GEOPHYSICAL
RESEARCH
LETTERS,
VOL. 16, NO. 12, PAGES
1425-1428, DECEMBER
1989
TELESEISMIC
BODYWAVEANALYSIS
OFTHE1988ARMENIAN
EARTHQUAKE
J.F. Pachecol,
2, C. H. Estabrookl,
2,D. W. Simpson
2 andJ.L. N•b•lek3
Abstract.Long-periodandbroadband
bodywavesfrom 14
digital seismicstationsare usedto investigatethe rupture
processof the December7, 1988 earthquakenear Spitak,
Armenia,USSR. The inversionof thesedata gives the
followingcentroidal
source
parameters:
strike299ø,dip64ø,
rake 151ø, depth6.3 km and seismicmoment1.5x1019Nm,
indicatingthaton averagethe earthquake
had a strike-slip
mechanismwith a substantialreverse component.The
broadband waveforms, however, show significant
50 ø
i'........."
.--.GC
.,!-•,".<•
-,
30 ø
complexity;they are bestfit with a sourcemodelthatincludes
threesub-events,very similarin size, but with distinctfocal
mechanisms
andlocations.Ruptureapparentlyinitiatedasa
shallowreversefault at a point of maximumbendingon a
right-lateralstrike-slipfault, andthenextendedbilaterally,
'
40 ø
20 ø
60 ø
40 ø
41.0
first towards the southeast and then towards the west. This
interpretationagreeswith the aftershockdistributionandfault
lineations
observed
onLANDSAT images.
Introduction
40.9
40.8
Focal parametersof the devastatingearthquakeof
December
7, 1988nearSpitak,Armenia,USSR,reported
by
the National EarthquakeInformationCenter (NEIC) are:
40.7
43.8
43.9
44.0
44.1
44.2
44.3
44.4
44.5
Ms=6.8, origin time 07h 41m 24.9s, latitude 40.987ø N,
longitude44.185ø E, and 10 km depth.Extensivecasualties
anddamageoccurredat thetownof Spitak,just southof the
epicenter,and at the city of Leninakan,nearthe westernend
of the aftershock
zone(Figure1). The rupturereachedthe
Fig. 1. Top figure shows NEIC location of the 1988
surfacesouthof Spitakalongan approximately
8-km-long,
west-northwest
trendingfault segment(Figure 1). The
displacements
alongthesurfacetraceshowreversefaulting
steeply-dippingto the north with a large right-lateral
component:
themaximummeasured
slipwasvertically1.6m
.andlaterally0.5 m [Sharp,1989].TP-:aftershock
epicenters
[Simpsonet al., 1989] appearto divide into two groups,
probablydelineatingat leasttwo fault segments
(Figure1):
the major segment,orientedroughly east-west,extends
LC = Lesser Caucasus,GC = Greater Caucasus,AP =
westwardfrom the epicenterof the mainshock;a second
segment extends from the mainshock epicenter to the
southeast.The hypocentersextendto a depthof 15 km only
in thewest,but mostoccurat a depthshallowerthan 10 km.
The earthquakeoccurredalong the southernedge of the
LesserCaucasus.In the Lesserand GreaterCaucasus,major
structurestrend west-northwestand fault plane solutions
indicateprimarily reversefaulting [Jacksonand McKenzie,
1984]. However, two moderate events (Mw = 5.6 and 5.8)
that occurred in 1978 and 1986 within 50 km of the Armenian
Armenianearthquake
(star).Main plateboundaries
andfaults
aredrawnon themap assolidanddashedlines.Shadedareas
areelevations
higherthan2 km. Largearrowshowsdirection
of motionof the Arabianpeninsula
with respectto Eurasia.
Arabianplate,BM = Bitlismassif,ZM = Zagrosmountains,
T = Turkey.Bottomfigureshowspreliminary
epicenters
of
aftershocks
recorded
by theU.S. Geol.SurveyandLamontDohertyGeol.Obs.(qualitya andb locations).
Epicenter
of
maineventreported
byNEIC is indicated
bythestar,triangle
showsthelocationfrom 19 localSovietstations[Cisternas
et
al., 1989].Symbolsizeis scaledto eventmagnitude,
with
magnitudesrangingbetween1 and 4. Dashedline indicates
mappedsurfacerupture.Thin solidlinesaremainroads.
Bitlis massifandZagrosmountainsto the southandalongthe
Lesser and Greater Caucasusto the north. The region
between the Lesser Caucasus and the Bitlis massif is an
elevatedvolcanicplateauwith Quaternaryvolcanicedifices
alignedin a north-southdirection,suggesting
an almosteastwestorientationof theprincipaltensionalaxis [Kazminet al,
1986].The deformation
hereis characterized
by dispersed
earthquakeepicenter, show strike-slipmechanismswith Paxesorientednorth-south[Dziewonski et al., 1987a, 1987b].
The regional tectonicsis characterizedby north-south
compression
as the Arabianplateindentsthe Eurasianplate
with a convergence
velocityof 30 mm/yr acrossnorthernIran
(Figure 1). This collisionresultsin the lateralescapeof the
TurkishandIranianplatesandmajorthrustzones,alongthe
conjugatestrike-slipfaults[e.g.,$eng0ret al., 1985].
Seismicitystudiesof the convergence
zone[Jacksonand
McKenzie, 1984, 1988;AmbraseysandAdams,1989]reveal
thatonlya smallpercentage
of thedeformation
takesplacein
earthquakes.
Large (Ms > 7) historicalearthquakes
in the
regionarescarce,
although
anearthquake
catalogcompiledby
Ambraseysand Adams[1989] showsseveraldevastating
earthquakes
with magnitudes
lessthan7.
Long-periodand broadbandbody wavesrecordedat 14
1Deparm•ent
ofGeological
Sciences
of Columbia
University. digital stationsfrom the Global Digital SeismicNetwork
(GDSN) and the IncorporatedResearchInstitutions for
2Lamont-Doherty
Geological
Observatory.
Seismology
(IRIS) networkareusedhereto studytherupture
3College
ofOceanography,
Oregon
State
University.
process
of theArmenianearthquake.
Data andData Analysis
Copyright 1989 by the American Geophysical Union.
The analysis of the mainshockfollows procedures
Paper number 89GL03366.
0094-8276/89/89GL-03366
describedby Ngb61ek [1984]. We use the P, SH and SV
$03 . 00
wavesrecordedby GDSN and IRIS stations,takingonly
1425
1426
Pachecoet al.: Teleseismic
Analysisof the ArmenianEarthquake
COL 5'
TABLE1.Crustal
Structure
UsedinComputing
Theoretical
Seismograms
Thickness Vp
Vs
Density
km/s
km/s
g/cm3
8.10
4.68
3.30
Receiver half-space 6.00
3.46
2.75
Source
km
1
5
30
4.00
5.60
6.50
half-space
2.14
3.23
3.75
TOL 285'
SLR 196'
CHTO 98'
2.35
2.70
2.85
stationsin a distancerange between30ø and 83ø from the
epicenterto avoid strong,regionallyvariable,uppermantle
arrivalsand corephases.The crustalstructure(Table 1) we
assumedin computingtheoreticalseismograms
is similarto
thatusedby Simpsonet al. [1989] to locatethe aftershocks.
The seismograms
areinvertedin a least-squares
sensefor the
source model parameters. For long-period data and
intermediatemagnitudeearthquakessuchas the Armenian
earthquake,
a singlepointsourcemodelusuallyprovidesan
adequate
sourcedescription.In thatcase,theinversionyields
theaverage(centroidal)parameters
includingthefaultplane
solution(strike, dip, and rake of the slip direction), depth,
seismicmoment and time function. If requiredby the data,
additionalpoint sourcesseparatedin spaceand time can be
introduced,revealingmore detail aboutthe ruptureprocess.
In this study,we startedwith the long-periodseismograms
to
determinethe averagepropertiesof the source,and then
refinedthe modelusingthe broadbandrecords.
Inversion Results
sec
Fig. 2. Observed(solid) and theoretical(dashed)broadband
seismogramsfor different source models from several
azimuthallydistributedstations.Shownare seismograms
for
Alaska (COL); Spain (TOL); Thailand (CHTO); and South
Africa (SLR). The numbernext to stationname is its azimuth
from the source. 1S shows the effect of first subevent(see
Table 2 for source parameters); 2S shows effects of
subevents1 and 2; 3S shows combined effect of 3 subevents
(our final model); and BB-PS for broadbandpoint source
model.
individualpoint sourcesinto our model,eachhavingits own
mechanism,time functionanddepth. The two pointsources
thatdescribethe laterpart of the ruptureare allowedto have
arbitrarytime delay and locationwith respectto the first. In
the subsequentdiscussion,we will refer to the three point
sourcesas subevents,althoughtheir contributionsin mostof
thewaveformsoverlapconsiderably.
The parametersof the three subeventsare summarizedin
Table 2 andthe matchesto the completedatasetareshownin
Figure 3. The initial eventexhibitsprimarilyreversemotion
with a small right-lateral strike-slip component. Its
mechanismis in good agreementwith the mappedsurface
rupture.The nodal plane correspondingto the fault plane,
based on field observations,strikes 299ø and dips 45 ø
northward.The meandepthof this part of the ruptureis 3.1
The results of a series of inversions we performed are
summarizedin Table 2. The long-periodmodel (modelLPPS) indicatesthat,on average,theArmenianearthquake
hada
strike-slip mechanism with a large reverse-faulting
component.The mean (centroid)depth is about 6 km, the
seismicmomentis 1.4x1019Nm, and the sourcedurationis
about18 s. When we invert the broadbanddatausinga single
point source,we obtain essentiallyidenticalresults(model
BB-PS), however, the model does not provide adequate
matchto thedata(Figure2), indicatingthatthe sourceprocess
wasmore complicated.The main featuresof the datathat the
singlepointsourcemodelcannotadequately
reproducearethe
amplitudes in the early portion of the waveforms (e.g.,
compareBB-PS and 3S for COL in Figure 2) and the large
amplitudearrival about10 s after the first motion(e.g., SLR
andCHTO). We haveattemptedto explainthe complexityof
azimuth of 153ø relative to the first subevent(nucleation
point). This subeventruptureda nearly pure right-lateral
strike-slipfault striking 309ø and dipping 90ø, using the
aftershockdistribution(Figure 1) as a guidein choosingthe
faultplane.The meandepthis 7.2 km andthe sourceduration
is 9 s. The third subeventhasa strike-slipmechanismwith a
substantialreversecomponentand is located38 km west of
the broadband waveforms
thenucleation
point.The seismicmomentis 5.1x10•8 Nm
in terms of the effects of crustal
structure(Moho, low velocity layers),but sucheffectsfail to
havecorrectpolarities,sufficientamplitudesandtherequired
arrivaltimesto fit thedata simultaneously
at all the stations.
To explain thesefeatures,we were forced to subdividethe
ruptureinto three phases.We do this by introducingthree
km andthe seismicmomentis 5.9x1018Nm. The rupture
builds up gradually and lasts for almost 9 s. The second
subevent with a moment of 7.5x10 •8 Nm locates 17 km at an
andthe centroiddepthis 7.3 km.
Using a moment tensor summation, the combined
mechanismof the three subevents(BB-SUM, Table 2) is a
puredoublecouple,in goodagreement
with thatobtainedby
invertingthe data with a singlepoint source(BB-PS, Table
TABLE2. Parameters
of theSource
ModelsDerivedby theInversion
of BodyWaves
Moment
1018Nm
LP-PS
BB-PS
BB-SUM
1st subevent
2nd subevent
3rd subevent
Depth
km
Strike
deg
Dip
deg
Rake
deg
Duration Delay
s
s
Distance Azimuth
km
deg
14.4
14.8
6.3*
6.3
307
299
60
64
159
151
18
17
-
-
-
14.1
5.9
7.5
5.1
6.0
3.1
7.2
7.3
293
299
309
264
67
45
90
69
141
127
150
148
18
9
9
8
3.8
9.3
17
38
153
277
The formaluncertainties
(2(•) are: depth0.4 km, orientation4ø, duration0.5 s, delay0.5 s, distance4 km, azimuth5ø, and
moment10%. LP-PS = Long-periodpointsource,BB-PS= Broadband
pointsource,BB-SUM = bestdoublecouplefrom
sumof subeventmomenttensors,"*" = depthfixed.Distanceandazimuthof thecentroidsof thesecondandthirdsubevents
aremeasured
with respectto thenucleationpoint(epicenter)of the firstsubevent.
Pacheco et al.' Teleseismic Analysis of the Armenian Earthquake
P waves
..-.
'
SC•--""--•4•'; •:•T" WMQB
2.65 •
....
0 10 20
0
sec
1S
0,001
0 •
SeC
•
SeC
SH waves
c
SCP
•
BJI
wa
0
. .
• •.88]
o.oo,[•
10 20
sec
the surfacerupturesouthof Spitakrepresents
only aboutone
third of the momentreleasedduringthe earthquake,the rest
occurredon faultsat depth[seealsoCisternaset al., 1989;
Stein and Yeats, 1989].
The predominantlystrike-slipmechanisms
of the Armenian
earthquakeand the pfior smaller earthquakesin the region
indicatethat the strike-slipmode of deformationin eastern
TurkeyandwesternIran extendsat leastpartlyinto theLesser
Caucasusas the matefial is squeezedlaterally from the vise
formedby the north-southcollisionof Arabia with Eurasia.
To explain the broadbanddata, we subdividedthe model
into three subeventscorrespondingto three stagesof the
rupture process. The mechanism of the first subevent
indicates
thattheruptureinitiatedon a reversefault.The fault
strikeandthe senseof slipfor the first subeventareconsistent
with the surfacebreakmappedsouthof Spitak.The shallower
depth of this subeventwith respectto the other two, and
perhapsalsothe differencein mechanism,may explainwhy
onlythispartof therupturereachedthe surface.The seismic
momentestimatedfrom the field measurements
(thevertical
slipof 1.6m; lateralslipof 0.5 m; faultlengthof 8 km; and
widthof 8.8 km, basedon the seismologically
determined
fault dip and centroiddepthof 3.1 km) is 3.5x1018Nm,
..,•
• 8.•
o.•] •"...
1427
0 20 •
sec
SV waves
which is somewhat smaller than the moment we deducedfor
the first subevent.This indicatesthat the faulting
corresponding
to the first subeventextendedbeyondthe
surface
faultbreak,probably
westward,
eventually
triggering
the third subevent.The duration of the first subeventis at
. ••A•*A•
4.28.
õ o.oo/. ß
0
2040
$0c
Fig. 3. Observed
(solid)andtheoretical
(dashed)
P, SH, and
SV wavesfor best-fitting3-subevent
model(Table2). Fault
planesolution
for thebestdouble-couple
of thesumof three
subeventsis shownat the centerof the figure, and total
source time function at the bottom of the P wave section.
HRV and stationnamesendingwith "B" are broadband
seismograms,
therestarelong-period.
"BB"and"LP"denote
broadband and long-period scales. Seismograms are
normalized
to a gainof 1500at a distance
of 40ø.
2). The moment released in the sequence(BB-SUM) is
equivalent
to anearthquake
of magnitude
Mw = 6.7.
Discussion
The centroidalparametersof the Armenianearthquake,
obtained from inversion of both the long-period and
broadbanddata (Table 2), are well resolvedand stable.On
average, the earthquakehad a right-lateral strike-slip
mechanismwith a large reversefaulting component. The
orientationof theprincipalaxes,depthof 6.0 km, andseismic
momentof 1.6x1019Nm determined
by the HarvardCMT
method [Dziewonski et al., 1989] are very similar to those
determinedhere. The centroiddepthof the mainshockis in
good agreementwith average depth and extent of the
aftershocks
[Simpsonet al., 1989].Our centroidalresultsare
also broadly consistentwith the estimatesfrom very-long-
period(100-300s) surfacewavesby Cisternas
et al. [1989].
Their mechanism,however,has a somewhatsteeperdipping
(~70ø)faultplaneandtheseismic
moment
is larger(2x1019
Nm) thanours.The highermomentestimatecanprobablybe
attributedto the differencein dip of the mechanismsand
frequencyof the waves.The momentreleaseassociated
with
least9 s, considerably
longerthanwouldbe neededunder
normalcircumstances
torupturean8 km segment.
The secondstage of the rupture initiated 4 s after the
earthquake origin time, before the first stage came to
completion.The mechanismof the secondsubeventand its
positionto the southeastwith respectto the first subevent
indicate that this rupture was associatedwith fight-lateral
stfike-slip faulting on a northwest-southeast
striking fault,
which appearsto be delineated by the aftershocksand a
pronounced topographic valley to the southeast of the
epicenter(Figure4). Becausethe first and secondsubevents
overlap in time, the derived parameterstrade-off to some
extent with each other (for example,it is possiblethat the
secondsubeventincludeseffectsof somestrike-slipfaulting
on the westwardpart of first event'srupture).However, the
agreementof our teleseismicallyderivedparameterswith the
field
observations
and the aftershock
distribution
adds
confidenceto our interpretation.Moreover, in spite of the
differences in the faulting mechanisms,the principal
compressionalaxes for these two subeventsare ofiented
north-southas is generallyobservedfor eventsin this area,
reflectingthe north-southcollisionof Arabia with Eurasia.
The third subeventis well separatedin time from the other
two and therefore its parameters are well resolved. It
correspondsto strike-slipfaulting at the westernend of the
aftershockzone. The maximumcompressional
axesfor this
mechanism,however,is rotatedcounter-clockwise
by about
43ø with respectto that of the first and secondsubevents,
indicatingthat this faulting is likely to be associatedwith
somepre-existingzoneof weaknessor stressredirectiondue
to heterogeneity.Because well-located aftershocksat the
westernend define a clear east-westlinearionfollowing the
trendof themajorvalley at 40.9øN, we choosetheeast-west
striking nodal plane as correspondingto the fault plane,
implyingright-lateralstrike-slipmotion.A LANDSAT image
of the regionshowsthat, in the areawhereour third subevent
is located,the major east-westtrendingvalley is intersected
by a secondprominentvalley,whichmay be a strike-slipfault
about25 km longwith a strikeof about265ø (Figure4). The
third subeventpossiblyoccurredat the westernend of the
aftershockzoneeitheron this intersectingfeatureor the eastwest trending valley. We interpret the third subeventas
faulting on a pre-existing zone of weakness,thereby
explainingthedifferencein P axesorientation.
1428
Pacheco et al.: Teleseismic Analysis of the Armenian Earthquake
Barka, A. A., and K. Kadinsky-Cade, Strike-slip fault
geometry in Turkey and its influence on earthquake
activity,Tectonics,7, 663-684, 1988.
Borcherdt,R., G. Glassmoyer,A. Der Kiureghian and E.
Cranswick,Effect of siteconditionson groundmotionsin
......
........
.•--..•-•...•. ••••••g.•.•.•.....•.•:•..
::•.:•..'
........
:.......
.
.. .- .-::•.W•:.•.•,:..
Leninakan, Armenia S.S. R., R. Borcherdt ed., Results
and datafrom seismologic
and geologicstudiesfollowing
earthquakesof December 7, 1988, near Spitak,Armenia
S.S. R. , U.S. Geological Survey Open-File Report,
40.9 .-;33•' *
. ß
'*:•-•i•.
3
.....
•............
--2 - '•"
40.8
.'-: '-•..
....
':'.
.....
89-163A, 86-108, 1989.
-•-•j•,•....
Cisternas,A., et al., The Spitak (Armenia) earthquakeof
December 7, 1988: field observations,seismologyand
tectonics, Nature, 339, 675-679, 1989.
Dziewonski, A.M., G. Ekstr6m, J. E. Franzen, and J. H.
40.7 .......
43.8
43.9
44.0
44.1
44.2
44.3
44.4
44.5
Fig. 4. Schematic
showingbodywaveinversion
results,
aftershock
disffibution
•d surface
rapture(smalldashedline)
for the•menian e•thqu•e. Fkst subevent
nucleated
north
of the townof Spit•. Centroids
of the secondandthird
subeventslocate 17 km southeastand 38 km west of the first
subevent,
respectively.
Faultplanesolutions
(labelled1, 2
and3) •e thoseobtainedin thisstudy.Heavydashedline
•awn through
theaftershocks
represents
thepresumed
hult
surhceat an averageaftershock
depth.Faultobserved
on
LANDSAT•ages is shown
witha thindashed
line.Shading
levelincreases
with 500 m stepsin elevation.
Woodhouse,Centroid-momenttensorsolutionsfor AprilJune 1986, Phys. Earth and Plan. Int., 45, 229-239,
1987a.
Dziewonski, A.M.,
G. Ekstr6m, J. E. Franzen, and J. H.
Woodhouse,Global seismicityof 1978:centroid-moment
tensor solutionsfor 512 earthquakes,Phys. Earth and
Plan. Int., 46, 316-342, 1987b.
Dziewonski, A.M., G. Ekstr6m, J. H. Woodhouse, and G.
Zwart, The Armenian earthquakeof 7 December 1988,
EOS Transactions, A.G.U., 70, 394, 1989.
Jackson,J., andD. McKenzie, Active tectonicsof the AlpineHimalayan belt betweenwesternTurkey and Pakistan,
Geophys.J. R. Astr. Soc., 77, 185-264, 1984.
Jackson,J., and D. McKenzie, The relationshipbetween
The damage sustainedat Spitak and Leninakan can be
partially attributedto their proximity to the first and third
subevents.Contributingfactorsto the destructionincludethe
locationof Spitak on the hangingwall of a shallowreverse
fault andthepossibilitythat softsedimentsmay havecaused
siteamplificationat Leninakan[Borcherdtet al., 1989].
Broadbandseismogramshave allowed us to resolve the
rupturehistoryof the Armenianearthquake
with greaterdetail
thancould be donewith only long-periodrecords. The main
event started at the intersection
of a reverse fault with two
right-lateral strike-slip faults. Four and ten secondsafter
nucleation,strike-sliprupturespropagatedbilateralysoutheast
andwest. This model indicatesthe presenceof a restraining
bendat theintersectionof two strike-slipfaultshavinga 45ø
differencein strike. The bendingof the fault is also seenin
the aftershocklocations(Figure 1). The westernend of the
rupturesequencecorresponds
to the two valleysclearly seen
in the LANDSAT image and topographicmaps (Figure 4).
The generalscenariofor the ruptureprocessof the Armenian
earthquakeis in agreementwith the studies by King and
NgbBlek[1985] and Barka and Kadinsky-Cade[1988] who
showedthat earthquakesoften nucleate and stop near the
bendsand intersectionsof faults. The aftershockactivity
(Figure 4) shows concentrationsat the epicenter of the
mainshock and the southeastern and western ends of the fault
zone,mostlikely indicatingstressconcentrations
dueto fault
intersections.
plate motions and seismicmoment tensors,and the rates
of active deformation
in the Mediterranean
and Middle
East, Geophys.J. R. Astr. Soc., 93, 45-73, 1988.
Kazmin, V. G., I. M. Sbortshikov, L-E. Ricou, L. P.
Zonenshain,J. Boulin, andA. L. Knipper,Volcanicbelts
as markersof the Mesozoic-Cenozoicactivemarginof
Eurasia, Tectonophysics,123, 123-152, 1986.
King, G., and J. N•ibBlek,Role of fault bendsin the initiation
andterminationof earthquakerupture,Science,228, 984987, 1985.
N•ibBlek, J. L., Determination of earthquake source
parametersfrom inversionof body waves,Ph.D. Thesis,
Mass.Inst. Technol.,Cambridge,Mass.,pp. 361, 1984.
Sharp,R., Surfacefaultinginvestigations,
R. Borcherdted.,
Resultsand datafrom seismologicand geologicstudies
followingearthquakes
of December7, 1988, near Spitak,
Armenia S.S. R. , U.S. Geological SurveyOpen-File
Report, 89-163A, 21-34, 1989.
Simpson,D., C. Langer, E. Cranswick, M. Andrews, and
G. Glassmoyer, Preliminary aftershock locations
(December21, 1988-January4, 1989), R. Borcherdted.,
Resultsand datafrom seismologic
and geologicstudies
followingearthquakes
of December7, 1988,nearSpitak,
ArmeniaS.S. R. , U.S. GeologicalSurveyOpen-File
Report, 89-163A, 75-85, 1989.
$eng6r,A.M. C., N. G6rtir, and F. $aro•lu,Strike-slip
faulting andrelatedbasinformationin zonesof tectonic
escape: Turkey as a case study, K. T. Biddle and N.
Christie-Blick eds., Strike-slip deformation, basin
formation, and sedimentation,Society of Economic
Acknowledgements.
We thankKlausJacob,LynnSykes,
andthreeanonymous
reviewers
for thoughtful
re.views
and
Paleontologists
andMineralogists,specialpublication,
the staffat the Albuquerque
Seismological
Laboratoryfor
37, 227-264, 1985.
supplyingthe networkday tape. CharleyLanger,Ed
Stein,R. S., andR. S. Yeats,Hiddenearthquakes,
$ci.
Cranswick,
RogerBorcherdt
andMaryAndrews
of theU.S.
Am., 6, 48, 1989.
GeologicalSurveyprovideddata usedin the aftershock
locations.This projectwas supported
by Departmentof
Energy grant DE-FG02-ER13221D[C.H.E.]; National
C. H. Estabrook,J. F. Pachecoand D. W. Simpson,
ScienceFoundation
grantsEAR-8896187[J.L.N.],EAR87Lamont-Doherty
GeologicalObservatory,
Palisades,
NY 10964.
07719 [J.F.P.]; andU.S. GeologicalSurveygrantUSGS
14-08-0001-G1365 [D.W.S.]. L-DGO contribution4543.
J. L. N•ibElek, College of Oceanography,Oregon State
University,Corvallis,OR 97331.
References
Ambraseys,
N. N., andR. D. Adams,Long-termseismicity
(Received:July31, 1989;
of North Armenia, EOS Transactions,A.G.U., 70, 145-
revised: October 18, 1989;
154, 1989.
accepted:
October18, 1989.)