Source mechanism of February 4, 1965, Rat ... - Wiley Online Library
VOL.
78, NO. 26
JOURNAL
OF GEOPHYSICAL
RESEARCH
SEPTEMBER
10, 1973
Source Mechanism ofFebruary 4, 1965,RatIslandEarthquake FaA•cIs
T. Wu
Department o] Geology, State University o] New York Binghamton, New York 13901 HIROO KANAMORI !
Earthquake Research Institute, University o] Tokyo Tokyo, Japan
The Rat Island earthquake of February 4, 1965 (origin time 05h 01m 21.8s, h = 40 km), is one of the largest earthquakes recorded in recent years. On the basis of the radiation patterns and the amplitudes of the great circle Rayleigh and Love waves, the earthquake is found to have the following characteristics: fault plane dip, 18ø; fault plane dip direction, N19øE; rupture propagation direction, N51øW; rupture propagation velocity, 4.0 km/sec; faul't length, 500 km; moment, 1.4 X 10•ø dynes cm; stress drop, 30 bars; and average dislocation, 2.5 meters. It is interesting to observe that the attitude of the fault plane and the general features of this earthquake are similar to those of the 1963 Kurile earthquake (Kanamori, 1970a) and the 1964 Alaskan earthquake (Kanamori, 1970.b). The first 35 see of P waves were also well recorded at many stations. The wave forms suggesta multiple-event nature of the earthquake for at least the initial 35 sec. By using the relative location method, the events were loca.t.ed progressively south of the initial hypocenter. It seems plausible that the earthquake started off at depth, first propagated southward, and then westward along the Aleutian arc. Owing to the long-period nature of the surface waves, only the average feature of the fault is seen, and P waves reveal some of the detailed initial
behavior.
In the period 196.2-1965,during which the World-Wide Standard Seismograph Network (WWSSN) operated with To -- 30 sec and Tg -- 100 sec,severallarge shallowearthquakes occurred in the circum-Pacific
island arc chain.
With To -- 30 sec, enough long-period (T > 150 sec) wavesappear on the recordsfor a study of the dynamicsof faulting of theselarge events. Kanarnori [1970a, b] has studied the 1963 Kurile earthquake (M -- 8.2) and the 1964 Alaskan earthquake (M -- 8.5). Both studies reveal the shallow thrust
nature
of the events
and the fact that rupture propagatesmore or less along the strike directions. Besides, the studies also clearly demonstrate the relative 'size' of these earthquakes: the moment of the Alaskan event is approximately 10 times larger than that
of the Kurile
event.
Undoubtedly, the February 4, 1965, Rat Island earthquake is one of the major events. It excited large-amplitude free oscillationsof the earth [Nowro,ozi, 1966], and at many stations with low magnification(X750) one could see well-recordedRo and Go, and it is not uncommonto seeR9 on many of the WWSSN records.
The
aftershock
area
extends
over
a
distance of approximately 600 km along the Aleutian arc, and there are more than 10 aftershockswith magnitudeshigherthan 6. It would be therefore important to know the source mechanism and to see whether this earthquake shares some common characteristics
of the two
already studied.Stauder [1968b] worked out the partial focal mechanismsolution for the main shock and complete solutions for many of the aftershocks. These data will assist us in
choosingour preliminary trial solutions. • Now at Seismological Laboratory, California Institute of Technology, Pasadena, California 91109.
Copyright ¸
1973 by the American Geophysical Union.
6082
The waves used for the surface wave source
mechanism for this earthquake are mainly G5 and Rs. The method of analysis is much the same as that used earlier by Kanamori [1970a,
Wu ANt) iKANAMORI: RAT ISLANt) EARTI•QUAKE
6083
P WAVES AND INITIAL SOURCE BEHAVIOR b]. The frequency content of these waves is concentratedin the range 0.002-0.01 cycle (corFor most large earthquakes the long-period respondingto 500-100 sec). As a consequence, P waves show complicated groups of pulses. the detailedbehavior,with a time scaleof the The amplitude relations of these pulses vary order of 50 sec or less,will not be revealedby from station to station, and the time intervals these waves at all. What we observe are the between them do not bear any relation to the averagepropertiesof the faulting process. known P phasesrefracted and reflected at the In addition to the source mechanismstudy, various discontinuities inside the earth. These the method above also presents information pulses are therefore most probably the result that enablesus to determine the relative magof a complicated source-time function. nitude of large earthquakes. This ability is Some examplesof the long- and short-period rather important. For large earthquakes the P waves for the Rat Island mainshock are magnitudeis alwayssomewhatambiguous, be- shown in Figure 1. With the exception of those cause a typical P wave usually consistsof a recorded at United States continental stations, groupof pulses,and the conceptof magnitude long-period P waves are found to correlate determination in this case breaks down; the very well at stationsof all distancesand azimuth long-periodsurface waves effectivelyintegrate angles.The exceptionsare probably due to the these individual events and differentiate very closenessof those stations to the radiation patwell the relative magnitude of the large events. tern minimum. In Figure I we can see that This earthquakehas a U.S. Coast and Geo- there are usually identifiable short-period P deticSurvey (USCGS) mb- 6.0 but a surface phasesassociatedwith a long-periodpulse; the corwave magnitude of about 7•. For the 1964 reverse is not true. The station-to-station Alaskan earthquakem, ----8.5 and M8 -- 8.4relation of the short-period records is very 8• (USCGS Seismological Bulletin MSI 279), poor. In the following,the long-periodP waves are the ones used for analyses.In Figure 2 the and for the 1963 Kurile earthquakeM8 -- 7•differential travel times tp, -- tp, where t•, is 8•/• (USCGS SeismologicalBulletin MSI 274; no m,value published). The m, magnitudesare the time of the ith arrival and t• is that of the first motion, are plotted as a function of based on a particular part of a complicatedP azimuthal angle. wave train, and M• is basedon a particularpart To locate the secondaryevents, one can use ('20-sec Rayleigh waves') of a dispersedsura simple least squares procedure [Fuka.o., face wave; neither is representativeof a large 1972]. Assume that the individual events are earthquake. On the other hand, if we measure separatedin time by •o and in spaceby 3A in magnitudes by waves with periods commena certain direction; then at the ith station we surate with the source-timeduration, we have have effectively summed the multiple events and = avoided the lateral refraction and other problems associated with
where St, is the differential time between two
the 20-sec waves.
For this particular earthquake we were able to use P waves to study the initial multiple source event
behavior
of the
source.
Similar
analyses have been carried out by Wyss and Brune [1967], amongothers.They used shortperiod (mainly Wiechert) records and concluded that the Alaskan earthquake can be representedas a multiple event. That a large earthquake consists of a number of small events is only natural, when the diverse nature of frictional characteristicsalong the entire fault zone spanning hundredsof kilometers is
arrivals, r• is the P wave velocity, and •, is the angle betweenthe line element connectingthe two hypocentersand the ray leaving the first hypocenter. By assuminga direction of the line element,one can computethe 6, and then use a linear regressionprocedureto find •o and 8A. The direction of the line element is given
by the azimuth.• measuredfrom the north and the dip angleD. The methodto determinethe best set of •o and 8.A values consistsof varying the direction of the line element until the minimums in standard error of the estimation of
considered. We are able to follow the details
the distanceS, or standard deviation of the time e, can be identified.It was found that the
of sourcedevelopmentin the first 35 sec or so.
azimuth •
of the line element can be de-
6084
Wu AND KANAMORI' RAT ISLAND EARTI-IQUAK•
termined well for each D assumed, but the variation
of the inclination
of S or e varies
little with D (Figure 3). The results are presented in Figure 4 and Table 1. In both Figure 4 and Table 1, first event refers to the first multiple event after the first break and so on. Values of to, 3A, and the dip angle are determinedrelative to the first break. We see that D is positive for the first and third events and negative for the second event; i.e., within the initial 21 sec of the earthquake, the source developssouth and southeastward,the first and third events develop downdip, and the third event developsupdip from the hypocenter. As we shall see in the next section,the dominant growth of the fault was westward along the Aleutian arc over a distance of 500 km in about 2.5 min; the body wave results presented here indicate the initial source behavior. This southward growth may be simultaneous with the westward smoother propagation or precedethe main phase of fault propagation.
AKU LPN
•, = 62.6 o
AKU
SPZ
AZ = 7,7 ø
•
0511
CH6
0512
COP LPN
HKC LPE
o5•'
8A6 LPN
0511'
SURFACE WAVES
In this study the R5 and G5 are the chief waves used. Where it is necessary,R, and G4 and R6 and C6are usedto supplementthe data. The seismogramsare digitized within group velocity windows of 3.7 and 3.4 kmfiec for Rayleigh waves and 4.5 and 4.3 km/sec for Love waves and equalizedin frequencydomain for instrument response,attenuation, and dispersionto a standardepicentraldistanceof 9w/ 2. The equalized spectra are then inverseFourier-transformedto time domain (seeKanamo.ri [1970a, b] for details). The equalized R5 are presentedin Figure 5 and G• in Figure 6. The central figures are azimuthal plots of the maximumamplitudesof thesewaves; these are the radiation patterns.The Rayleigh radiation pattern is apparently dipolar and shows a strong asymmetry accentuating the upper
KON $PZ
A : 69
0ø
AZ:-5
/•:
9 ø
38 5 ø
AZ:-90.2
ø
MINUTE
lobe and the westernflanksof both the upper and the lower lobes. The Love wave pattern is not distinctly lobed, but a strong northwest
waves
accentuation is clearly indicated.
Noticethat the short-period records on the
Fig. 1. Examples of long- and short-period P recorded
at
several
WWSSN
stations.
The methodof interpretingthese equalized right are muchmorecomplicated than the longwavesand radiationpatternhas alreadybeen periodrecords. expounded in detail elsewhere [Kanamo. ri,
1970a,b]. It comprises two mutually related patternand the Rayleigh/Loveamplituderatio steps: (1) findinga double-couple equivalent and (2) accounting for the asymmetryin the forcesystemthat yieldsthe correctradiation individualpatterns by assuminga moving
WU AND KANAMORI' RAT ISLAND EARTHQUAKE
./
/
/o x•• x x •
.W
•x x
ß
• x
o
x
x lOS
opea c•c]es, secoadevea•; so]id c•rc]es,•hi•d e•ea%s.
1st
EVENT 185
$ = 200 ø
1.09 108
18 3 ß
ß
ß
ß
ß
1.07
2nd
I
ß
'
ß
18.2
EVENT
I.I
108
1.09
I0 7
165o
108
106
o
s
I 07
S
ß
10.5
ß
106
I0 4
1.05
10.3
3rd
EVENT
1.76
22
1,75
22.5
1,74
o
o
22
ß
o ß
1.73 o
o
75
25
S
o
22.0
o
ß
21.75
1.71
I.......... -12 -9
-15
-6
-3
0
3
$
9
12
I
15
I
18
I ;d ;7
21
30
20.5
D
Fig. 3. Variation of S, standard error of •/x in percent, and e, standard deviation in seconds, as a function of D, the dip of the line element.
6086
Wu AND KANAMORI' RAT ISLAND EARTHQUAKE
source model with fault length, propagation velocity and directionof faulting as parameters. We have additionalguidancefor this interpretation at our disposal:(1) Stauder'spartial focal mechanism solutions for this main event (Figure 7a) and also complete solutions for many
large aftershocksin the seriesand (2) spatial distribution of aftershocks(Figure 7b).
• ' vo=20.8 sec 8/5
It shouldbe notedthat, because of the long-
= 140 km
periodwavesthat we usehere, the featuresderived concerningthe source are average properties. Temporally, we are not able to resolve
detailsbelow100 sec.The advantageof this
ß
r o = 14.7 sec
method is that a definite choice of fault plane
•
can be made; in fact, for large shallowthrust earthquakes this method is the only one available to resolvethe low-angle thrust plane. The best fit of the experimental data is presented
•A =86 kmß
in Figure 8a. In Figure.8b, theoreticalradiation patterns for rupture velocities of 3, 3.5, and 5 km/sec are compared with the pattern for 4.0 km/sec. The results obtained indicate that the fault plane strikes N70øW and dips 18ø in N20øE direction. The auxiliary plane constrained by Stauder's[1968b] partial solutionstrikesN60øE and dips 78ø in S30øEdirection. The fault
6.5 sec 55 km i,
,
i,,
,
C0S
0
Fig. 4. Least squares lines for the solutions. Here 0 is the angle between the line element IA and the ray leaving the source for the station. DIscUsSiON
lengthL i• about500km, the rupturevelocity
It may seem that the conclusions from the is about 4 km/see, and the seismicmoment Mo = 1.4 X 10• dynescm.If we a•sume,from studies above are contradictory. We have seen the aftershock area, the width W of the fault that, whereasthe body wavesindicatenot only plane to be 150 km, the average.dislocation a jerky propagationbut alsoa southwardpropagation,the long-periodsurfacewavesshbwa (u) on the fault can be estimated as (u) = smooth WNW propagation.These conclusions Mo/!•LW = 2.5 meters,where • = 7.0 X 10• for the multiple dynes/em• is used for an appropriate value of are not, however,inconsistent, the rigidity. The stress drop may be estimated, accordingto Knopo# [1958], by A•r ----8.l•(u)./,rW= 30 bars. The fault plane solutionobtainedhere is very similar
to that
obtained
from
the aftershocks
that occurredalong the island chain [Stauder, 1968b]. The slip angle measured clockwiseon the fault plane from the strike direction, N70øW, is 41.4ø. This means that the foot wall side (oceanicside) moved in about N30øW direction with respect to the hanging wall side (continental side). This slip direction is more or lessparallel to the slip directionsdetermined for the 1964 Alaskan earthquake, 1963 Kurile Islands earthquake, and the 1968 Tokachi-Oki earthquake [Kanamo.ri, 1971].
events in the initial 21 sec are not necessarily
relatedto the main propagation phase,which could have commencedsimult,aneously, and more smoothly,or 21 seeafter the initial break.
Because of thelargeamplitudes of thecomplex
long-period P waves, •theywerenotrecorded well on the WWSSN' seismogramsand hence
werenot subjected to studywith the sametech-
TABLE 1.
Multiple
Events
Error,
Event
1 2 3
D
6 -6 15
•
•
•o
200 165 140
1.1 1.2 1.7
6.5 14,7 20.8
33 86 140
Apparent Velocity, km/sec
6h
18.3 10.4 21.6
5.1 5.9 6.7
WU AND KANAMOn•' RAT ISLAND EARTHQUAKE •
o
6087
6088
Wu ANDKANAMORI' RATISLAND EARTHQUAKE
F1Oh •55m F1oh•55m NUR I10 cm VAL
ATU
PDA
DAI
I0
--,270
2o
901 PEL
180
AF I
,
;500
Fig. 6.
,
I
sec
Lovewave radiation pattern. A]]waves reduced to• ,: 9•r/2.
nique. These wavescould have their sources
uationis moreserious, because hereonlythe
lying alongthe path of propagation as de- firstfewcycles oftheP wave(orother phases) lineatedby the surfacewavemethod.That the
150- to 300-secsurface waves do not reveal
are used.As we havementioned before, the
USCGSv•bfor the Rat Islandmainshock was the sameinformation as the bodywavesis only 6.0 (USCGSSeismological BulletinMSI only natural, sincethesewavescannot discern
behaviorof shortduration.Thereforefrom the
290),whichimpliesthat the energyof this earthquake is lessthanabout11100,000 that
surface waves wecandeduce onlytheaverage of the1964Alaskan earthquake andabout1/
velocityof rupturepropagation togetherwith 100,000 that of the 1963Kurileearthquake therupturelengthbutnotthepossible accelera- [Gu•enbe'rg, 1956].Fromthe seismic moment tions and decelerations with a time scaleof
it is clearthat the Rat Islandearthquake is smaller thantheAlaskan earthquake, but not The multiple-event natureof largeearth- by 100,000timesin energy,and in fact is quakesmakesthe meaning of the magnitudeslightly larger thanthe1963Kurileearthquake. ambiguous. If M•. is used,then,depending on If themagnitude istorepresent thetotalenergy the proximityof the stationto a particular oftheearthquake, anappropriate sumofmagperhaps50 sec or less.
sectionof the fault, the magnitude will reflect nitudes of theindividual eventsshould beused.
the characteristics of that section, sinceby
Theideaofmultiple events hasbeen proposed
definition ML is determined by the maximum by seismologists forseveral largeshallow earthrecorded motionof a part of the seismogram quakes[Wyssand Brune,1967;Florenso•) and
andthenearest section of a faultislikelyto Solonenko, 1963], foronedeep earthquake [Fukao,
contribute mostof the energyowingto the 1972],andevenforearthquakes of magnitudes near-field effect.If mbisused,however, thesit- around 6 [Wu,1968; Niazi,1969]. Conceptually,
Wu AND KANAMORI' RAT ISLAND EARTHQUAKE 55÷
!
,
!
,
,
,
6089
,
!
54
œ
ß5
ß
55-
o
4
o
00
ß
0
0ß 3
NODAL
0
0
GO ø 00 0
GO
7ß ø
14
o•DOo
0 0
øQD•DO
0
GO %o
0
GO
o
o
8
o
o ß9
o
o0 0
GO øoo o
I0 0
o
0
o
0•0 0 ß6 00
ß11 00 ß0
0
00 00 0010 00 0
00
0
ß12
0% 0
o
o •16 o
00
0
o
o
oooOo o
o
o
00
0000
0
0
oø
0
0
o
0
0
o o 000
o.,,
ß ß15
0 0
oo •, 0
0ß2 0
0
o
(a) RAT
LAT
1965
o o
ISLAND
•
Main
&
m_•6
0
6>m_•5
AFTER
SHOCKS
shock
0
0
0
N
70
175
LONG œ
180
LONGITUDE
Fig. 7. (a) Large aftershocks.Numbers attached to the solid triangles mark the time sequenceof events with m• •_ 6 (from USCGS Preli-minary Determination of Epicenters). (b) Stauder's partial solution for the main shock.
there are two ways to view multiple events in terms of sourcespaceand time functions. One is
B**(•, t) (Figures 9c, d). In this casethere are 'dead' sectionsalong the fault during the main that the time futictionat eachpoint is uniform propagation phase, and aftershock could occur but consistsof a delayedsum of ramps (Figure at theseplaces.In reality it is highly possiblethat 9a). This ramp could propagate with a uniform both modesoccur during the main shockof the velocity. At each instant along the fault there Rat Island series. will be a numberof activesectionscorresponding If the earthquakeis composedsimply of a to the acceleratingand deceleratingparts of the propagating rupture, the durations of P waves ramp (Figure 9b). The other possibilityis that a in different azimuthal directions would be difpropagatingdislocationhas a changingdensity ferent. Thus in the direction of propagationat function;i.e., B(•), the Burgersvector[Weertman, large distance we would expect a duration of 1969],is alsoa functionof x, the coordinate along about 100 sec and in the opposite direction of the fault, B*(•, x), or a function of time t, about 150 sec. But on actual short-period
6090
Wv AND KANAMORI: RAT ISLAND ]•ARTI-•QUAKE
seismogramsfor the Rat Island earthquake we do not see such a variation.
N
The P waves in all
azimuths have decayingtails of about the same length. Thus a sequenceof aftershocksoccurring behind the main propagating dislocation line accompaniesthe main propagation, and the number
of these aftershocks
decreases ex-
ponentially as a function of time. For shallow large earthquakes behind the trench, body wave focal mechanismsusually fail to give a completesolution [Stauder, 1968a] becauseof the lack of readable close-inrecords, and hence it is difficult
to determine
a shallow
dipping plane. When they are combinedwith the surfacewave method, however,the parameters for this plane can be estimated. Stauder [1968a, b] has shown that in the Aleutian Island chain the events with
thrust
mechanisms
Fig. 8a. Preferredfit of the long-periodsurface wave data. Thin line representstheoreticalRay-
all have very similar high-angle planes. For
leigh wave data; heavy line, theoretical Love wave data; solid circles,observed'Rayleighwave
the
data; crosses,observed Love wave data.
1965 Rat
Island
series the same
All these events show a NE-SW
strike
is true.
and a
southwarddip of about 75ø. That this plane cannot be the fault plane for the main shock is establishedby lookingat the asymmetryof the Rayleighand Love wave radiationpatterns, which imply a NW rupture propagationdirection. Sincethe NE-SW plane is steepand dipping south,a nonstrikeNW directionof propagation is unreasonable,unless small segments of the fault line up en echelon;however,the continuoustrench and the shape of tim island arc do not agree with such geometry. BenMenahem and Rosenman [1972] adopted a high-angle solution for their interpretation of this earthquake;the strike of the planeis not consistentwith Staude'r's[1968a] solutionsfor earthquakes in this region. Stauder [1968a] also showedthat the shallowplane dips about
Both the P wave and the surface wave re-
sults yield rupture velocities.The multiple eventsimply velocitiesof 5.1-6.7 km/sec (Table 1). These valuesare higher than thoseof shear wave velocity and approach those of
15ø-20 ø in the direction of N50øW-N90øW for most of the thrust events in the aftershock
series. By using a shallow-angleplane consistent with the partial solutionof the main
shock,we foundthat the directionof propagation has to deviate slightly from the strike. However,sincethe dip is shallow,the 10ø or so deviationonly impliesa dip directionpropagationof about 85 km, which couldbe accommodated very well by the width of the fault plane. The same sort of deviation has been ob-
servedfor the 1964Alaskanearthquake[Kanamori, 1970b].
Fig. 8b. Other rupture velocities. Calculated radiation patterns with rupture velocities of 3.0 (R•, /_a; heavy dashed line), 3.5 (Re, La; medium dashed line), and 5 km/sec (Rs, I_•; thin dashed line).
WU AND KANAMORI' RAT ISLANDEARTttQUAKE (a)
6091
Finally, it shouldbe noted that the parallelismof the slip vectorsfor major earthquakes in the northern Pacific is conformable to the
Pacific basin'smoving as a rigid plate [McKenzie and Parker, 1967]. REFERENCES
Ben-Menahem,A., and M. Rosenman,Amplitude patterns
t=t 1
(b)
•o
n>
x =ct t
B(•',X•)
(c)
/"
of
tsunami
waves
from
submarine
earthquakes,J. Geophys. Res., 77, 3007-3128, 1972.
Florensov,N. A., and V. P. Solonenko,The GobiAltai earthquake, Izd. Akad. Nauk SSR, 1963. (English translation by Israel Program for Scientific Translations, p. 392, 1965.) Fukao, Y., Source processof a large deep-focus earthquake and its tectonic implications--The western Brazil earthquake of 1963, Phys. Earth Plan.et. Interiors, 5, 61-76, 1972.
Gutenberg,B., The energyof earthquakes, Quart. J. Geol. Soc. London, 112, 1-14, 1956. Johnson,T., F. T. Wu, and C. It. Scholz,Source parametersfor stick-slip and for earthquakes, Science,179, 278-279, 1973.
t=t I
(d)
t =t 2
Kanamori, H., Synthesisof long-periodsurface wavesand its applicationsto earthquakesource studies--Kurile Islands earthquake of October 13, 1963,J. Geophys.Res., 75, 5011-5028,1970a. Kanamori, It., The Alaskan earthquake of 1964: Radiation of long-period surface waves and source mechanism,J. Geophys.Res., 75, 50'295040, 1970b.
Kanamori, II., Focal mechanism of Tokachi-Oki
earthquakeof May 16, 1968,Tectonophysics, 12, Fig. 9. Generation of multiple-event type of seismograms.(a) Displacement of a point along the fault. (b) Particle velocity function alongthe fault at t = t for uniform propagating fault. This function will move down the line as t increases.(c) Burgers vector as a function of • and x, B(•, x•.) -- 0. (d) Dislocation along the fault at different t.
compressionalwaves. Recently, such values have been observedin the laboratory [Wu et a/., 19'72; Johnson et al., 1973] and attained theoretically (R. Burridge, personal communication, 1973). In both instances,a low frictional coefficientis deemed responsible.For the experiments,low friction was achievedthrough acceleratingcreep; it is conceivablethat this explanationis also applicable in real faulting. The surfacewave resultsindicatean average value of 4 km/sec for the whole fault. In view
of the complexprocessin the sourceregion,the meaning of this value deserves further consideration.
1-13, 1971.
Knopoff, L., Energy releasein earthquakes,Geophys. J., 1, 44-52, 1958.
McKenzie, D. P., and R. L. Parker, The North Pacific: An example of tectonicson a sphere, Nature, 216, 1276-1280,1967. Niazi, M., Source dynamics of the Daslit• Bayaz earthquake of August 31, 1968, Bull. Seismol. Soc. Amer., 59, 1843-1861,1969. Nowroozi, A. A., Terrestrial spectroscopyfollowing the Rat Island earthquake, Bull. Seismol. Soc.Amer., 56, 1269-1288,1966.
Stauder, W., Tensional characterof earthquake foci beneath the Aleutian trench with relation to
sea floor spreading,J. Geophys.Res., 73, 76937701, 1968a. Stauder, W., Mechanism of the Rat Island earth-
quake sequence of February 4, 1965, with relation to island arcs and sea floor spreading, J. Geophys. Res., 73, 3847-3858, 1968b. Wu, F. T., Parkfield earthquake of June 28, 1966: Magnitude and sourcemechanism,Bull. $eismol. Soc. Amer., 58, 689-709, 1968. Wu, F. T., K. C. Thomson, and H. Kuenzler, Stick-slip propagation velocity and seismic source mechanism, Bull. Seismol. Soc. Amer., in press, 1972.
6092
Wu ANI) KANAMORI: RAT ISLANI) EARTHQUAKE
Weertman, J., Dislocation motion on an interface with friction that is dependent on sliding velocity, J. Geophys. Res., 74, 6617-6622, 1969. Wyss, M., and J. N. Brune, The Alaskan earthquake of March 28, 1964, a complex multiple
structure, Bull. Seismol. Soc. Amer., 57, 10171023, 1967.
(Received June 26, 1972;
revised May 16, 1973.)
78, NO. 26
JOURNAL
OF GEOPHYSICAL
RESEARCH
SEPTEMBER
10, 1973
Source Mechanism ofFebruary 4, 1965,RatIslandEarthquake FaA•cIs
T. Wu
Department o] Geology, State University o] New York Binghamton, New York 13901 HIROO KANAMORI !
Earthquake Research Institute, University o] Tokyo Tokyo, Japan
The Rat Island earthquake of February 4, 1965 (origin time 05h 01m 21.8s, h = 40 km), is one of the largest earthquakes recorded in recent years. On the basis of the radiation patterns and the amplitudes of the great circle Rayleigh and Love waves, the earthquake is found to have the following characteristics: fault plane dip, 18ø; fault plane dip direction, N19øE; rupture propagation direction, N51øW; rupture propagation velocity, 4.0 km/sec; faul't length, 500 km; moment, 1.4 X 10•ø dynes cm; stress drop, 30 bars; and average dislocation, 2.5 meters. It is interesting to observe that the attitude of the fault plane and the general features of this earthquake are similar to those of the 1963 Kurile earthquake (Kanamori, 1970a) and the 1964 Alaskan earthquake (Kanamori, 1970.b). The first 35 see of P waves were also well recorded at many stations. The wave forms suggesta multiple-event nature of the earthquake for at least the initial 35 sec. By using the relative location method, the events were loca.t.ed progressively south of the initial hypocenter. It seems plausible that the earthquake started off at depth, first propagated southward, and then westward along the Aleutian arc. Owing to the long-period nature of the surface waves, only the average feature of the fault is seen, and P waves reveal some of the detailed initial
behavior.
In the period 196.2-1965,during which the World-Wide Standard Seismograph Network (WWSSN) operated with To -- 30 sec and Tg -- 100 sec,severallarge shallowearthquakes occurred in the circum-Pacific
island arc chain.
With To -- 30 sec, enough long-period (T > 150 sec) wavesappear on the recordsfor a study of the dynamicsof faulting of theselarge events. Kanarnori [1970a, b] has studied the 1963 Kurile earthquake (M -- 8.2) and the 1964 Alaskan earthquake (M -- 8.5). Both studies reveal the shallow thrust
nature
of the events
and the fact that rupture propagatesmore or less along the strike directions. Besides, the studies also clearly demonstrate the relative 'size' of these earthquakes: the moment of the Alaskan event is approximately 10 times larger than that
of the Kurile
event.
Undoubtedly, the February 4, 1965, Rat Island earthquake is one of the major events. It excited large-amplitude free oscillationsof the earth [Nowro,ozi, 1966], and at many stations with low magnification(X750) one could see well-recordedRo and Go, and it is not uncommonto seeR9 on many of the WWSSN records.
The
aftershock
area
extends
over
a
distance of approximately 600 km along the Aleutian arc, and there are more than 10 aftershockswith magnitudeshigherthan 6. It would be therefore important to know the source mechanism and to see whether this earthquake shares some common characteristics
of the two
already studied.Stauder [1968b] worked out the partial focal mechanismsolution for the main shock and complete solutions for many of the aftershocks. These data will assist us in
choosingour preliminary trial solutions. • Now at Seismological Laboratory, California Institute of Technology, Pasadena, California 91109.
Copyright ¸
1973 by the American Geophysical Union.
6082
The waves used for the surface wave source
mechanism for this earthquake are mainly G5 and Rs. The method of analysis is much the same as that used earlier by Kanamori [1970a,
Wu ANt) iKANAMORI: RAT ISLANt) EARTI•QUAKE
6083
P WAVES AND INITIAL SOURCE BEHAVIOR b]. The frequency content of these waves is concentratedin the range 0.002-0.01 cycle (corFor most large earthquakes the long-period respondingto 500-100 sec). As a consequence, P waves show complicated groups of pulses. the detailedbehavior,with a time scaleof the The amplitude relations of these pulses vary order of 50 sec or less,will not be revealedby from station to station, and the time intervals these waves at all. What we observe are the between them do not bear any relation to the averagepropertiesof the faulting process. known P phasesrefracted and reflected at the In addition to the source mechanismstudy, various discontinuities inside the earth. These the method above also presents information pulses are therefore most probably the result that enablesus to determine the relative magof a complicated source-time function. nitude of large earthquakes. This ability is Some examplesof the long- and short-period rather important. For large earthquakes the P waves for the Rat Island mainshock are magnitudeis alwayssomewhatambiguous, be- shown in Figure 1. With the exception of those cause a typical P wave usually consistsof a recorded at United States continental stations, groupof pulses,and the conceptof magnitude long-period P waves are found to correlate determination in this case breaks down; the very well at stationsof all distancesand azimuth long-periodsurface waves effectivelyintegrate angles.The exceptionsare probably due to the these individual events and differentiate very closenessof those stations to the radiation patwell the relative magnitude of the large events. tern minimum. In Figure I we can see that This earthquakehas a U.S. Coast and Geo- there are usually identifiable short-period P deticSurvey (USCGS) mb- 6.0 but a surface phasesassociatedwith a long-periodpulse; the corwave magnitude of about 7•. For the 1964 reverse is not true. The station-to-station Alaskan earthquakem, ----8.5 and M8 -- 8.4relation of the short-period records is very 8• (USCGS Seismological Bulletin MSI 279), poor. In the following,the long-periodP waves are the ones used for analyses.In Figure 2 the and for the 1963 Kurile earthquakeM8 -- 7•differential travel times tp, -- tp, where t•, is 8•/• (USCGS SeismologicalBulletin MSI 274; no m,value published). The m, magnitudesare the time of the ith arrival and t• is that of the first motion, are plotted as a function of based on a particular part of a complicatedP azimuthal angle. wave train, and M• is basedon a particularpart To locate the secondaryevents, one can use ('20-sec Rayleigh waves') of a dispersedsura simple least squares procedure [Fuka.o., face wave; neither is representativeof a large 1972]. Assume that the individual events are earthquake. On the other hand, if we measure separatedin time by •o and in spaceby 3A in magnitudes by waves with periods commena certain direction; then at the ith station we surate with the source-timeduration, we have have effectively summed the multiple events and = avoided the lateral refraction and other problems associated with
where St, is the differential time between two
the 20-sec waves.
For this particular earthquake we were able to use P waves to study the initial multiple source event
behavior
of the
source.
Similar
analyses have been carried out by Wyss and Brune [1967], amongothers.They used shortperiod (mainly Wiechert) records and concluded that the Alaskan earthquake can be representedas a multiple event. That a large earthquake consists of a number of small events is only natural, when the diverse nature of frictional characteristicsalong the entire fault zone spanning hundredsof kilometers is
arrivals, r• is the P wave velocity, and •, is the angle betweenthe line element connectingthe two hypocentersand the ray leaving the first hypocenter. By assuminga direction of the line element,one can computethe 6, and then use a linear regressionprocedureto find •o and 8A. The direction of the line element is given
by the azimuth.• measuredfrom the north and the dip angleD. The methodto determinethe best set of •o and 8.A values consistsof varying the direction of the line element until the minimums in standard error of the estimation of
considered. We are able to follow the details
the distanceS, or standard deviation of the time e, can be identified.It was found that the
of sourcedevelopmentin the first 35 sec or so.
azimuth •
of the line element can be de-
6084
Wu AND KANAMORI' RAT ISLAND EARTI-IQUAK•
termined well for each D assumed, but the variation
of the inclination
of S or e varies
little with D (Figure 3). The results are presented in Figure 4 and Table 1. In both Figure 4 and Table 1, first event refers to the first multiple event after the first break and so on. Values of to, 3A, and the dip angle are determinedrelative to the first break. We see that D is positive for the first and third events and negative for the second event; i.e., within the initial 21 sec of the earthquake, the source developssouth and southeastward,the first and third events develop downdip, and the third event developsupdip from the hypocenter. As we shall see in the next section,the dominant growth of the fault was westward along the Aleutian arc over a distance of 500 km in about 2.5 min; the body wave results presented here indicate the initial source behavior. This southward growth may be simultaneous with the westward smoother propagation or precedethe main phase of fault propagation.
AKU LPN
•, = 62.6 o
AKU
SPZ
AZ = 7,7 ø
•
0511
CH6
0512
COP LPN
HKC LPE
o5•'
8A6 LPN
0511'
SURFACE WAVES
In this study the R5 and G5 are the chief waves used. Where it is necessary,R, and G4 and R6 and C6are usedto supplementthe data. The seismogramsare digitized within group velocity windows of 3.7 and 3.4 kmfiec for Rayleigh waves and 4.5 and 4.3 km/sec for Love waves and equalizedin frequencydomain for instrument response,attenuation, and dispersionto a standardepicentraldistanceof 9w/ 2. The equalized spectra are then inverseFourier-transformedto time domain (seeKanamo.ri [1970a, b] for details). The equalized R5 are presentedin Figure 5 and G• in Figure 6. The central figures are azimuthal plots of the maximumamplitudesof thesewaves; these are the radiation patterns.The Rayleigh radiation pattern is apparently dipolar and shows a strong asymmetry accentuating the upper
KON $PZ
A : 69
0ø
AZ:-5
/•:
9 ø
38 5 ø
AZ:-90.2
ø
MINUTE
lobe and the westernflanksof both the upper and the lower lobes. The Love wave pattern is not distinctly lobed, but a strong northwest
waves
accentuation is clearly indicated.
Noticethat the short-period records on the
Fig. 1. Examples of long- and short-period P recorded
at
several
WWSSN
stations.
The methodof interpretingthese equalized right are muchmorecomplicated than the longwavesand radiationpatternhas alreadybeen periodrecords. expounded in detail elsewhere [Kanamo. ri,
1970a,b]. It comprises two mutually related patternand the Rayleigh/Loveamplituderatio steps: (1) findinga double-couple equivalent and (2) accounting for the asymmetryin the forcesystemthat yieldsthe correctradiation individualpatterns by assuminga moving
WU AND KANAMORI' RAT ISLAND EARTHQUAKE
./
/
/o x•• x x •
.W
•x x
ß
• x
o
x
x lOS
opea c•c]es, secoadevea•; so]id c•rc]es,•hi•d e•ea%s.
1st
EVENT 185
$ = 200 ø
1.09 108
18 3 ß
ß
ß
ß
ß
1.07
2nd
I
ß
'
ß
18.2
EVENT
I.I
108
1.09
I0 7
165o
108
106
o
s
I 07
S
ß
10.5
ß
106
I0 4
1.05
10.3
3rd
EVENT
1.76
22
1,75
22.5
1,74
o
o
22
ß
o ß
1.73 o
o
75
25
S
o
22.0
o
ß
21.75
1.71
I.......... -12 -9
-15
-6
-3
0
3
$
9
12
I
15
I
18
I ;d ;7
21
30
20.5
D
Fig. 3. Variation of S, standard error of •/x in percent, and e, standard deviation in seconds, as a function of D, the dip of the line element.
6086
Wu AND KANAMORI' RAT ISLAND EARTHQUAKE
source model with fault length, propagation velocity and directionof faulting as parameters. We have additionalguidancefor this interpretation at our disposal:(1) Stauder'spartial focal mechanism solutions for this main event (Figure 7a) and also complete solutions for many
large aftershocksin the seriesand (2) spatial distribution of aftershocks(Figure 7b).
• ' vo=20.8 sec 8/5
It shouldbe notedthat, because of the long-
= 140 km
periodwavesthat we usehere, the featuresderived concerningthe source are average properties. Temporally, we are not able to resolve
detailsbelow100 sec.The advantageof this
ß
r o = 14.7 sec
method is that a definite choice of fault plane
•
can be made; in fact, for large shallowthrust earthquakes this method is the only one available to resolvethe low-angle thrust plane. The best fit of the experimental data is presented
•A =86 kmß
in Figure 8a. In Figure.8b, theoreticalradiation patterns for rupture velocities of 3, 3.5, and 5 km/sec are compared with the pattern for 4.0 km/sec. The results obtained indicate that the fault plane strikes N70øW and dips 18ø in N20øE direction. The auxiliary plane constrained by Stauder's[1968b] partial solutionstrikesN60øE and dips 78ø in S30øEdirection. The fault
6.5 sec 55 km i,
,
i,,
,
C0S
0
Fig. 4. Least squares lines for the solutions. Here 0 is the angle between the line element IA and the ray leaving the source for the station. DIscUsSiON
lengthL i• about500km, the rupturevelocity
It may seem that the conclusions from the is about 4 km/see, and the seismicmoment Mo = 1.4 X 10• dynescm.If we a•sume,from studies above are contradictory. We have seen the aftershock area, the width W of the fault that, whereasthe body wavesindicatenot only plane to be 150 km, the average.dislocation a jerky propagationbut alsoa southwardpropagation,the long-periodsurfacewavesshbwa (u) on the fault can be estimated as (u) = smooth WNW propagation.These conclusions Mo/!•LW = 2.5 meters,where • = 7.0 X 10• for the multiple dynes/em• is used for an appropriate value of are not, however,inconsistent, the rigidity. The stress drop may be estimated, accordingto Knopo# [1958], by A•r ----8.l•(u)./,rW= 30 bars. The fault plane solutionobtainedhere is very similar
to that
obtained
from
the aftershocks
that occurredalong the island chain [Stauder, 1968b]. The slip angle measured clockwiseon the fault plane from the strike direction, N70øW, is 41.4ø. This means that the foot wall side (oceanicside) moved in about N30øW direction with respect to the hanging wall side (continental side). This slip direction is more or lessparallel to the slip directionsdetermined for the 1964 Alaskan earthquake, 1963 Kurile Islands earthquake, and the 1968 Tokachi-Oki earthquake [Kanamo.ri, 1971].
events in the initial 21 sec are not necessarily
relatedto the main propagation phase,which could have commencedsimult,aneously, and more smoothly,or 21 seeafter the initial break.
Because of thelargeamplitudes of thecomplex
long-period P waves, •theywerenotrecorded well on the WWSSN' seismogramsand hence
werenot subjected to studywith the sametech-
TABLE 1.
Multiple
Events
Error,
Event
1 2 3
D
6 -6 15
•
•
•o
200 165 140
1.1 1.2 1.7
6.5 14,7 20.8
33 86 140
Apparent Velocity, km/sec
6h
18.3 10.4 21.6
5.1 5.9 6.7
WU AND KANAMOn•' RAT ISLAND EARTHQUAKE •
o
6087
6088
Wu ANDKANAMORI' RATISLAND EARTHQUAKE
F1Oh •55m F1oh•55m NUR I10 cm VAL
ATU
PDA
DAI
I0
--,270
2o
901 PEL
180
AF I
,
;500
Fig. 6.
,
I
sec
Lovewave radiation pattern. A]]waves reduced to• ,: 9•r/2.
nique. These wavescould have their sources
uationis moreserious, because hereonlythe
lying alongthe path of propagation as de- firstfewcycles oftheP wave(orother phases) lineatedby the surfacewavemethod.That the
150- to 300-secsurface waves do not reveal
are used.As we havementioned before, the
USCGSv•bfor the Rat Islandmainshock was the sameinformation as the bodywavesis only 6.0 (USCGSSeismological BulletinMSI only natural, sincethesewavescannot discern
behaviorof shortduration.Thereforefrom the
290),whichimpliesthat the energyof this earthquake is lessthanabout11100,000 that
surface waves wecandeduce onlytheaverage of the1964Alaskan earthquake andabout1/
velocityof rupturepropagation togetherwith 100,000 that of the 1963Kurileearthquake therupturelengthbutnotthepossible accelera- [Gu•enbe'rg, 1956].Fromthe seismic moment tions and decelerations with a time scaleof
it is clearthat the Rat Islandearthquake is smaller thantheAlaskan earthquake, but not The multiple-event natureof largeearth- by 100,000timesin energy,and in fact is quakesmakesthe meaning of the magnitudeslightly larger thanthe1963Kurileearthquake. ambiguous. If M•. is used,then,depending on If themagnitude istorepresent thetotalenergy the proximityof the stationto a particular oftheearthquake, anappropriate sumofmagperhaps50 sec or less.
sectionof the fault, the magnitude will reflect nitudes of theindividual eventsshould beused.
the characteristics of that section, sinceby
Theideaofmultiple events hasbeen proposed
definition ML is determined by the maximum by seismologists forseveral largeshallow earthrecorded motionof a part of the seismogram quakes[Wyssand Brune,1967;Florenso•) and
andthenearest section of a faultislikelyto Solonenko, 1963], foronedeep earthquake [Fukao,
contribute mostof the energyowingto the 1972],andevenforearthquakes of magnitudes near-field effect.If mbisused,however, thesit- around 6 [Wu,1968; Niazi,1969]. Conceptually,
Wu AND KANAMORI' RAT ISLAND EARTHQUAKE 55÷
!
,
!
,
,
,
6089
,
!
54
œ
ß5
ß
55-
o
4
o
00
ß
0
0ß 3
NODAL
0
0
GO ø 00 0
GO
7ß ø
14
o•DOo
0 0
øQD•DO
0
GO %o
0
GO
o
o
8
o
o ß9
o
o0 0
GO øoo o
I0 0
o
0
o
0•0 0 ß6 00
ß11 00 ß0
0
00 00 0010 00 0
00
0
ß12
0% 0
o
o •16 o
00
0
o
o
oooOo o
o
o
00
0000
0
0
oø
0
0
o
0
0
o o 000
o.,,
ß ß15
0 0
oo •, 0
0ß2 0
0
o
(a) RAT
LAT
1965
o o
ISLAND
•
Main
&
m_•6
0
6>m_•5
AFTER
SHOCKS
shock
0
0
0
N
70
175
LONG œ
180
LONGITUDE
Fig. 7. (a) Large aftershocks.Numbers attached to the solid triangles mark the time sequenceof events with m• •_ 6 (from USCGS Preli-minary Determination of Epicenters). (b) Stauder's partial solution for the main shock.
there are two ways to view multiple events in terms of sourcespaceand time functions. One is
B**(•, t) (Figures 9c, d). In this casethere are 'dead' sectionsalong the fault during the main that the time futictionat eachpoint is uniform propagation phase, and aftershock could occur but consistsof a delayedsum of ramps (Figure at theseplaces.In reality it is highly possiblethat 9a). This ramp could propagate with a uniform both modesoccur during the main shockof the velocity. At each instant along the fault there Rat Island series. will be a numberof activesectionscorresponding If the earthquakeis composedsimply of a to the acceleratingand deceleratingparts of the propagating rupture, the durations of P waves ramp (Figure 9b). The other possibilityis that a in different azimuthal directions would be difpropagatingdislocationhas a changingdensity ferent. Thus in the direction of propagationat function;i.e., B(•), the Burgersvector[Weertman, large distance we would expect a duration of 1969],is alsoa functionof x, the coordinate along about 100 sec and in the opposite direction of the fault, B*(•, x), or a function of time t, about 150 sec. But on actual short-period
6090
Wv AND KANAMORI: RAT ISLAND ]•ARTI-•QUAKE
seismogramsfor the Rat Island earthquake we do not see such a variation.
N
The P waves in all
azimuths have decayingtails of about the same length. Thus a sequenceof aftershocksoccurring behind the main propagating dislocation line accompaniesthe main propagation, and the number
of these aftershocks
decreases ex-
ponentially as a function of time. For shallow large earthquakes behind the trench, body wave focal mechanismsusually fail to give a completesolution [Stauder, 1968a] becauseof the lack of readable close-inrecords, and hence it is difficult
to determine
a shallow
dipping plane. When they are combinedwith the surfacewave method, however,the parameters for this plane can be estimated. Stauder [1968a, b] has shown that in the Aleutian Island chain the events with
thrust
mechanisms
Fig. 8a. Preferredfit of the long-periodsurface wave data. Thin line representstheoreticalRay-
all have very similar high-angle planes. For
leigh wave data; heavy line, theoretical Love wave data; solid circles,observed'Rayleighwave
the
data; crosses,observed Love wave data.
1965 Rat
Island
series the same
All these events show a NE-SW
strike
is true.
and a
southwarddip of about 75ø. That this plane cannot be the fault plane for the main shock is establishedby lookingat the asymmetryof the Rayleighand Love wave radiationpatterns, which imply a NW rupture propagationdirection. Sincethe NE-SW plane is steepand dipping south,a nonstrikeNW directionof propagation is unreasonable,unless small segments of the fault line up en echelon;however,the continuoustrench and the shape of tim island arc do not agree with such geometry. BenMenahem and Rosenman [1972] adopted a high-angle solution for their interpretation of this earthquake;the strike of the planeis not consistentwith Staude'r's[1968a] solutionsfor earthquakes in this region. Stauder [1968a] also showedthat the shallowplane dips about
Both the P wave and the surface wave re-
sults yield rupture velocities.The multiple eventsimply velocitiesof 5.1-6.7 km/sec (Table 1). These valuesare higher than thoseof shear wave velocity and approach those of
15ø-20 ø in the direction of N50øW-N90øW for most of the thrust events in the aftershock
series. By using a shallow-angleplane consistent with the partial solutionof the main
shock,we foundthat the directionof propagation has to deviate slightly from the strike. However,sincethe dip is shallow,the 10ø or so deviationonly impliesa dip directionpropagationof about 85 km, which couldbe accommodated very well by the width of the fault plane. The same sort of deviation has been ob-
servedfor the 1964Alaskanearthquake[Kanamori, 1970b].
Fig. 8b. Other rupture velocities. Calculated radiation patterns with rupture velocities of 3.0 (R•, /_a; heavy dashed line), 3.5 (Re, La; medium dashed line), and 5 km/sec (Rs, I_•; thin dashed line).
WU AND KANAMORI' RAT ISLANDEARTttQUAKE (a)
6091
Finally, it shouldbe noted that the parallelismof the slip vectorsfor major earthquakes in the northern Pacific is conformable to the
Pacific basin'smoving as a rigid plate [McKenzie and Parker, 1967]. REFERENCES
Ben-Menahem,A., and M. Rosenman,Amplitude patterns
t=t 1
(b)
•o
n>
x =ct t
B(•',X•)
(c)
/"
of
tsunami
waves
from
submarine
earthquakes,J. Geophys. Res., 77, 3007-3128, 1972.
Florensov,N. A., and V. P. Solonenko,The GobiAltai earthquake, Izd. Akad. Nauk SSR, 1963. (English translation by Israel Program for Scientific Translations, p. 392, 1965.) Fukao, Y., Source processof a large deep-focus earthquake and its tectonic implications--The western Brazil earthquake of 1963, Phys. Earth Plan.et. Interiors, 5, 61-76, 1972.
Gutenberg,B., The energyof earthquakes, Quart. J. Geol. Soc. London, 112, 1-14, 1956. Johnson,T., F. T. Wu, and C. It. Scholz,Source parametersfor stick-slip and for earthquakes, Science,179, 278-279, 1973.
t=t I
(d)
t =t 2
Kanamori, H., Synthesisof long-periodsurface wavesand its applicationsto earthquakesource studies--Kurile Islands earthquake of October 13, 1963,J. Geophys.Res., 75, 5011-5028,1970a. Kanamori, It., The Alaskan earthquake of 1964: Radiation of long-period surface waves and source mechanism,J. Geophys.Res., 75, 50'295040, 1970b.
Kanamori, II., Focal mechanism of Tokachi-Oki
earthquakeof May 16, 1968,Tectonophysics, 12, Fig. 9. Generation of multiple-event type of seismograms.(a) Displacement of a point along the fault. (b) Particle velocity function alongthe fault at t = t for uniform propagating fault. This function will move down the line as t increases.(c) Burgers vector as a function of • and x, B(•, x•.) -- 0. (d) Dislocation along the fault at different t.
compressionalwaves. Recently, such values have been observedin the laboratory [Wu et a/., 19'72; Johnson et al., 1973] and attained theoretically (R. Burridge, personal communication, 1973). In both instances,a low frictional coefficientis deemed responsible.For the experiments,low friction was achievedthrough acceleratingcreep; it is conceivablethat this explanationis also applicable in real faulting. The surfacewave resultsindicatean average value of 4 km/sec for the whole fault. In view
of the complexprocessin the sourceregion,the meaning of this value deserves further consideration.
1-13, 1971.
Knopoff, L., Energy releasein earthquakes,Geophys. J., 1, 44-52, 1958.
McKenzie, D. P., and R. L. Parker, The North Pacific: An example of tectonicson a sphere, Nature, 216, 1276-1280,1967. Niazi, M., Source dynamics of the Daslit• Bayaz earthquake of August 31, 1968, Bull. Seismol. Soc. Amer., 59, 1843-1861,1969. Nowroozi, A. A., Terrestrial spectroscopyfollowing the Rat Island earthquake, Bull. Seismol. Soc.Amer., 56, 1269-1288,1966.
Stauder, W., Tensional characterof earthquake foci beneath the Aleutian trench with relation to
sea floor spreading,J. Geophys.Res., 73, 76937701, 1968a. Stauder, W., Mechanism of the Rat Island earth-
quake sequence of February 4, 1965, with relation to island arcs and sea floor spreading, J. Geophys. Res., 73, 3847-3858, 1968b. Wu, F. T., Parkfield earthquake of June 28, 1966: Magnitude and sourcemechanism,Bull. $eismol. Soc. Amer., 58, 689-709, 1968. Wu, F. T., K. C. Thomson, and H. Kuenzler, Stick-slip propagation velocity and seismic source mechanism, Bull. Seismol. Soc. Amer., in press, 1972.
6092
Wu ANI) KANAMORI: RAT ISLANI) EARTHQUAKE
Weertman, J., Dislocation motion on an interface with friction that is dependent on sliding velocity, J. Geophys. Res., 74, 6617-6622, 1969. Wyss, M., and J. N. Brune, The Alaskan earthquake of March 28, 1964, a complex multiple
structure, Bull. Seismol. Soc. Amer., 57, 10171023, 1967.
(Received June 26, 1972;
revised May 16, 1973.)
