The October 1980 Earthquake Sequence Near the New Hebrides

GEOPHYSICALRESEARCH LETTERS, VOL. 10, NO. 12, PAGES1137-1140, DECEMBER1983

THE OCTOBER 1980 EARTHQUAKE SEQUENCE NEAR THE NEW HEBRIDES

John Vidale and Hiroo Kanamori

SeismologicalLaboratory, California Institute of Technology,Pasadena, California 91125

Abstract. Four large earthquakesoccurred in 1980 in a magnitudegreater than 6 in this region between 1963 and 1979. seismicgap near the Loyalty Islands in the New Hebrides. At However, since the sequence in 1980, three shocks of surface- 3:25 UT on October 24 an event with Ms -- 6.7 initiated the wave magnitude 6.6 or 6.7 have been reported (2/17/81, 9/17/81, sequence. Three events,Ms -- 6.7, 7.2, and 6.5, followedon the and 11/24/81) by NEIS. Also, an event of Ms = 7.3 occurredon next day. We investigatedthis sequenceby usingthe seismicity, 7/6/81 100 km southeastof the regionconsidered in this paper. first-motion, and waveform data and long-periodsurface waves. Another notable feature associatedwith this sequenceis that The first-motion data constrain one of each pair of nodal planes. the aftershockzone expandedduring the first week to an area 10 With this constraint,inversion of Rayleigh-and Love-wavespec- to 20 times larger than that for the first few hours. Some of the tra at 256 seconds determines the other nodal plane. The aftershocks occurred near the trench axis. A similar feature has mechanismsof all four events are almost pure thrust on a plane been noted for other eventsin the New Hebrides [e.g. Isackset dipping about 20 degreeseast and striking parallel to the local al., 1981]. Also, the aftershockzone during the first few hours strike of the New Hebridestrench. The first-dayaftershocks indi- became quiescentwithin six hours after the third mainshock. catean initial rupturezone of about2,000 km2, whichis con- Although the relationshipbetween seismicity and the state of sistentwith the estimatedseismic moment of 3 x 1027 dyne-cm. stressin the fault zone is not fully understood,recent studies indi- During the next two days, the aftershockactivity expandedto an cate that the spatio-temporalvariation of seismicitymay reflect area of 10,000to 20,000 km2 in the directionsboth alongand the mode of stressbuildup, release,and readjustmentin the focal perpendicularto the trench. Within 5 hours after the third and zone (see [Lay et al., 1982] for summary). In view of the impor- largestevent, the initial rupture zone had become mostly quies- tance of this sequenceas a gap-filling event, we made a detailed cent. Modeling of waveforms suggestsa body-wave moment of analysis of seismicity patterns preceding and following this between0.5 and 1.0x 1027dyne-cm and a sourceprocess time of sequenceand of the mechanismsof the mainshocks. 11 seconds. This pattern suggeststhat the initial rupture zone representsa zone of increasedstrength (i.e. an asperity),and the Seismicity stresschange due to failure of this asperitysubsequently migrated outward. During the two-year period before the main event, The temporal variation of the seismicityas well as the focal seismicityin the initial rupture zone was very low exceptnear the mechanisms,seismic moments, and magnitudesof the four larg- point where the first mainshockinitiated. A very tight clustering est earthquakesare shown in Table 1 and Figure 2. The loca- of activity occurredthere. This pattern indicatesgradual stress tions of the events are taken from the EDR listings(Earthquake concentrationnear the asperity which finally failed during the Data Reports, publishedby NEIS), and the nearestseismic sta- mainshock sequence. tions used are shownby the x's in Figure 1. The station coverage is good. Coudert et al. [1981] showthat the ISC (International Introduction SeismologicalCenter) locationsand locationsfrom a local seismic network agreeto within 10 km for the centralportion of the New In October, 1980, an earthquake sequenceconsisting of 4 Hebrides arc. We expectthe NEIS location errors to be lessthan large events(Ms -- 6.5, 6.7, 6.7, and 7.2; hereafterthese events 30 km. Engdahlet al. [1977] showedthat locationerrors of shal- are referred to as the mainshocks)and about 160 aftershocks, low events in subduction zones, if present, will tend to shift which were reportedby NEIS (National EarthquakeInformation apparent earthquakelocations toward the island-arc side of the Service),occurred in the New Hebridessubduction zone. Figure trench, so the existence of events on the oceansideof the trench 1 [McCann, 1980] showsthe plate tectonicsetting of the region. is probably real. The black square in the figure contains all the events in the Focal mechanisms of the four largest events are determined sequence. McCann [1980] assignedthe boxed area a seismic potential of 2, which means that the region had experiencedan earthquakeof Ms greater than 7.0 in the last 100 years but not 135'E 1,50' 165øE 180' 165øW the last 30 yearsand hasrelatively high seismicpotential. In the southern section of the New Hebrides subduction zone "3::•/"'•"•";• *qo PACIFIC earthquakeswith Ms of 7.0 to 7.5 appear to recur with at least __ •A '"'•--- • NEW 30-40 year intervals. Nine eventsof Ms = 6.5 to 7.5 occurredin =I 1 I Iø' the black square between 1943 and 1947 [relocationsfrom McCann, personal communication, 1981; the magnitudes are " "•?X':,.• •'•• 'I t • • 0 from Gutenbergand Richter, 1949], but none that largeoccurred from !947 to 1980. Becauseof the uncertaintiesin the magni- ':•?;• AUSTRALIAN tudes of old events, the overall magnitude of the activity from _ :.] PLATE • •o• 1943 to 1947 is somewhatuncertain [McCann, personal communication, 1982]. However,recently reevaluated magnitudes of these old events in the southern New Hebrides trench near the '•:.•...:.•'}".... .::;.• "" •W ' Loyalty Islandsare smallerthan the magnitudesgiven by Guten- bergand Richter [1949] by only 0.2 units on the average[Abe, • a;STATI• 1981]. Thus the activity in 1980 appearssimilar in magnitudeto that in the 1940's. The NOAA catalog lists no earthquakesof Fig. 1. The tectonic settingof the southwestPacific ocean. The New Hebrides subductionzone is part of the boundary betweenthe Pacific Copyright 1983 by the American Geophysical Union. plate and the Australian plate. The events describedin this paper occurred in the black square, near the Loyalty Islands, centered on Paper number 3L1226. 170 E and 22 S. The crosseslocate the stationsused by NEIS for epi- 0094-827 6/83 / 003L-1226503.00 center determination.

1137 1138 Vidale and Kanamori: Earthquake Sequence in the New Hebrides

TABLE 1. Parametersof the Four LargestEvents in the Sequence

Date Time Moment Latitude Lon•tude First Nod•Plane SecondNod•Plane

, (GMT) (in 1027dyne-cm) (degrees) (degrees) (Strike) (Dip) (Strike) (Dip) 10/24/80 3:25 .20 -21.989 170.165 321 20 141 70 10/25/80 7:00 .46 -21.982 170.025 316 30 136 60 10/25/80 11:00 2.00 -21.890 169.853 322 25 142 65 10/25/80 16:20 .10 -22.313 170.380 297 20 117 70 from long-periodWWSSN (WorldwideStandardized Seismograph thetic seismograms.The observedwaveforms are digitizedfrom Network), IDA (International Deployment of Accelerometers), film chips of long-periodWWSSN recordsfor COL (College, and SRO (SeismicResearch Observatories) and ASRO (Abbrevi- Alaska), KIP (Kipapa, Hawaii), LON (Longmire, Washington), ated SeismicResearch Observatories) records. First motions from and MSO (Missoula, Montana). The syntheticseismograms are WWSSN film chips constrain the more steeply dipping nodal computed from a simple dislocation model as described in plane. With this one nodal plane constrained,the amplitude and Kanamori and Stewart [1976]. The mechanism listed in Table 1 phase spectra of Rayleigh and Love waves at a period of 256 is used for the third event. A homogeneoushalf space is secondsobtained from IDA, SRO, and ASRO data are inverted assumed,and direct P, pP, and sP phasesare includedin the syn- by the methoddescribed in Kanamori and Given [1981]. thetics. We assumea point sourceat a depth of 15 km and usea The mechanismsof the four largestevents are indistinguish- symmetric trapezoidal time function with the rise time to and able from pure thrust (Figure 2). The first mainshockwith Ms = effectivewidth tl (i.e., total width to + tl; seeFigure 3). Figure 3 6.7 and seismicmoment of 2 x 1026 dyne-cm occurred at 3:25 shows the observedrecords as well as the syntheticsfor three UT on October 24, 1980, and was followed by the second sourcetime functionswith (to -- 3 sec,tl -- 8 sec),(4, 11), and (5, mainshockwith Ms -- 6.7 and seismicmoment of 4 x 1026 dyne- 14). The sour6etime function with (to -- 4 sec,h = 11 sec)fits cm at 7:00 UT on the next day. Figure 2d showsthe locationsof the observed records best. thesefirst two earthquakesand their aftershocks,which preceded The amplitudesof the recordsfor COL, KIP, LON, and MSO the third and largestevent. The third event occurredfour hours yield body-wavemoments of 0.56, 0.96, 0.72, and 0.44 x 1027 afterthe second,with Ms = 7.2 and momentof 2 x 1027 dyne- dyne-cm,respectively, with an averageof 0.6 x 1027 dyne-cm. A cm. Figure 2e showsthe activity from the third main event until second source of body-wave radiation with about one-third the 5 hours later. The fourth mainshock(with Ms = 6.5, shown in amplitude of the first pulse may be seenabout 50 secondsafter Figure2a) hasa smallermoment of 1 x 1026 dyne-era and may the first arrival on the observedrecords. The body-wavemoment be considered to be an aftershock of the third mainshock. The is about 30% of the 256 secondperiod surface-wavemoment of 2 distribution of events from 5 hours after the largestevent until x 1027 dyne-cm. Althoughthere is sometrade-off between the December 31, 1980, is shown in Figure 2f. All the events that depth, the shape of the sourcetime function, and the moment, occurredin the first week had depthslisted as lessthan 33 km. we will use the values obtained here (the effective width of the The sourcetime function and the body-wavemoment of the sourcetime function-- 11 sec,the seismicmoment = 0.6 x 1027 third event are estimatedby comparisonof the observedand syn- dyne-cm)as grossparameters of the body-wavesource.

(b) 36-]8 MONTHSPRIOR (C) PREVIOUS18 MONTHS

ß " '-'" * 4-49 '•'+L' .... .••o fo,•..,,. + • (' _•gJ.'_t\',, ß 5-5.•

i1•[ i1• i?r i•( I?O' I?l' i•J •O MM (f) 39-2400 HOURSRFTER

'%,,

life i?r L so K.

Fig. 2. Seismicityplots of the area coveredby the black squarein Figure 1. Starsindicate earthquakes; larger sym- bols indicatelarger events. The solid lines contourthe bathymetry(Mammerickx et al., 1974) in fathoms. Figure 2a showsthe hypocenters,mechanisms, surface-wave magnitudes, and seismicmoments for the four largestevents in the sequence.Figures 2b and 2f showactivity in 5 differenttime windows. The outer dotted line containsthe later aftershockzone and was derived from Figure 2f. The inner dotted line containsthe immediate aftershockzone and was drawn from Figure 2e. Vidale and Kmnamori: Earthquake Sequence in the New Hebrides 1139

Discussion MSO COL

The aftershockarea was about 2000 km2 immediatelyafter B the third event (Figure 2e) and expandedto fill an area of about 15,000km 2 in the followingweek. We excludeboth the area C , D , oceanward of the trench and the area so far to the east that the shallow aftershockscould not have been on the interplate boun- KIP b 3'0s LON,i:•. t,,:',l t

The spreadingpattern could indicate severaldifferent rupture A A mechanisms. The first possiblemechanism is that the immediate aftershockarea (the smallerarea definedby dotted lines in Figure B B 2), which is much smaller than the later aftershock area (the c c larger area defined in Figure 2), may be the entire extent of the O O coseismicrupture, and the later aftershockarea is in responseto Fig. 3. Body-wavesynthetics for 4 WWSSN stations. Trace A in a slow diffusion of the stresschange from the mainshocksout- each group is the observedseismogram. Traces B have a trapezoidal ward througha zone of viscousmaterial. The total moment Mo time functionof (to - 3 scc,t• - 8 scc)as illustratedschematically in of the threelargest events in the sequenceis 2.7 x 1027dyne-cm. the figure and as discussedin the text. Traces C and D have time If the coseismicslip took place over the immediate aftershock functionsof (4,11) and (5,14) area shownin Figure 2e, the amount of slip D would be D--Md•A-3.2m (1) Conclusion wheret• is the rigidity(here 5 x 10TM dyne/cm 2 is used)and A is the size of the immediate aftershock area, which is estimated to In our preferred model, the immediate aftershock area that be 2000 km2 from Figure2e. If the thrustplane between the broke in the first three events(Ms • 6.7, 6.7, 7.2) representsa Australian and Pacific plates had been completely locked during strongersection of the fault zone, where most of the preseismic the interseismicperiod and the interval is about 30 years, the stressaccumulation took place. In this sense,we call this area a convergencerate of 10 cm/yr [Duboiset al., 1977] would cause fault asperity[see e.g. Lay et al., 1982]. Within the asperity, accumulated slip of 3 m, which is approximately equal to the there is an even stronger region. When the strongestregion coseismicslip estimatedabove. breaks, body-wave and long-period surface-waveenergy is radi- As mentioned earlier, the averagebody-wave moment, 0.6 x ated. The rupture subsequentlypropagates into the rest of the 1027dyne-cm, is significantlysmaller than that obtained from the asperity,which then radiatesthe rest of the long-periodsurface- surface waves, suggestingthat the body-wave source is much wave energy but little body-waveenergy, except possiblyfor the smaller than the surface-wavesource. Although the size of the small arrival 50 secondsinto the record(visible in Figure 3). The body-wavesource cannot be determinedaccurately, the effective effect subsequentlypropagates outward, causingexpansion of the width of 11 sec indicates a source dimension of about 22 km if aftershockactivity. unilateral rupture with a velocity of 2 km/sec is assumed. This With this asperity model, we interpret the seismicitypatterns dimensionis considerablysmaller than that of the immediate aft- in the area beforethe mainshocks.Figure 2c showsthe activity ershockarea. If the rupture is bilateral or circular, the dimension during the 18-monthperiod beforethe mainshocks.Clustering of of the body-wavesource could be as large as 44 km, but it is still the events near the point of initial rupture (epicenterof the first considerablysmaller than that of the immediateaftershock area. event) seemsto indicate stressconcentration on one edge of the The result that the moment and the source area estimated asperity,perhaps in the strongestregion. During the 18-month from body-wavesare smallerthan thosefrom surfacewaves is not period prior to this period (Figure2b), no obviousseismicity pat- unusual, especially for large earthquakesin some subduction tern is found that may indicatethe existenceof the asperity. zones. Although this interpretation is not unique, the accuratedeter- An alternative possibility is that the entire later aftershock minations of seismic moments together with spatio-temporal area could have ruptured in the mainshocks,but the averagedis- mapping of seismicityprovide an important clue to the distribu- placementwas only about 30 cm insteadof 3 meters. If this is tion and the nature of asperitiesthat control the mode of stress the case,the absenceof earthquakesin the prior 35 years in the accumulationleading to a large earthquake. Detailed informa- black squarein Figure 1 requireseither that 9/10 of the slip has tion on the locationand the characterof an asperitysuch as the been aseismic or that much of the fault plane slips in several one indicatedby Figure2e and identificationof clusteringactivity earthquakesduring an earthquakecycle. suchas the one shownin Figure 2c are important for evaluating Although either one of thesecases or a combinationof them the seismicpotential of a gap. is possible,we considerthe first casemost satisfactoryin view of For long-term earthquakeprediction, areaswith the potential the good agreement between the amount of coseismicslip for large earthquakescan be identified on the basis of seismic estimatedfrom the seismicmoment and that predictedby plate gaps[McCann, 1980]. In the intermediateterm, the gapscould motion. We use this model in the rest of this paper. be monitored for large areas of quiescenceand clusteringof The other striking feature is the expansion of the zone of eventsto indicate areasthat might rupture and the point where quiescencenear the center, which grew to a radius of 50 km by rupture might initiate, respectively. the end of the first week. This feature can be seenin Figures 2d, 2e, and 2f, where a zone of quiescencedevelops in the area of the Acknowledgments.This work was partly supportedby USGS inner dotted line, which representsmost of the immediate aft- contract No. 14-08-0001-21223. We thank Jeff Given for pro- ershock zone. viding us with computerprograms. William McCann generously The development of quiescencein the immediate aftershock supplieda preprint of his paper, some of his data, and permission area may reflectthe differencein mechanicalproperties between to use a figure from his paper. Larry Ruff, Thorne Lay, and Jim the immediate aftershock area and the surrounding area. In Pechmann provided valuable advice. The figures were drafted terms of the mechanism proposed above, the immediate aft- skillfully by Laszlo Lenches. J.E.V. was supportedby a Guten- ershockarea is more brittle (undergoesless aseismic slip) than the berg Fellowship and an NSF Fellowship. The IDA data were surroundingarea, so that stressreadjustment after the mainshock providedby the IDA Projectteam at the Universityof California, may have been completedquickly, resulting in a shorterduration San Diego, and the SRO and ASRO data were provided by the of aftershockactivity. U.S. Geological Survey. Contribution 3856 of the Division of 1140 Vidale and Kanamori: Earthquake Sequence in the New Hebrides

Geological and Planetary Sciences,California Institute of Tech- Hebrides island arc, in Ewing Conf Symposiumon Earthquake nology, Pasadena,California 91125. Prediction, edited by D. W. Simpson and P. G. Richards, pp. 97-101, Am. Geophys. Union, Washington,D.C., 1981. References Kanamori, H., and J. W. Given, Use of long-periodsurface waves for rapid determinationof earthquake-sourceparameters, Phys. Earth Planet. Inter., 27, 8-31, 1981. Abe, K., Magnitudes of large shallow earthquakesfrom 1904 to 1980, Kanamori, K., and G. S. Stewart,Mode of strainrelease along the Gibbs Phys. Earth Planet. Inter., 27, 72-92, 1981. fracture zone, Mid-Atlantic ridge, Phys. Earth Planet. Inter., 11, Coudert, E., B. L. Isacks,M. Barazangi,R. Louat, R. Cardwell,A. Chen, 311-32, 1976. J. Dubois, G. Latham, and B. Pontoise, Spatial distribution and Lay, T., H. Kanamori, and L. Ruff, The asperitymodel and the nature mechanismsof earthquakesin the southern New Hebrides arc from of large subduction zone earthquakes, Earthquake Prediction a temporary land and ocean bottom seismic network and from Research,I, 35-103, 1982. worldwideobservations, J. Geophys.Res., 86, 5905-25, 1981. Mammerickx, J., T. E. Chase,S. M. Smith, and I. L. Taylor, Bathymetry Dubois, J., J. Dupont, A. Lapouille, and J. Recy, Lithosphericbulge and of the SouthPacific, Chart No. 12 of 21, ScrippsInst. of Oceanogra- thickening of the lithospherewith age: examplesin the Southwest phy, La Jolla, Calif., 1974. Pacific, in International Symposium on Geodynamics in the McCann, W. R., Large- and moderate-sizeearthquakes; their relationship SouthwestPacific, pp. 371-380, Technip, Pads, 1977. to the tectonicsof subduction,Ph.D. thesis, 194 pp., Columbia Engdahl, E. R., N. Sleep, and M. Lin, Plate effectsin northern Pacific Univ., New York, 1980. sulxluctionzones, Tectonophysics,37, 95-116, 1977. Gutenberg, B., and C. F. Richter, Seismicity of the Earth, Princeton Univ. Press,Princeton, 273 pp., 1949. Isacks, B., R. Cardwell, J. Chatelain, M. Barazangi, J. Marthelot, D. (Received June 28, 1983; Chinn, and R. Louat, Seismicit.vand tectonicsof the central New accepted July 19, 1983.)