priori homogeneous model (4 km S-') in order to try to reconstruct the introduced anomalies. Results of this process, for the 3-km layer, clearly show (sec Plates 4b and 4c) tha t the period 4 ar~omaly is well reproduced in location for the two pcriods and both inversions indicate similar amplitude attmmlation. The designed 15% anomaly is recovered, for both nlodels, as a 4% velocity anornaly. This con- firms the result of the previous test. Had the period 4 a ~ ~ o ~ ~ l a l y I~een preserli during period 2 , ihe tornography rnrthotl wuilld havc uullirled it irl tlle sarrle rr~arli~er as in pcriod 4. Henc~; comparison of the motlels withill the induced seismicity zone can be conducted with greater confidence. It is concluded that the injection of water in wcll MGZRU, and in other nearby wells: has induced an increase of P wave velocity in the vicinity of the wells. hut evaluation of all~plitudes has to be considered with care.

However. it shi>uld he noted t,hnt Pla1,es 4b and 4c show t.liai t.lie amplitude of the period 2 anomaly was not reconstructed in the same manner for {.he period 2 and period 4 da ta distributions, so tha t it is difficult 10 corrLpare t l~eru. 11, addition, the reconstrurtioli of lllr 2 a n o ~ i ~ a l y has it>troducecl some smearing for the period 2 data set. Helice the existcncc of period 'L a n o i ~ ~ a l y is qi~esrionahle.

The secorld test has also hem r~pen t , rd wit,h a rlat,a, set of 231 eartiiquakes generated on a ilniforni grid at, 6 km depth. In this test the amplit,ude of the designed anomaly is recovered as well as a 10% velocity anorrraly. This suggests bhat our model computation might havc underest,imated the real arnplitudc of the anomaly he- cause of the seismicity distribution.

PRIOUL ET AL.: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

Plate 1. Three-dimensional P wave velocity model (km S-') (a) for period 2 and (h) for injection period 4, at layer node 0 , 1.5, 3 km (plan view). The inversion grid has been rotated from 37' NE. Color scale represents P velocities in between 2 and 6 km S-'. Only results that exhibit a resolution highcr than 0.2 for both pcriods 2 and 4 are shown, in ordcr to sclcct rcliablc information for both models (see text). P wave velocity anomalies can be identified hetween the two models, especially near the injection sites.

Figure 10: Plntr 2. Cross sections of tlie 3-D velociiy nlotlel (km S - ' ) across ( Y - 1.5) and along (X - 0) the fault. The period 4 anomaly is clearly identified (see text) . Depths are given with respecl lo bhc network elevation reference, i.e., 400 m above sea Ievcl.

13,608 PRIOUL E T AL.. IXDUCED SEISMICITY ACROSS PHILIPPIYE FAULT

Plate 3. Test for the detcctcd pcriod 4 anomaly. (left) Data from period 2 were inverted again n i n g an a priori model where the anomaly of period 4 has been int,roduced in an homogeneous layered model ( ID) at layer node 3 km. (right) Results show the resulting V; model (layer 3 km). The introduced anomaly has been re- moved out from the a priori model.

At other geothermal sites, low Vb/lj8 velocities have also been reported [.Julzan et al.. 19961 but no evi- dent anomaly was observed on V, alone. In our case the results from the velocity models have shown that

changes in fluid content because of injections (period 4 anomaly), have induced at least a 14% increase of the P wave velocity in a rock volume distributed in depth ranging between 1360 and 2860 m bgs.

6 . Discussion

5.1. Seismicity and Philippine Fault

The microseismicity observations during one and half year have shown that very few events (magnitude 0.5- 3) occurred, in Tongonan, along the Central Fault Line a n d t,he \Vest Fault T,ine. Moreover, the stimulation experiment by forced fluid injection across the Central Fault Line did not induce any microeveilts along the fault. GPS measurements that have been conducted before and after the ~xperiment (June and July 1997) do not yield any detectable displacement along the fault that may indicate any faster aseismic slip because of the experiment. Irl between 1991 and 19Y7, less than 10 earthquakes of magnitude 4 and none with largcr mag- rlilude hake l~eeri recorded itlong Lhe 50-km Y C ~ I I I ~ I I ~ uf the Philippi~le Fault i l l rrurthcrr~ Leytr (ll"05'-11"30°'N) by tbe Philippino national network or ~vorldwide seismicity.

Only one earthquake of magnitude 5.4 was located 12 km east to the East Fault Line on May 17; 1993 (Harvard epicenter: 11.05'Y, 124.84"E). This event was attributed to the Philippine Fault after a seisinic survey recorded 70 aftershocks in the vicinity of the East and Central Fault Lines, south to Tongonan (latitude

Plate 4. Tcst for the detected period 2 and 4 anomalies. The velocity perturbation (9%) to an homogeneous velocity model (Vp) of 4 km S-' is shown. (a) Two 15% velocity perturbations have been introduced in the a priori homogen~ous model. This model was used for the co~~lputatioil of synthetic travel times for both seismicity distributions of periods 2 and 4 . (b and c) Results of the P wave velocity model after thr inversion of synthetic travrl t,imes a.re shown for periods 2 and 4 at layer nodc 3 km.

PRIOUL ET AL.: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT 13,609

1O009'-11°06'N, [Narag et al., 19931). The possibility of occurrence of this event within the geothermal field has been rejected for two reasons. First, the Harvard location is biased in the direction 50.80' and 230-260" due t,o the lack of stations in the Indian and Pacific Oceans, so that the latitude of the expected event along the East Fault Line might be in between 11.02' ahd 11.03'N (i.e., south of the geothermal field). Second, geodetic measurements conducted in February 1991, June 1993, May 1994, November 1995, and June 1997 all along the West and Central Fault Lines have shown a yearly repeated shear velocity close to 2.4 cm yr-l [Dvquemoy et al., 1994; Ouquesnoy, 1997; Bacolcol, 19991, which do not indicate any faster slip because of such an earthquake within the geothermal site.

The questio~l which is Lo be discussed riext concerns the possibility of establislli~rg the seislrlic or aseislrlic behavior of t h e e fault segmerlts fro111 seisniic observations. Considering t,he cunlulated shear displacement up to 15.2 cm ihat occrtrred within 76 months, it is of intcrest to cvaluatc thc magnitude and seismic moment that would produce a single event for the correspond ing observcd displaccmcnt. For a purc uniform shcar motion the seismic momcnt (M") is rclatcd t,o t,hc shcar modulus (C), to the area of the source (S), and to the dislocation amplitude (d) by MO = GSd Define S a s the product of the length of lhe fault where displacerrlenls have bee11 measured, i.e, 15 krn, by the deplh of the seismogenic zone which is close to 8 km. If the shear modulus is taken as 15 GPa, the observed displacement should produce a seismic moment close to 2.74 X 10" N m. Using Kanamori and Anderson's [l9751 empirical relationship between magnitude and seismic moments for major earthquakes, one might have expected an eart,hqi~ak~ with ma.gnitude of the order of 5.6, when none was observed. We consider next the possihilit,y in which slip occurred in multiple seismic events; 220 events of magnitude 4 or 220,000 events of magnitude 2 would be needed to generate the expected slip motion. Also, Pearson [l9821 has developed a similar relationship between magnitude and seismic moments for microseismic events induced by forced fluid flow with observations from The Geysers geothermal field in northern California. This relation leads to higher expected level of microseismicity. Hence il is co~rcluded thal lhe observed displacements are not yelleraled seis~nically. This is confirmed by the lack of events observed for t,he period during which our dense seismic network was operat,ed. Hence it is concluded that the fault mot,inn is indeed aseismic and insensitive to the large water injection conducted during our experiment. These results are consistent with the fact that the continental vertical strike-slip Philippine Fault creeps at a rate of 3.5 cm yr-' along its northern Leyte 50-km segment [Duquesnoy, 19971.

5.2. Velocity Anomaly

'I'he injection exprrirrlerrt in well MG2RD and the injection prograrrls in the six other wells have induced >400 seismic events in a volume of rock estimated at 1.75 X 10d m3. Hence the volume of the induced seismic cloud coincides wit,h a volume in which the pore pressure has been incremed. The injection temperature was 25'C for the water used in MG2R.D and close to 15U°C for the hot brines injected in the other wells. The total injected volume has been estimated at 36,000 m3 for the MGZRD injection and at 327,000 f 20,000 m3 for all other injections combined together.

Tlle factors affecting seismic velocity include porosity, confining pressure, lithology, pore pressurc, saturation or phase transitions, and temperature [Nur, 19871. In an attemp to interpret rapid P wave velocity increase, only pore pressure, saturation, phase transitions, and temperature effects are discussed since no variation from porosity, lithology, or confining pressure are anticipated.

If t,he pore pressure increases in a saturated rock, so does t,hr pore volume and t,his induces a correlative decrease of the P wave velocity. This effect has been clearly illustrated by Wyllie et al. [l9581 r~sirrg experimental data. So it is anticipated thal 1.t1e flow of liquid water within a liquid saturateil 11c)dy would induce a decrease in P wave velocity. Yet an increase has bccn detected -1 km east of well MG2RD.

Regarding temperature effects caused by dilatation, P and S wave velocities increase when the tcrnperaturc decreases, but this velocity increase relrlilims smallcr than 5% for a lO0"L' temperature variation according to LiourbzP et al. [1986]. Thus the perturbation to be expected from the injection of water at 25'C (MG2RD) and brines at 160°C (others wells) should not have been larger than a few percent. On the contrary, Fehler [l9811 described a velocity decrease with a temperature decrease, but this effect was attributed to an increase in the microcrack porosity as a result of long-term heat extraction. Given the small temperature variation anticipated from the injections, this mechanismis not considered here.

At The Geysers geolherrrlal fields in California, low- velocity ratios were interpreted as low pore pressure and dry conditions, caused by the boiling of water as steam was extracted [Julian et al., 19961 It is proposed her? that the observed increase in velncit,y, in Tongonan, is linked with the invasion of liquid water in a nonliquid saturated rock body, i.e., within a pore volumc partially filled with vapor or gas. Some laboratory measurements [Domenico, 19761 havc pointed out nonlinear variations of P wave velocity with saturation. This phenomenon of imbibition (displacement of gas by waler), predicled by

13,610 PRTOUT, ET ;\L : 1KT)TJCED SEIS!vIICITY ACROSS PHILIPPIYE FAULT

the Biot-Gassmann-Domcnico thcory [Domcnzco, 19761, shows that Vp drastically increasrs (-302,) >\-hen saturation is within a range of 80% to 100%. First, C$ gradually decreases with increasing saturation because of the effective density increase, bnt when the full sat- uratiorr is r~early rerrched, i.b drastically iricl.eases because the pore fluid compressihility effect o~ltweigl~s the density effect. More recently, Le llaunlec and Gu4guenx [l9961 have modeled thrs effect as the result of two scale heterogeneities, one relabcd to the size of pores and cracks, the other one to the sizc of the heterogeneities of the fluid phase distribut,ion

The argument for considerirlg nonsat.i~rat,ed conditions, in this area of the geotherrnal reservoir, is three- fold.

First, thermodynamical condit,ions in the reservoir have been exarnined in order t o ascertain the possibility of having pore vnlrrme partially filled with liquid and vapor. For example, boreholr hIG21 D exihi l ,~ (,ern- perature and pressure values equal to 290°C and 7.4 MPa, respeciively, a t depth 1000 m bgs. This corm- sporlds to t l ~ r vapor-liquid water transition ('rable 2) [Lide, 19921. Unsaturated cor~dit ior~s call also be observed i ~ r tlle rescrvoir where temperrtture reaches 320°C a t -1000 m bgs. Moreover, the te~nperature profile in MG21D is characteristic of a convectiol~ process and indicates t l ~ a l higher trrrrperaturc cxists in the vicinity of the well. Hence we expect lerr1pi:raLures lligtr(:c than 300°C that would explain nonsaturated conditions a t depths >l000 m bgs. When we try to evaluate the weight of a water column given a dens~ty gradient related t o the temperature gradient on site: the results indicate tha t liquid pressure is close to 1.1 MPa for a temperatnre of 340°C a t 2000 nl bgs, it still corresponds t o partially saturat,ed conditions. This evaIrlat,ion seems realistic since pore pressure will not increase drastically if pore volume contains vapor.

Second, it has already been mentioned that the injec- tivity index has heen st,rongly reduced for well \'INRRD 2 nront,hs arter period I n.nd slighty for rvells MN2RD and MG21D. This may he related to the fact that saturat,ion occnrs in these urells anti tha t niaximumcapacit,y for injection is reached rapidly

'Third, let 11s assume that the prrtnrhecl volume of rock was initially unsaturated with a saturaiiorl within

Table 2. Vapor Prrssnrr-Trmperature Coriditiot~i of LVater [Lide, 19921

a range of 80% t o 100% to allow a nonlinear P wave vclocity increase. In t ha t case, the initial conditions must br close to fr~ll saturation, so that a small anioul~t of liquid woilld saturate the pore volume and would allow Lhe observed increase in pore pressure. We consider an initial sa tura t io~l equal lo 9 5 % II we eslirrrale l l ~ e volume of rock perturbed by the invasion of the injected liqnid (porosity of 6%), this would yield a volume eqnal to 1 . 3 4 ~ 10' 11i3. This computed volume is snlaller bhan the volume of rock in which induced seismicity has been observed but has the same order of n~agllit,odc. I t means that most of the inc111i:ed seismicit,) o r r l~ r r rd in a, vol- urne where ilie increase in pore pressure is allowed only after some vapor has been replaced by some liquid water.

Hence we propose t,he following n~echanism to interpret the observation. Before injection, a nonsaturated mrdinm prevailed wit,h a pore vol~lme (V,) partially filled with a liquid v o l ~ ~ m e (&) ancl a gas vapor volume (1:;) and with a pore pressure P,. Tlie saturalior~ was close t o 100%. After the beginning of Lhe iujeclior~s (injected volume V]), invasion and slight cooling oct:orri:d, lceding t o a rcduction of the volume of vapor followed l)y a pl~asc chengc (constant P , , C ' V

p - 1 +I$ + , \$+l is the volume of liquid due to the phase change). When saturation was fully reached. pore pressure increased, and a srrlall amount of liquid volurl~e (AMs) was absorbed by the cornpressihility of the liquid (5, = Vj + V, + V,,, - AKs). The main contribution of this volume effect is provided by the in- vasroil phenomenon. 'This evaluation is consistent with the order of magnitude that would induce a slgnificarit pertnrbation to the observcd travel tirnes. Wecause the injected water is distribnted nonuniformly close t o each well and the saturation and porosit,y will be heteroge- neous, we suggest tha t the perturbed volume may be even larger than that which has heen computed.

Accordingly, it seems reasonahle to interpret thc oh- srrvrd P wave velocit,y incrcase as indnred hy a derrea.sr i r r vapor i.ont,rnI.. I n Tongonan t,his soggest,~ t,hal, t,orno- graphic inversion of \~elocity could provide valuahle in- ibr~liations For iden1,ifying geothcrmal resources and for monitoring them during exploitation.

6. Conclusions

4 large water injection has beer1 undertaken in a well that intersects a creepirig segment of the Philippine Fault, a t the Tongonan geotherrnal field. Various other i r ~ j r c t , i o ~ ~ prop;ra,ms were iconductrd sirrrnlt,ar~rnr~sly in tlie viciuit,y of t,he well wit,]> larger injected vnh~mes but lower wellhead pressures. 'The MG2RD experiment has show11 that most of the water was injected 1.0 the easl of the Cerlt,ral Fauli Liue.

More t,han 400 i,vt:nt,s werc rrcorcled ill the vicinity of thc well and have been located through a simultaneous 3-D velocity-hypocenter inversion procedure. None of the induced microeathquakes is locatecl along the creep-

PRIOUL ET '\L.: INDUCED SEISMICITY ACROSS PIIILIF'FIKI: FAULT 13,611

iiig Central Fauit Line, and all of thern occrlrred below t,he casing shoe t o thr east of t,he fatilt line, i e , wilhin the geothermal res2rvoir. The increase i n wellhead pressure with associa,ted nricroscisrr~icity strongly suggests that pore pressure is the nlechanism responsi- ble for tlie observed induced srismicit,y for tile injection experiment.

The absenre of induced event,s along t,hr C!entraS Fault Line, combined with the lack of inicroseisniicity along the two central and west hrarrches of the fault over a broad range of t,inle; indicate an aseisntic behavior of the fault a t this location. These results have to be considered with the fact tha t the contitleiital vertical strike-slip Philippine Fault creeps a t a r ;~te of 3.5 cm yr-' along its northern Leyte 50-km segment.

The 3-D velocity rrrodrl irrlaged during injections. cwupared t o tha t obtained f i o r ~ , seisrrtir: ~rto~~ii.orirrgi:or~- ducted prior tu the injection experirr~er~t, shows a localized significant increase of P wave velocity. This anonraly is within tlie seislnicity cloud associated wiih the water injection. This effect is Interpreted as being the results of an increase in l iqu~d content. w~t,hin a liquid-vapor multiphase part of tire reservoir. Thcsc observations may reveal significant for constraini~rg t,he pore pressure in the vicinity of the fault.

Acknowledgments. The hydraulic stinlulation was sup- ported by PNOC (Philippine National Oil Company) and a grant from the Philippino Uepartment of Sciences and Tech- ~iolugy, wlliclr was oLl;ti~lrd thmks to R. Punongbnyan and E. Ranlos (PHIVOLCS). All t,hr srismir work has hpcn slip

ported by the French Ministry of Foreign Affairs and by PHIVOLCS within the framework of the French-Philippino cooperation. This project was iniliated by R. Gaulan (IPGP) and E. Barrier (UPMC) and benefitted from t,he supporl of H. Ferrer (PNOC). Most of the seismmo irlslrumenls were pru~ided by COPGS. Spccidl t l ld~~ks dre dut. LU J. Sal11 (EOPGS) for. the enlergency repairs of insiruments struck by lightning and for cfficicnt handling al the ielememcd network. We would like also tn thank E. Cnhallrs (PiiOC). D. Martinez (PFIIVOLCS), S. Luther JPIIIVOLCS) and F. Maneja (PNOC) for partzcipating in the data gathering and maintenance or the instruments. We also thank h1. Frogrreux (EOPGS) and A. Nercrssian (IPGP) for their help and expe- ricrlce in the preparation of a clean database and Tor inirinl data processing. This work benefitted from fiuitful cliscus- sions with T. Mossop. This is also IPGP contribntion l67?.

References

Alien. C.R., Circum-Pacific faulting in the Philippines-Taiwar region, J . Ceophys. Km.> 67, 4,795-4,811, 1962.

Aurrliu, M , Tecto~riquc du st.8~r~r1li c r ~ ~ t r a l dr la Faillr Philippine, Etude structurde, cin&matiqut e t kvolution gtuclynamiclue, l l~ tse 'le doctorat, Llniv. Paris 6; I'aris, 1992.

Bacolcol, T. C.; Etude d'un segment de la F'aille Philippine B partir de donnges GPS; stage de IIEA. liniv. Paris 6, I'aris, 1999.

Barrier, C.. M. Aurelicl, C. Muller, M. Pubellier, R. Que- bral, and C. Rangin. La Faille Philippinc : Un exernplr de grand dkcrochement nctif i I'arrikrc d'une zone dr subduction, C. R. Acad. Sci. Ser. 11 311, 151-188; 1990.

Barrier., E. , P. Huchon, and M. .\urelio, Philippine Fault: ..\ key to Philippine kinematics, Gtology, 19, 33-35, 1991.

Ik,r~ducuy, D., MG2RD cur~~p le l iu~~ l~.st I zpurt. trrhuical ie- port. PNOC - Energy Dev. Corp., hlakati City, Philip- pines. 1993.

nu~~rh ik , T., 0. COIISSY. and R. Zinszn(.r; A C O Z L S ~ Z ~ Z ( ~ de3 Miliczrz Poreua. Technip, Paris, 1986.

C:ornet; I'.H.; Fracture processes induced by forced Huid percolaliorr, Volcanic Seismology, 14 VCEI Proceedings in Volcanology; Vol. 3, edited by P,, Gasparini, R.. Scarpa and K . Aki, pp. 407-431. Springer-Vcrlug, Ncw York.

l Q0) .. Chrnet, T.II., and R. Jones. Field evidenc~ or, the oricnta-

t,ion of forced water flow with respect to the rrgional prin- cipal st.ress directions, Rock ,2lechar~ics-~l.foclels and Mra- rtrrements, edited by P.P.. Nelson and S.E.. La~rhach, pp. 61-71, A.A. Ralkema, Brookfield; VT., 1994.

Cornet., FIT., J . I-Ielrrl; H. Pt,il.l.er~aud, and A.. Etchecupat.. Seismic and aseismic slips incluced by large scale fluid ill- jections, P u r e Appl Geophys., 150, 563-583, 1997.

Delfin, F. .Jr. F. llaneja, D. Layogan, and 11. Zaidr- Delfirl. Stratigraphic and geophysical constrainrs on in- iection targets in the greater Tongonam geothernial field. Leyte, Philippines. technical reporl, PNOC - Energy Dev. Corp.. Makati City, Philippines, 1995.

Domenico, S.. Effect of brine-gas mixture on velocir,). in an ~~nronsnliclated fancl reservoir. Crolihy.~ii .s . { I , 882X94, 1976.

I)orhath. (:.. I). Oppenlirimel., F. Amehmg. and (4. l<irrg. Seismic tomagraphv and deformation rrlo<lcling of t,he

j~lnciion of the San r\ndreas m d Calnvrras f a ~ ~ l t s , .l. Gpo. phys. Rcs.. 101, 27,917-27,941, 1006.

Duquesr~oy, T., <>ontributiun di. la gCodesie B l'etude de grands dkcrochen~enis aclils assuciCs h drs ZUIIC~ de ~ u L - iluci,iu~i i cuuvelgrnce oblique, t,lllse de dnrtorat, Uiliv. Paris-Sud, Orsay, Prance. 1997.

Duquesrloy; T.. E. Harrier, 21. Kasscr, h1 At~r~l io . R . Ganlon, R . Punongbayan, C. Rangin, and the French- Philippine Cooperatior~ Team; Ilet,ect.ion of creep along the Philippine Fault: First result ot geodet~c measurements on Levte island. central Phili~uines. Ceovhus. lles. . . . "

Lel t . , 21, 972-978, 1994. Eberhart-Phillius. D.. Three-dimensional vclocitv structure

in northern California Coast R a n g ~ s from inversion of In- rnl rarrho~~nke arrival t,imps. Rull. .Sezsn~ol. .Soc. .4rn., 76. 1,025-i,osz. 1 9 s ~ .

Eherhart-l'hilli~s. U,. 'l'hrec-dimensional P arid S velocitv . . structure in the Coalinga region, California, J . Geophys. Res., 95, 15,313-15.363, 1990.

Eberhurt,-Phillips, D., Local carthquukc t,ornography: Earth- quake sorrrca regions, S e i s n ~ i c Turnography: Tl~eory and Practrce. edited by H.?II., Iyer, and I<., Hirahara, pp. 613- 6.13. Chnpman and Hall, New York, 1993.

Evar~s. J . , D. Eberhart-Ptlillips, and C. Thurher, User's manual for SIMULPSlZ for imaging I.; and V,/'/., a derivative of the i'hurber tomographic inversion S1ML~L:I for local earthcluakcs and explosions, LT.S. G ~ o l S u r v . Open File Rrp., 94-431. 142 pp., 1994.

Evans, J R . , and C. Achauer, Teleseismic v~lncity tomog- 1.i1pl8y ~ l r i~ jg /hr ACH met,hocl: Tllrory ancl application io continental-scale studies, Seidmic Tomography: Theory and Practice, edited by H.M., lycr, and K. . Hirahara. pp. 319-360, Chapman and Hall: New York. 1993.

Fehler. hl., Czhanges in P waw vplocity dnring operation of a hot dry rock geothermal svstern. J. Geophus. R c s . 86. . . 2.92.i3.938, 1981.

Fehler. M , Stress control of seismicity patterns ohserved ~ ~

cluri~lg hydraulic fracturing experiments at the Fenton Hill

13,612 PRIOUL E T AL.: INDUCED SEISMICITY ACROSS PHILIPPINE FAULT

hot dry rock geoth~rrnal energy site, New Mexico, Int. J . Rock Mech. Mzn. Sci., 26, 211-221, 1989.

Fitch, T. , Plate convergence, transcurrent taults, and in- ternal deformation adjacent to SE Asia and the wei1rt.n

Pacific, .l. C~0phy.5. Re.?.: 77, 4,432-4.460, 1972. Foulgrr, G., and A. Miller, Three-dimensional V, and

V,/V, structure of the Hengill Triple Junction and geothermal area, Iceland, and the repeatability of tomographic inversion, Geophys. Res. Lett.. 22, 1,309-1,312: 1995.

.Julian, B., A., Ross, G. Faulger, and J . Evans, Tln-er- dimensional seismic image of a geothermal reservoir: The Geysers. California, Geophys. Res. L e t t , 93, 685-688. 1996.

Kanamori, TT., and D. L. Anderson, Theoretical basis of some empirical relations in Seismology; Bull. Seismol. Soc. Am., 65, 1073-1095, 1975.

Kissling, E., W. Ellsworth, D. Chcrhart-Phillips, and U. K~.adolfer, Initial reference models in local earthquake tomography, J . Gcophys. Rcs., 99, 19,635-19,646; 1994.

Klein, F. , User's gl~ide 1.0 HYPOINVERSE; a program for vax computers to solve [or earthquake locations and magnitude, il..S. Geol. .Strnl. Open File Re,?, 89-,714, 49 pp., 19x9.

Le Ravalec, M,; and Y. Guegnen. Elastic wave velocities in oarliallv saturated rocks: Saturation hysteresis. . l Gro- jhys. 101. 837-841> 1996.

Lide. D., Handbook o f Chemistrv and Phvsics. 73rd ed.. CRC . , " ,

Press, Boca Raton, Fla.. 1992. Mossop. A. , Seismicity a t The Geysers geot,hrrmal held:

Correlations and mechanisms, Eus Trans., .4GU, 79 (17). Spring Meet Soppl., S227, 1998.

Narag, I . , B. ;LIarte, and A. Valerio, May 17, 1993 Tong- onan Leyte earthquakes-Preliminary Report of Investiga- tion, technical report, PHIVOLCS, Quezon City, Philip- pinta, 1993.

N I I ~ ; A., Seismic rack properties for rpsrrvoir descriptions and monitoring, Seismic Tomography- With Applications in Global Se i smolos~ and Exploration Geophysics, edited by G. Nolet, pp. 203-237, D. Keidel, Norwell, Mass., 1987.

Pearson, C.. The relationship between microseismicity and high pore pressures during hydraul~c st~mulatian exper-

iments in low permeability granitic rocks, J . Geophys. Res., 8G, 7,855-7,864, 1981.

Pcnrson, C., Parameters and a magnitude moment relationship fro,^ rrr~all earlhquakes observed doring hydraulic fracturing experiments in cristalline rocks, Geophys. Res. Lett., 9, 404-407, 1982.

Pine, R. and A. Batchelor; Downward migration of shearing in jointed rock during hydraulic injections, Int . J. Rock Mech. Mzn. Sct., 21, 149-263, 1984.

Scgdll, P., Earbhquakrs Lriggered by fluid extraction, Geol- ogy, 17, 942-946, 1989.

Tfiurher, C , Earthquake locations and three-dimensional crustal structure in the Coyote Lake area, ccnt,ral Califor- nia; J. Geophys. Rer., 88, 8,226-8,236, 1983.

Thurber, C., Local earthquake tomography: Velocities and V,/V, theory, Sezsmzc Tomogrnphy: Theory and Prac- ticc, edited l ~ y H . M . Iyrr-, and K., Hirahara, pp. 563-583, Chapman and Hall, New York, 1993.

Wyllic, M., A. Grcgory, and C . Gardner, An experimental investigation of fact,ors affert,ing elasiir wavp velocities in porous media, Geophysics, $3, 459-493, 1958.

F.H. Cornet and R. l'rioul, Dlipt. de Sismulugie, I ~ l a l i - tut de Physiquc du Clobe de Paris, 4 place Jussieu, F- 75252 Cedex 05 Paris, France ([email protected]; pri- [email protected])

C. Dorbath and L. Dorhath, Dkpt. de Sismologie, Institut de Physique du Globe de Strasbourg, 5 rue Rene Descartes, F-67084 Strasbourg cedex, France (cath0sismo.u-st,rasbg.fr; luuisQsisrr~o.u-strasbg.fr)

M. Ogena, Geoscient,ific Division, Ph~ l ipp~nc National Oil Compagny, Energy Development Corporation, Fort Donifa- cio. Makati City, Philippines ([email protected]

E. Ramos, Philippine Institute of Volcanology and Seis mology, C.P. Garcia Avenue, U.P. Diliman, Quezan City, Philippines (eramosQphivolcs.dost.~ov.ph).

(Received July 12, 1999; revised January 26; 2000; accepted February 9, 2000.)

an induced seismicity experiment across a creeping...

Documents