GEOPHYSICALRESEARCH LETTERS, VOL. 13, NO. 13, PAGES 1426-1429, DECEMBER1986

OF TI-IE MAY 7, 1986 EARTHQUAKE SOURCE PARAMETERS

Lorraine J. Hwang a,nd Hiroo Kanamori

SeismologicalLaboratory, California Institute of Technology, Pasadena, CA 91125

Abstract. Source characteristics of the May 7, 1986 averageabout 80 yrs [Jacob,1984]. However, they are AndreanorIslands earthquake (51.412 ø N, 174.830øW, estimated to be as low as 50 yrs and may exceed 100 NEIC) are investigatedfrom WWSSN, GDSN and yrs [Sykes,et. al, 1981]. IDA records. First motions from over 60 stations The March 9, 1957 Great Aleutian earthquake, determine one steeply dipping nodal plane. We con- Mw--9.1 , located at 51.3øN, 175.8 øW, is one of the strained this nodal plane and inverted long-period sur- largest earthquakes in recorded history and occurred face waves at a period of T=256 sec and determined very close to this recent event. This earthquake rup- the second nodal plane to be dip 18 ø, rake 116 ø, and tured a 1200 km segment of the arc. The aftershock strike 257 ø. This shallowly dipping thrust mechanism sequence defines two segments of the rupture zone- is consistent with plate motions in this region. one west of Pass (172øW) and one east Seismic moment from surface-wave inversion is [Mogi, 1968]. Since the arc ruptured here just 29 1.3X102sdyne-cm corresponding to Mw----8.0. Ampli- years ago, seismic potential in this area has been rated tudes of body and surface waves from short-period very low [Sykes,1971; McCann, et. al, 1979;Sykes, et. instrumentsyield magnitudesof ff•b----6.8and Ms=7.7. al, 1981;Jacob, 1984]. The teleseismic average P-wave moment rate spec- The 1986 Andreanof Island earthquake occurred trum from 17 short- and intermediate-period instru- within the western aftershock zone of the 1957 Great ments is slightly lower than that of an average Aleutian earthquake. Portions of the western aft- Mw•8.0 subduction-zone event. We constrained the ershock zone have ruptured before in major events. fault plane as determined above to deconvolve the Kisslinger[1985] examined the seismicityin a region first 90 secsof the long-period body wave at 11 telese- between 175-178.8 øW called the Adak Seismic Zone. ismic stations to determine the source time function This zone ruptured in a sequenceof large earthquakes and the spatial distribution of moment release. The (Ms----7.8)between 1901-1905 and in 1957. No great source time function consists of 4 moment-releasing earthquakes have occurred in this region until 1986. episodes which have a total moment release of The 1986 earthquake is unique because of the unex- 9.4X1027 dyne-cm. The fault rupturedbilaterally pectedly short recurrence time and the partial overlap with the largest moment releasing subevent occurring of its rupture zone with the 1957 earthquake. To between 30-45 sec. This subevent nucleates approxi- better understand this earthquake and its relationship mately 75-90 km west of the determined epicenter. to the 1957 earthquake, we examine the source This region correspondsto the epicentral area of the characteristics of the May 7, 1986 earthquake over a 1957 Great Aleutian earthquake which is one of the. broad periodrange (1-256 sec). largest earthquakes in recorded history.

Introduction 4.0-5.0 + 5.0 -6.0 x The May 7, 1986 Andreanor Islands earthquake 6.0-70 • 7.0-8.0 • (22h 47m 10.2s UTC, 51.412øN, 174.830øW, h--33 --- ACTIVITY OF MAY 7-8, 1986 55N km, Ms=7.7 , NEIC) is the largest event that has + • ,• occurred in this portion of the Aleutian arc since the 1957 Great Aleutian earthquake. The earthquake was felt on Atka approximately 65 km north of the epi- " •AT K center and on Adak 130 km northwest and caused damageon both islandsbut no fatalities(Figure 1). 2000-%•__•• The Aleutian arc is one of the most seismically active regionsin the world, and has generatedseveral of the largest earthquakes in recorded history. These large earthquakes are a part of an earthquake sequencethat ruptured most of the -Aleutian _ • -- - ,. / - ••• + arc during the period 1938 to 1965. Areas that did 6000 -• / " MAY7, 1986 not rupture during this period of increased activity • • Ms= 7.7 50 N 600• •• have been identified as seismic gaps. These seismic + I00 km gaps, the Yakataga., Shumagin, Unalaska and Kom- mandorski gaps have not broken in at least 80 yrs. 179 W 172 W Recurrence times of great earthquakes in this region Figure 1. Map of a portion of the showing earthquake activity during May 1986 from Copyright 1986 by the American Geophysical Union. the PDE catalog. The May 7, 1986 earthquake is plotted as a filled star. The aftershock area within 24 Paper number 6L7064. hrs following the main event is outlined in dashed 0094-8276 / 8 6 / O06L-7 064 $ 03. O0 lines.

1426 Hwang and Kanamori' The 1986 Andreanor Islands Earthquake Source 1427

IO9 (describedbelow) determined the secondnodal plane to be dip 18 ø, rake 116 ø and strike 257 ø. This thrust mechanism is consistent with the orientation of the plate boundary and direction of plate motion in IO8 this region[Minster and Jordan,1978].

DIP:18RAKE:112 STRIKE:252 Seismic Moment and Rupture Area • 107 Seismic moment, Mo, is determined from long- period surface waves from International Deployment of Accelerometers Network and GDSN stations follow- • •o• & ing Kanamoriand Given [1981].Amplitude and phase spectral data at 256 sec are inverted using excitation functions for a source depth of 16 km. Since the source depth is less than 30 km, two elements of the 1o5 seismic moment tensor cannot be resolved. We there- fore use a double-couplesource constrainingone plane from the first-motion data. The source time delay is 1o4 varied from 10 to 80 sec. Using 48 Rayleigh waves, ß , I0-2 10-1 100 o DILATATION Frequency(Hz) R2-R4, the best fitting solution occurs at a source x NODAL a b time delay of 60 sec which yields Mo----1.3X102s Figure 2. (a) Focal mechanismfor the May 7, 1986 dyne-cm, Mw----8.0. Adding 12 Love waves, G2-G4, earthquake. Sixty-seven stations are used to constrain gives approximately the same seismic moment. Since the steeply dipping nodal plane. Large symbols are this is a very large event, directivity of the good quality picks and small symbols poor quality. propagating dislocation causesthe source time delay (b) Averagemoment rate spectrafor the May 7, 1986 to be azimuthally dependent. Assuminga simple rup- earthquake. Theoreticalspectra for an w-2 model are ture propagating a distance of 150 km to the west at shown by thin lines. The dashed line shows the aver- 3 km/sec and consideringazimuth to the receiver, age spectrum from 7 events for an Mw:8.0 subduc- delay times are computed at each station. Using tion zone event. these times in the inversion doesnot significantly alter the solution. For simplicity, here we use a constant source delay time of 60 sec. Aftershock Area The aftershock data within the first 24 hrs of the mainshock suggest an immediate rupture zone of The 1986 Andreanof Islands earthquake was pre- approximately 220 km in length and 65 km in width. ceded by several foreshocks,the largest, Ms----6.0, Assuming rigidity is /•----5X10TM dyne-cm -2, the occurring a little over two hours before the main estimated slip during this event is 180 cm correspond- event. Numerous aftershocks followed over the next ing to a slip rate of 6.3 cm/yr since1957. This esti- severaldays (Figure 1). The aftershockarea extends mate is somewhat lower than the average plate over a 220 km segment of the are. The western extent motioncalculated by Minsterand Jordan[1978]. The of the aftershock zone is bounded by Adak Canyon. Minster and Jordan[1978] slip rate, 8.1 cm/yr, indi- The eastern extent coincideswith the eastern edge of cates a deficiency of slip of approximately 53 cm. the western aftershock zone of the 1957 event, This deficiency can be substantially reduced if ase- strengthening the argument for a structural discon- ismic slip is occurring(1.8 cm/yr) or rigidity of the tinuity (or barrier) here as proposedby Mogi [1968]. fault plane surface is closer to crustal values. The aftershock zone grew northward toward Atka within the following weeks (Figure 1). Ensuing activity has been primarily restricted to within the above aftershock zone. Kisslinger[1985] forecasted an earthquakemeasur- ing Ms>_7 to occur in late 1985 in the immediate ,,.•-'',•, :/.j •..•• ,,:•o vicinity of Adak Island and to rupture soon after into Adak Canyon. The epicentral region of this predicted ': V_ event is over 150 km from the epicenter of the 1986 PMG earthquake. Although the immediate aftershock / / \,.•o --•,, sequenceextends into the Adak Seismic Zone, it did not break into the Adak Canyon sub-region.

Focal Mechanism ,.40 ---: v 'C' ' '9' '•' Figure 3. Comparisonof actual (solid lines) and syn- The focal mechanism for the main event was deter- thetic (dashed lines) seismogramsfrom body wave mined from 67 World-Wide Standardized Seismo- inversion. ?eak amplitudes for the data in cm is graphNetwork (WWSSN) and GlobalDigital Seismo- given below each station for a long period WWSSN graph Network (GDSN)stations. The first-motion instrument with a magnification of 1500, The focal data fix one of the nodal planesreasonably well (Fig- mechanism and stations used in the inversion are ure 2a). Inversion of long-period surface waves shown. 1428 Hwang and Kanamori' The 1986 Andreanof Islands Earthquake Source

a--0.34 and a--0.32 respectively. The average 3.30 x 1026 moment rate spectrum computed from 17 teleseismic P-wave seismograms from short- and intermediate- dyne-cm. period vertical componentGDSN instrumentsusing the method of Houston and Kanamori [1986] is slightly lower than the averageMw--8.0 subduction- I i I i ' zoneevent (Figure 2b). 0 15 30 45 610 75 90 TIME (sec) Rupture Pattern Figure 4. Source time function from the simultaneous body wave deconvolution. Peak moment release rate The source time function is determined by simul- is 3.30X 10•6 dyne-cm/sec. taneous inversion of the first 90 sec of the body wave from 6 long-period WWSSN and 5 intermediate-period GDSN vertical component instruments following Magnitudeand SourceSpectra Kikuchi and Fukao [1985].We use the focal mechan- ism as determined by the surface-wave inversion and Amplitudesof body and surfacewaves from short- extend the fault-plane surface discussed above from periodvertical componentWWSSN and GDSN instru- the aftershock data up to the trench axis. Figure 3 mentsyield a magnitudefrom Houstonand Kanamori shows the best fitting synthetics compared to the [1986]of rhb--6.8(40 stationaverage) and an Ms--7.7 data. The source time function for these synthetics is (33 station average)with a standarddeviation of shown in Figure 4. For all models considered, the

TIME (SEC)' O- 15 TIME(SEC)' 16- 30

60 N

30 :• z

0 m

_

--30 >.

- (9

' -60•.

v -90 S

ß .. - ...... i i i i i i i i i i i i i i i i i i i i i i i i

TIME (SEC)' 31- 45 TIME (SEC)' 46- 90

-60N _ _30:• - 0 "'u

-30>.

ß .. - --60 •. ß . .

v . --90 S

ß . .

...... - - ...... -. -1'50''1'20'-9'0 '-6'0 '-3'0 ' 6' '" 3b ' 6b ' -1'50i -i20i i '-90'i 'i '-60"i i -3'0 i 6 i 3b i 6b i W ALONG STRIKE DISTANCE (KM) E W ALONG STRIKE DISTANCE (KM) E Figure 5. Fifteen secondtime slicesof the spatial distributionof momentrelease pro- jected onto the fault surface. The radiusof each circleis proportionalto the seismic momentof the point sourceit represents.Shading is proportionalto the amountof moment releasein a given area and is normalizedto the maximumvalue in each time slice. The asterisk marks the hypocenter. Hwangand Kanamori:The 1986Andreanor Islands Earthquake Source 1429 general features of the source time function are stable. Division of Geological and Planetary Sciences,Califor- The first moment-releasing episode is of lower ampli- nia Institute of Technology, Pasadena, CA 91125. tude and duration than the second and largest episode. Moment release then falls off with several References smaller episodes occurring in the remaining 45 sec. Total moment release for this model is Anderson, D.L., Accelerated plate tectonics, Science, Mo--9.4X 1027dyne-cm. 187, 1077-1079, 1985. With our distribution of stations, the rupture pat- Houston, H. and H. Kanamori, Sourcespectra of great tern (Figure 5) is resolvedbetter along strike than earthquakes: teleseismic constraints on rupture perpendicular to strike. The first subevent occurs processand strong motion, Bull. Seism. Soc. Am., near the epicenter. The rupture area expands and 76, 19-42, 1986. moment release culminates in the second subevent Jacob, K.H., Estimates of long-term probabilities for between 30-45 sec. During this largest subevent, the great earthquakes in the Aleutians, Geophys.Res. point sources cluster about 75-90 km west of the epi- Lett., 11, 295-298, 1984. center. Remaining activity is diffuse. Kanamori, H. and J.W. Given, Use of long-period sur- The areas of the first two major subevents can be face waves for rapid determination of earthquake thought of as asperities along the fault surface. The source parameters, Phys. Earth Planet. Inter., 27, first area is associated with the initial failure of the 8-31, 1981. fault surface which perhaps triggered the failure of the Kikuchi, M. and Y. Fukao, Iterative aleconvolutionof second. Within the resolution of the data, the second complex body waves from great earthquakes- the and largest subevent corresponds to the epicenter of Tokachi-Oki earthquake of 1968, Phys. Earth. the 1957 earthquake. Planet. Inter., 37, 235-248, 1985. Kisslinger, C., Seismicity patterns in the Adak Seismic Conclusion Zone and the short term outlook for a major earthquake, in Minutes of the National Earth- The 1986 Andreanof Islands earthquake is the larg- quake Prediction Council, September 8 F• 9, est event to occur within the Islands since 1957. This Anchorage, Alaska, edited by C.F. Shearer, U.S. great event, Mo--1.3X102sdyne-cm, Mw--8.0 , rup- Geological Survey Open File Report, 86-92, 120- tured only 20% of the arc which failed in 1957. The 134, 1985. end points of the aftershock region are well defined by Li, V.C. and C. Kisslinger, Stresstransfer and non- Adak Canyon to the west and the eastern terminus of linear stress accumulation of subduction-type the western aftershock zone of 1957 to the east. Body plate boundaries- applicationsto the Aleutians, wave modeling shows that this region is composed of Pure Appl. Geophys.,122, 812-830,1984/85. 2 strong asperities, one which is associated with the Lyzenga, G.A., A. Raefsky and B.H. Hager, Time- 1986 epicenter and the other with the 1957 epicenter. predictable earthquake recurrenceat subduction The effect of rupturing of these two asperities within zones?(abstrac't,), EOS Trans. A GU, in press, such a small portion of the 1957 aftershock zone is 1986. unclear.Li and Kisslinger[1984/85] find that rupture McCann, W.R., S.P. Nishenko, L.R. Sykes and J. in adjacent fault segments can accelerate the loading Krause, Seismic gaps and plate tectonics: Seismic rate and cause a coseismicstress jump in neighboring potential for major plate boundaries, Pure Appl. segments. This possibly triggers rupture of neighbor- Geophys.,117, 1082-1147, 1979. ing segments if it is at a high stress state. Hence, Minster, J.B. and T.H. Jordan, Present-day plate failure of adjacent segments due to accelerated load- motions, J. Geophys.Res., 83, 5331-5354, 1978. ing from the 1986 earthquake depends on several Mogi, K., Development of aftershock areas of great poorly constrained parameters: current stress state, earthquakes, Bull. Earthquake Res. Inst. Tokyo loading rates, and failure strengths. Further model- Univ., •6, 175-203, 1968. ing of both stresstransfer along plate boundariesand Sykes, L.R., Aftershock zones of great earthquakes, possible accelerated plate motions following great seismicity gaps, and earthquake predication for earthquakes[Anderson, 1985; Lyzenga, et al, 1.986] Alaska and the Aleutians, J. Geophys.Res., 76, must be considered to further understand plate 8021-8041, 1971. interaction and seismic hazard in this region. Sykes, L.R., J.B. Kisslinger, L. House, J.N. Davies and K.H. Jacob, Rupture zones and repeat times of Acknowledgement. We thank the personnel from great earthquakes along the Alaska-Aleutian arc, the WWSSN stations, Project IDA at the Institute for 1784-1980, in Earthquake Prediction, An Interna- Geophysicsand Planetary Physics at University of tional Review, Maurice Ewing Series, •, edited by California, San Diego and the U.S. GeologicalSurvey D.W. Simpson and P.G. Richards, pp. 73-80, for making their data available to us. H. Houston 1981. helped us with the computationof the sourcespectra. Thoughtful reviews by L. Astiz and H. Houston were greatly appreciated. This researchwas supportedby USGS Grant No. 14-08-0001-Gl170 and partially by (ReceivedOctober 6, 1986; an NSF Graduate Fellowship. Contribution No. 4407, acceptedOctober 22, 1986.)