JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 89, NO. B12, PAGES 10,233-10,248,NOVEMBER 10, 1984
Factors Affecting SeismicMoment ReleaseRates
in Subduction
Zones
ERIC T. PETERSON
Department of Geophysics,Stanford University,Stanford, California
TETSUZO SENO
International Institute of Seismologyand Earthquake Engineering,Building ResearchInstitute,
Ministry of Construction, Tsukuba, Japan
The amountof shallowseismicactivityin subductionzonesvariesgreatlyfrom regionto region.
We quantifythis seismicity
by calculatingseismicmomentreleaseratesand seismicslip ratesfor 24
subductionzones. To calculatethe momentrelease,we sumthe seismicmomentfor all interplate
thrust-typeeventswith surfacewavemagnitudeM s > 7.0, occurringfrom 1904to 1980. We present
a time history plot of the seismicmoment releasefor each subductionzone; these exhibit the
differences
in the seismicreleasepatterns.For subduction
zoneswherethe time windowof our study
is not representative,
the total momentreleaseis correctedusinginformationon repeattimes. The
momentreleaseratesare comparedwith varioussubductionparametersin order to determinewhich
factorsinfluencethe degreeof coupling. Theseparametersincludethe age of the subducting
lithosphere,absolutevelocitiesof the upper and subductingplates,convergence
velocity,and length, maximum depth,and dip of the Wadati-Benioffzone. The momentreleaserate decreases
asthe ageof the
subducting
lithosphere
increases,
whenthe zonesbelonging
to a singlesubducting
plateare considered. This ageversusmomentreleaserelationis consistent
for the zonesin whichthe Pacific,Cocos,
PhilippineSea,andIndianplatesaresubducting.
Themomentre!ease
ratesfor the subduction
zones
in whichthe Pacificplateis subducting
are muchhigherthan for zonesof otherplateswith similar
age. The age versusmomentreleaserelationholdsamongthe zoneswhich belongto one subducting
plate; however,zoneswith similar agesbut belongingto differentplates do not have the same
momentreleaserates. This suggests
that within a singleplate the age is the dominatingfactor
affecting
thestrength
of seismic
coupling
butthateachPlateasa wholehasa characteristic
moment
release
budget.Zoneswithretreating
upperplatestendto havelowermoment
release
rates.The
momentrelease
ratedoesnot increase
withconvergence
velocity;
no simplerelationship
wasfound
betweenthesetwo parameters.The momentreleaserate dependsmost clearlyon the age of the sub-
ductinglithosphere
and the absolutevelocityof the upperplate. Theseare the two independent
subductionzone parametersamongthe variablesconsidered.The other variablesdependon thesetwo
parameters."
of subduction zones in order to determine which factors
INTRODUCTION
influence
the degreeof seismic
coupling.ThesecharacterisMost of the world's large earthquakesoccur at shallow
depths in subduction zones. Many previous studies have
discussedthe differencesin the shallow seismicityamong the
various subduction zones [e.g., Kelleher et al., 1974;
Kelleher and McCann, 1976; Kanamori, 1977a; McCann et
al., 1979; Ruff and Kanamori, 1980, Lay et al., 1982]. The
study of the factors which influence the seismicityin subduction zonesis important to understandthe generationof
large earthquakes. It will also shed light on the coupling
tics are the age and absolute velocity of the subducting
plate, the absolutevelocityof the upper plate, the convergence velocity, and the length, maximum depth, and dip
angle of the Wadati-Benioff zone.
The relative plate motion in a subduction zone takes
placeas•eismic
slipor aseismic
slip, andbothtypesof slip
can occur in a single subductionzone. Since aseismicslip is
not accompanied
by a radiationof seismic
energy,it cannot
be measured directly, and we must infer that it occurs from
betweenplates, which influencesplate tectonicprocesses a study of the seismicslip (or lack of it). By comparingthe
such as island arc evolution and back-arc Spreading[Molnar
seismicslip rate to the relative plate velocity, the amount of
andAtwater,1978;UyedaandKanamori,
1979;Sykes
and aseismicslip can be estimated. A large proportion of
Quittmeyer, 1981; Lay et al., 1982]. The amount of shallow
thrust-type seismicity in a subduction zone reflects the
aseismic
sliprepresentsweakcoupli•ng
betweenthe plates.
degreeof seismiccouplingat the plate interface. We quantify the amount of interplateseismicityin a region by calculating seismicmomentreleaseratesand seismicslip ratesfor
zones in Figure 1 are chosenin such a way that the rupture
zones of known eventsdo not crossthe boundaries. Physi-
The locations of the boundaries between the subduction
cal features
suchas bendsin the trenchand intersecting
the 24 subducti0n
zonesshownin Figure1. We compare aseismicridges or sharp variations in the size of known
these two parameterswith other quantifiable characteristics events are used to define the subduction zone boundaries.
For example,the LouisvilleRidge intersectsthe trenchat
Paper number 4B0608.
theboundary
between
•heTongaandKermadec
subduction
zones,
andthezones
alongthewestcoastof South•America
arebased
ontherupturezonelocations
of largeearthquakes
0148-0227/84/004B-0608505.00
as describedby McCann et al. [1979].
Copyright 1984by the American GeophysicalUnion.
10,233
10,234
PETERSON
ANDSENO:SEISMICMOMENTRELEASE
IN SUBDUCTION
ZONES
SUBDUCTION
.-'
ZONES
Alaska
Aleutlan•-east
Aleutlan•-we•t
Japan
Izu-Bonin
-%
Marianas
Mexk•o
Central
Amerk•
½olomble
Peru-north
Java
Peru-south
•K
Tønga
Chile-north
Chile-central
Ch#e-.outh
Fig.1. Thelocations
andlength
of the24subduction
zones
considered
in thisstudy.Some
subduction
zones
were
notincluded
because
of lackof seismicity
(JuandeFuca)or complications
suchasthesubduction
of continental
lithosphere(New Guineaand Solomons).
EARTHQUAKESIN THE CONTACTZONE
occurin the thrustzone, we assumea maximumdepthof 60
km, due to evidence for thrust earthquakesin several
In order to calculate a seismic moment release rate and regions(Japan, Kuriles, and Kamchatka)with rupture zones
seismicslip rate for a subductionzone, we must sum the extendingto this depth [Veith, 1974; $eno and Pongsawat,
seismicmomentsfor the interplatethrust-typeearthquakes. 1981; Kawakatsuand $eno, 1983]. The maximum depth of
As our sourceof earthquakelocationsand magnitudes,we the thrust zone may be slightlyshallower,e.g., 40-50 km, in
use the catalog compiled by Abe [1981], which contains the Aleutians[Engdahl,1977;Houseand Jacob, 1983].
shallow (depth _• 60 km) eventsfrom 1904 to 1980 with a
surfacewavemagnitude
Ms -• 7.0. After searching
Abe's
SEISMICMOMENT
list for all shallowearthquakes
in a givensubduction
zone,
The seismic moment is the best measure of the size of an
we wish to selectonly the interplatethrust-typeevents.
Thereforewe eliminatethoseeventswith publishednormal earthquake because it does not suffer from saturation for
or strike-slip focal mechanisms. For those events without
published mechanisms,we use the location of the event as a
large events,as do the various magnitudescales. Also, the
moment can be added linearly, while the magnitudescales
selectioncriterion. Earthquakeswith epicentersseawardof are logarithmic. After compilinga list of the thrust events
the trenchor landwardof the volcanicfront are eliminated, in a givensubductionzone, we quantifythe interplateseissince they clearly do not occur in the thrust zone between micity from 1904 to 1980 by calculatingthe total seismic
the two plates. Some of the events included in the calcula- moment released;thus we must assigna seismicmoment
tionswhichdo not havepublishedmechanisms
may not be value to each event. There are severalmethods available to
thrust events. However, the larger events(which are more obtain the seismicmomentof an earthquake,but the most
important in the calculations)usually have known mecha- accurate method is to determine the moment from seismo-
nisms. Thereforethe error causedby includingsmall gramsof long-periodsurfacewaves,and we usesuchpubnonthrust events should be small.
Belowthe thrustzone,the slipbetweenthe platesis aseismic, and the earthquakes
that do occurat thesedepthstake
placewithin the descending
slab [Isacksand Molnar, 1971;
Hasegawaet al., 1978; Yoshii,1979]. The maximumdepth
of the contactzone is not well constrainedand probably
varies among the subductionzones. To selecteventswhich
lished values whenever available. General lists of earthquakes whose moments have been measured are found in
works by Kanamori and Anderson [1975], Kanamori
[1977b], Wang [1981], Lay et al. [1982], and Seno and
Eguchi [1983]. There are also references too numerous to
list whichcontainmomentdeterminations
for singleearthquakesor tectonicregions.
PETERSON
ANDSENO:
SEISMIC
MOMENT
RELEASE
IN SUBDUCTION
ZONES
change in the moment estimated from (1). Therefore we use
the equation
Thrust Eartlxluakes
I
I
10,235
I
x
logM 0 = 1.5Ms + 16.3
9.0
(2)
There are several problems with using Ms to determine
the moment of an earthquake: the saturation of the Ms
scale for large earthquakes, the inconsistentscalesused by
various researchersto find Ms, and the fact that small
errors in determining Ms produce large errors in Mo.
As the rupture length of large earthquakes exceeds 100
8.0
km, the seismic energy released and the seismic moment
continue to increase, but the amplitude of the 20-s surface
waves does not increasesignificantly; thus the Ms scale saturates for earthquakes with Ms > -8.0
[Geller, 1976;
Kanamori, 1977b]. As a result of the saturation, estimating
7.0
the moment from Ms for earthquakes with Ms >- 8.0 will
tend to underestimate Mo. In our study there are 28 earthquakes with Ms >- 8.0. However, for all but four of these
events, the moment has been measured from long-period
7.0
8.0
9.0
surface waves or estimated from the aftershock area, so it
ObservedMs
was not necessaryto estimate M0 from Ms for most large
Fig. 2. The data set for this figure is the thrust earthquakes
whichhavehadthe•eismic
moment
measured
directly
fromsurface events. Therefore saturation was not a major problem in
x
x
x
x
x
x
x
waves. On the horizontalscaleis the observed
M s, and the vertical our moment determinations.
Various catalogs of earthquake magnitudes have used
scaleis the value of M s determinedfrom the measuredmomentby
using (2) in the text. Equation (2) givesthe best fit, which is illus- different definitions for the surface wave magnitude, with
trated by the line.
values for a single earthquake varying by 0.2 units or more
[Geller and Kanamori, 1977; Abe, 1981]. Obviously, a consistent magnitude scale is needed when estimating the
A less accurate method is to estimate the seismic moment
from the fault area$ of the earthquake,usingan empirical moment from Ms. Abe [1981] has compiled a uniform
relationin whichM0 is proportionalto $3/2 [Abe, 1975]. magnitude catalog for earthquakes from 1904 to 1980 by
Kanarnori[1977b]hasestimatedthe momentfor many large recalculatingMs using amplitude and period data from variearthquakesby this method, approximating$ by the area ous sources, together with the original definition of Ms
of aftershocks soon after the main event. We use the seis- given by Gutenburg and Richter [1956]. We use this catalog
mic moments6btained by this method for eventswhich do as the sourcefor the Ms values used to estimatethe seismic
not have direct measurements of the moment from surface
moments.
Uncertainty in the value of Ms will result in a large
uncertainty in the value of the moment predicted by (2)
determinedby one of the two methodsabove, we estimate because of the logarithmic relationship. For example, a
M0 from the surfacewave magnitudeMs. This is the least value of Ms which is 0.1 units too high causesthe moment
accurate method of evaluating the seismic moment and is to be overestimatedby 40ø70. However, the percent error in
used only when the previously discussedmethods are the total moment released for a given zone will not be as
large as the single event error, since some events will have
unavailable.
Ms valuestoo large, while otherswill be too small. We will
The relationusedto estimateM0 from Ms is
waves.
For earthquakes whose seismic moments have not been
discuss the errors in more detail and calculate
log M0 = ams + b
where a and b are constants.
Kanamori
(1)
tor for each subduction
an error fac-
zone in a later section.
and Anderson
[1975] established a theoretical basis for a = 1.5 from
dynamicdislocationmodelsand similarity conditions. This
holds for earthquakesin the approximatemagnituderange
of 6.5 to 8.0, which includesthe great majority of earthquakesfor whichwe must usethis methodto estimateMo.
Previousinvestigatorshave used empiricalvaluesfor b of
16.0 or 16.1 [Thatcherand Hanks, 1973; Wang, 1981;Seno
and Eguchi, 1983]. We modify this value as follows. In
our studythere are 46 thrust earthquakeswith Ms between
7.0 and 8.0 whichhavehad their seismicmomentsindependently determinedfrom surface waves, the most accurate
method. We found that usingthe previousvaluesof b in
(1) underestimatedmost of the measured moments for these
MOMENT
RELEASE CURVES
In Figure 3 we present seismicmoment release curves for
each subduction zone.
Plotted on the vertical axis is a run-
ning total of the seismicmoments for all earthquakessince
the beginningof the time interval(1904) in units of 10•1
Newton-meters (N-m).
For comparison, an event with
Ms = 8 has a momentof 2 x 1021N-m using(2). These
plots show that the characteristics of the moment release
patterns vary greatly from zone to zone, and the total
moment releasehas a range of several orders of magnitude
among the zones(note the differencesin the vertical scales).
For example, we can clearly see the variation in seismic
events.By usinga leastsquares
methodrelatinglogM0 and character along the western coast of South America. In
Ms for these thrust earthquakes,we obtained a value of southern Chile the great 1960 Chilean earthquake comb = 16.3 (see Figure 2), which results in a significant pletely dominatesthe seismicmoment release,with the other
10,236
PETERSONAND SENO:SEISMICMOMENT RELEASEIN SUBDUCTION
ZONES
Chile-south
200
I
I
I
I
I
Ch•e-central
-
I
I
I
I
I
i
i
i
i
15
ß
E
150
o
10
.u. lOO
E
.E
,'o
;o
;teor
E
Z:
0.4
i
i
i
i
i
i
i
--
;;
i
10-
0
0.3
8-
E
o
•
..•
6-
0.2
E
4-
½e o.1
2
yeor
E
Peru-nor•
I
i
i
i
i
Cobmb•
i
--
1.0
0.8
10
-
0.6
0.4
5
0.2
I0
P-O 30
40
SO
SO
70
80
cjeor
Fig. 3.
Seismicmoment releasecurves for each subductionzone. The vertical axis showsthe accumulated seismic
moment
released
since
thebeginning
of thetimeinterval(1904)in unitsof 1021
Newton-meters
(N-m). Thehorizontal
axisgivestheyear(10 is 1910,etc.). The patternof momentrelease
variesfromzoneto zone. Alsonotethediffering
verticalscales,
whichshowthat theamountof seismic
momentrelease
variesby severalordersof magnitude.
earthquakesin this time period contributinglessthan 2ø70 However,this regionhad two largeeventsin 1868and 1877,
In this subduction zone we
so this zone is not as aseismicas implied by the moment
have the greatesttotal seismicmoment release,more than releasecurve. The timeperiodtreatedis not longenoughto
2 x 1023N-m.
include representative
earthquakesin this region, so we
The central Chile region has a total seismic moment excludethis zone from the comparisonwith other subducreleasean order of magnitudesmallerthan SouthernChile. tion zone parametersin a later section. Similar problems
Also, the moment is not dominatedby one earthquake;the may occur in other regions,and these are discussedin the
largestis the 1922 earthquakewhich takes up about 40070of next section.
the total moment. The moment release is fairly evenly
In Table 1 we give values of the total moment released
distributed in time.
from 1904to 1980, and the percentageof the total moment
of the total moment released.
The northern Chile zone has one of the smallest total
momentreleases
for thistimeperiod,almostthreeordersof
takenby the largesteventin eachsubduction
zone. The
three largest earthquakesof this century (Chile, 1960;
magnitude less than southernChile. Only two interplate Alaska, 1964; Aleutians, 1957) take up more than 90070of
earthquakes with Ms > 7.0 occurred from 1904 to 1980. the total moment releasein their zones. None of the other
PETERSONAND SENO:SEISMICMOMENT RELEASEIN SUBDUCTIONZONES
Central America
i
10,237
Mexico
i
E
lO
8
-
6
4
2
l0
•.0
30
40
SO
S
70
80
10
•0
30
40
geor
Alaska
I
I
I
I
6O
4O
I
I
I
--
--
70
i
i
80
i
i
i
i
i
4
.
--
-
_
-
_
-
_
-
_
-
_
-
--
-
--
-
2O
2O
S0
Aleutians-east
1
E
8O
SO
•j©or
__L
I
'.
10
•0
30
•
40
I
I
I
I
,
SO
S0
70
B0
IO
I
•0
I
I
I
I
I
I
30
40
SO
SO
70
80
veor
veor
Aleutians-west
i
i
i
i
i
Kamchatka
i
i
i
-
i
E
4O
15
3O
10
20-
10-
i
I
IOP.o
I
30
•jeor
Fig. 3.
subduction
zones
has an event
which
accounts
40
S0
S0
•eor
for
(continued)
more
than 80ø7oof the total moment. Such large events rupture
the whole length of the arc, releasingthe strain and resulting in a lack of moderate size events. These three subduction zones also have the largest total moment releasesof the
24 regions studied. Thus we see that regions with large
moment releases are characterized by very large events
rather than by an abundanceof moderate sized earthquakes.
In regions such as southern Peru, Central America, Mexico, Kuriles, Japan, New Hebrides, and Tonga, the cumulative moment is not dominated by one large event. The rupture zones of the largest earthquakesin these regions extend
only along a portion of the arc. These segmentsbreak at
different times, causing the moment release to be more
evenly distributed in time than the regions dominated by
one large event.
A moment release plot is not given for Java, because
only one earthquake with Ms -> 7.0 occurred in this time
interval. Also, the Izu-Bonin and Marianas regions are
combined
because of the lack of events.
ERRORS IN THE MOMENT
RELEASE AND CORRECTIONS
USING THE TIME-PREDICTABLE
The total
moment
MODEL
released in each subduction
zone from
1904 to 1980 (Table 1) is used to calculate the moment
release rate and the seismic slip rate, so we will discussthe
errors involved in using this quantity. These errors arise
from two sources: (1) the error in calculating the actual
total moment releasedin this time period and (2) the inadequacy of the 77-year time interval to represent long-term
moment
release rates.
10,238
PETERSON
ANDSENO:
SEISMIC
MOMENT
RELEASE
IN SUBDUCTION
ZONES
E
Kuriles
I
I
i
i
I
30
40
SO
I
I
I
I
I
I
15
10
E
4)
60
I
I
70
80
•leor
Ljeor
Nankai
E
10
i
Ryukyus
2.0
i
ß
-
-"
i
I
8
1.5
_
1.0
0.5
•rl i i
30
40
SO
60
I
I
70
80
F
-
i
;0
,0
I
i
I
I
i
•leor
E
I
1.0
I
I
I
I
'
0.8
0.6-
0.4-
0.2-
iI
I
Fig. 3.
(continued)
The uncertaintyin the total momentreleaseis duein part
tional to S 3/2. Moments determinedfrom long-periodsur-
to the errors in the seismic moment values for individual
face waves are the most accurate, and an estimate of the
to an error factorof
events. We assignan uncertaintyfactor to each event, typicalerror is 30ø7o.This corresponds
which dependson the method used to determineMo. 1.3. Usingtheseroughestimatesof the error for individual
Moments which are estimated from the surface wave magni- events, we calculate an uncertainty factor for the total
tudeMs (equation(2)) havethe largestuncertainty.We can moment released in each subduction zone (see Table 1).
get an ideaof this uncertainty
from the scatterof the data This is calculatedby taking the root mean squareof the
pointsin Figure2. In the leastsquares
fit betweenlogM0 errors in moment for each earthquake. The large uncerand Ms the standarddeviationfor a singleeventis 0.27 tainty in the total momentfor somezonesresultsfrom the
unitsof Ms, whichcorresponds
to an uncertainty
in M0 by lack of events(northern Chile, Java, Izu-Bonin, and Maria factor of 2.5 when using (2) to estimatethe moment. anas), or the unavailabilityof direct moment measurements
Thereforewe assignan uncertaintyfactor of 2.5 to these (easternAleutians,Philippines,Ryukyus). The small error
events. When the moment is estimatedfrom the aftershock in the Kuriles' total moment value (---13ø70)is due to the
area, we assignan uncertaintyfactorof 2.0. This results fact that all earthquakeswith Ms > 7.5 have had direct
from the fact that aftershockareas can usually be reliably measurements of the seismic moment.
determined to within -50ø70 or better, and M0 is propor-
Another source of error in calculating the total moment
PETERSONAND SENO:SEISMICMOMENT RELEASEIN SUBDUCTIONZONES
E
Sumatra
4
i
10,239
New Hebrides
i
! iI ! ! i
I
1o
po
I
I
I
I
I
I
30
40
50
60
70
80
I
1o
•0
30
Tonga
I
I
co0
I
i
•o
70
80
i
i
1
yeor
yeor
E
40
Kennadec
I
I
I
I
I
I
I
_
I
!0
•0
i
I
I
I
i
i
30
40
50
60
70
80
I
•.0
yeor
30
40
SO
60
7
8
yeor
Fig. 3.
(continued)
release is that earthquakes with Ms • 7.0 were excluded
from the calculations. The seismic moment released by
these events is small for most subduction zones, even
though these events are more frequent. The exclusion of
the smaller earthquakesaffects the total moment values by
more than - 10070only in those zoneswith very low seismicity (Izu-Bonin, Marianas, and Java). Since the total
moment release is very small for these zones, the relative
ranking of subduction zones in terms of total moment
releasedis not affectedby the exclusionof smaller events.
Reliable data on surface wave magnitudes and seismic
moments are not available for earthquakesoccurring in the
1800's or earlier. Our study includes only those events in
the 77-year period from 1904 to 1980. For some subduction
zones the shortnessof this time period may present a problem when calculating moment releaserates and seismicslip
rates. We want these quantities to representlong-term averages, but the seismicityin this 77-year period may not be
sufficientlyrepresentative. We now discussthis problem and
an attempt to correct it based on the time-predictable model
of earthquake occurrence.
In several regions (southern Chile, Alaska, eastern and
western Aleutians, Kamchatka, and Nankai Trough) the
rupture length of the largest event is a large fraction of the
total length of the arc, and typical repeat times are longer
than the 77-year interval of our data. Therefore our total
moment values for these zones will give moment release
rates which are larger than the long-term average. Using
the time-predictable model [Shirnazaki and Nakata, 1980],
we apply a correction to the total moment values for these
zones. The basic assumption of the time-predictable model
is that the amount of slip in a given event is proportional to
the time interval
between
this event and the next event to
occur in the same location. The amount of slip and the
time interval between events may vary with time at a given
location. The time-predictable model is favored over the
slip-predictablemodel [Shirnazakiand Nakata, 1980; Mogi,
1981; Sykesand Quittrneyer, 1981]. Sykesand Quittmeyer
studied six regions which have had three or more large historical shocks. They found that the time interval predicted
by the amount of slip in the event at the beginningof the
time interval was in error by less than 50ø7ofor all cases.
The results using the slip-predictable model were not as
good.
Since the seismicmoment is proportional to the amount
of slip, we assumethat the moment of an event is proportional to the time period between this event and the next
event. Therefore, dividing the seismicmoment of an event
by the time interval to the next event will give a time average of the moment release. We apply this method to the
subductionzones listed above: the moment of the largest
event is multiplied by the factor (77-years/time interval).
The
values of the "corrected"
total
moment
release are
given in Table 1. In severalcases,the evidencepoints to a
similar amount of slip in the event precedingand the event
ending the time interval. In these caseswe use the moment
of the later
event because it is better
constrained.
The
events used and the correspondingtime intervals are southern Chile, 1960, 124 years; Colombia, 1906, 74 years; western Aleutians, 1965, 120 years;Kamchatka, 1952, 108 years;
10,240
PETERSON
ANDSENO:SEISMICMOMENTRELEASE
IN SUBDUCTION
ZONES
TABLE
Subduction
Zone
Chile,south
Chile,central
Chile,north
Peru,south
Peru,north
1.
Seismic Moment
Total
Corrected
Moment,
Total Moment,
1021N-m
1021N-m
203.
17.4
0.36
10.4
1.03
Release
Error Factor
LargestQuake ø7o
127.
-
1.30
1.47
2.06
1.40
1.67
98
40
50
26
25
Colombia
Central America
Mexico
Alaska
13.9
4.35
10.2
87.4
11.4
62.9
1.23
1.44
1.37
1.28
72
17
20
94
Aleutians,east
Aleutians,west
60.8
19.2
43.4
10.8
1.96
1.39
96
65
Kamchatka
Kuriles
45.8
20.2
35.7
-
1.25
1.14
76
33
Japan
8.84
-
1.29
32
Nankai
9.88
8.67
1.26
54
Ryukyus
Philippines
1.96
8.70
-
1.82
2.01
37
66
Izu-Bonin
Marianas
Sumatra
0.68
0.34
3.98
-
1.99
1.95
1.52
53
53
18
0.06
-
2.50
NA*
-
1.46
26
Java
New Hebrides
11.0
Tonga
7.01
-
1.35
33
Kermadec
4.55
-
1.73
32
Total moment' the total momentreleasedfrom 1904 to 1980; correctedtotal moment:total momentcorrectedfor
repeattimes(seetext); errorfactor:thetotalmomentis in errorby thisfactor(seetext); largestquake%: thelargest
quake takes up this percentageof the total moment.
* Not applicable'
OnlyoneeventwithMs>_7.0 occurred
in Java.
and Nankai Trough, 1944-1946, 92 years. The time intervals were taken from Sykesand Quittmeyer [1981] and Lay
et al. [1982].
Recurrence intervals are not well known for Alaska
the eastern Aleutians
and
because of the shortness of the histori-
cal record. A lower limit of 100 years was used for the time
interval in these cases. This will produce an overestimateof
small compared to the arc length and rupture zones often
overlap. This makes it difficult to define repeat times.
However, repeat times for these small rupture zones are
usually less than our 77-year time interval, and there are
many events along the whole arc, so our time window is
sufficient. No correctionswere applied to thesezones.
SEISMIC SLIP RATES
the seismic moment release rate for these two zones if the
recurrenceintervals are longer. In the Kuriles the are is segmented into several rupture zones which overlap very little.
Earthquakes occur in a coherent manner, and the average
recurrence interval is close to 77 years [Utsu, 1968], so no
correction was applied for this zone.
In regions with small moment releases(northern Chile,
northern Peru, Ryukyus, Izu-Bonin, Marianas, Sumatra,
and Java) there is the possibility that large eventssometimes
occur, but no such events happened to fall in our 77-year
time window.
In such cases the moment
The seismicmoment for a single earthquake is defined by
Mo -- I• 2•' S where /• is the rigidity of the material surrounding the fault, •' is the averagedisplacementon the
fault plane, and S is the area of the fault plane. This
definition can be generalizedto a whole subductionzone,
for which we sum the moments for all interplate events to
obtain the averageseismicslip:
• Mo= tt15 S
release rates will
be underestimated. As discussedin the previous section, the
northern Chile region had two large events in the 1800's,
and we exclude this region from future discussionsbecause
of the unrepresentativetotal moment releasevalue. There is
some evidencethat large events occurred in Sumatra in the
1800's [Newcomb and McCann, 1982]; thus the seismic
moment release rate for this zone may be underestimated.
For the remaining zones with low seismicityin this century,
no large historicalearthquakesare known, and it is believed
that the low seismicityis a long-term tectonic characteristic
of these subduction zones [Kelleher and McCann, 1976;
Seno and Eguchi, 1983].
In the other subduction zones the rupture lengths are
(3)
/5 is now the amount of slip betweenthe two plates aver-
agedoverthe wholesubduction
zone,andS is the areaof
the contactzone describedearlier. Equation (3) can be used
to estimatethe seismicslip rate V• in a subduerionzone by
considering
the total momentreleasedin a giventime interval T'
V• =
Ev/0
(4)
In our estimatesof the seismicslip rate for each subduction zone, we use the corrected total moment values for the
interval 1904-1980 from the previous section (Table 1). As
PETERSON
ANDSENO:SEISMIC
MOMENTRELEASE
IN SUBDUCTION
ZONES
TABLE 2.
Zone
•arn•,
Chile, south
Chile, central
Chile, north
Peru, south
Peru, north
1000
1480
820
1100
900
Colombia
{•,
degrees
SeismicSlip Rates
Vs,
mm/yr
Sykes
Frei
'
mm/yr
15
15
20
20
20
143
13
0.7
14
1.7
91
92
92
89
84
1.57
0.14
0.01
0.16
0.02
650
20
26
79
0.33
CentralAmerica
Mexico
Alaska
Aleutians,east
Aleutians,west
Kamchatka
Kuriles
Japan
Nankai
Ryukyus
Philippines
Izu-Bonin
Marianas
Sumatra
1580
1150
1350
1200
740
1280
1210
650
750
1300
1440
1240
2150
1930
40
40
15
25
25
30
30
25
25
25
30
20
20
20
7.7
25
52
66
27
60
36
25
21
2.8
13
0.8
0.2
3.1
81
65
68
79
85
90
100
105
42
60
ND*
65
44
68
Java
New Hebrides
Tonga
1800
1800
1450
20
35
20
0.0
15
7.2
75
95
90
0.10
0.38
0.77
0.84
0.31
0.67
0.36
0.24
0.50
0.05
ND*
0.01
0.01
0.05
0.00
0.16
0.08
700
20
9.6
74
0.13
Kermadec
10,241
0.88, 0.91, 0.62
0.56, 0.32
0.48
0.86, 0.92
1.05, 0.68, 0.30
0.38, 0.29, 0.35
Larc:
horizontal
length
ofthearc;8:dipof slab
indepth
range
0 to60km;Vs: seismic
sliprate;Vrel:
relative
platevelocity
averaged
overthearc;a: theratioVs / Vr½
1; Sykes:
values
ofa fromSykes
andQuittmeyer
[1981].
* ND: Vr½
1 is notwelldefined
for thePhilippines,
asexplained
in thetext.
an averagevalue of /• for the depth range of the contact ringontheeastandthewestsides
of thePhilippine
Islands,
zone, we use/• = 5 x 1020N/m e. The area of the contact sothe Philippineblockmay be decoupled
from the Eurasian
zone is approximatedby
S = Larcd sin/J
(5)
whereLa•cis the lengthof the subductionzone arc, d is the
depth extentof the contactzone (assumedto be 60 km for
all zones),and/• is the dip of the Wadati-Benioffzone from
0 to 60 km depth,measured
from seismicity
profilesfrom
plate[Seno,1977]. Sincethe upperplatevelocityis not
known,wecannotdefinea gre•for thisregion.
Theratioof theseismic
sliprateto therelative
platevelocityis a = g•/ grd; valuesfor eachzoneare givenin
Table2. Interpreted
in termsof interplate
coupling,
a high
valueof c• (c• *-1.0) represents
strongseismic
coupling,
whilea lowc•indicates
a largeamount
of aseismic
slip. The
four zoneswith the largestvaluesof c• (southern
Chile,
various sources. The errors in g• for each zone will be
l•ger than the error factorsfor the total momentrelease Alaska, easternAleutians,and Kamchatka)have several
listedin Table1, dueto theuncertainties
in g and$ in (4), other seismiccharacteristics
in common: large total
andthe inadequacy
of the 77-yeartimeintervalin represent- momentreleases,large maximumearthquakesize, moment
inglong-term
sliprates.
releases
dominated
by oneevent,rupturezonelengths
comThe calculated
•eismicslipratesandthe valuesusedfor parableto La•½, and long recurrence
intervals.The value
L•c and • are givenin Table2. g• variesby severalorders c• = 1.57 for southernChile is physicallyUnreal,as we
of magnitude,rangingfrom 143 mm/yr in southernChile to wouldnotexpect
theseismic
sllprateto begreaterthanthe
lessthan 0.1 mm/yr in the Javaregion. It is interesting
to relative
platevelocity.Thisdiscrepancy
maybedueto any
comparethe seismicslip rate of each zone to the relative of the following factors: overestimationof the moment of
plate velocity VrelThe relativeplate velocityvariesfrom one the 1960 Chilean event, underestimationof the recurrence
end of an arc to the other, so we have calculatedvalues 0f
interval
for thisevent,or uncertainty
in theshallow
dip
gre•averaged
overthelengthof eachzone(Table2), using angle(usinga valueof • = 10ø instead
of 15ø wouldgive
the rotation poles from the RM-2 model of Minster and an c• of 1.05). As Sykesand Quittmeyer
[1981]pointout,
Jordan [1978]and the rotationpole for the PhilippineSea the calculationof seismicmoments from surface waves
plate from Seno[1977]. Vre•represents
the motionof plates becomes
moreuncertain
for faultswithshallow
dips.
on a time scaleof millionsof years,while g• is estimated
Zones with very low valuesof c• (northernPeru,
from data covering--100 years. In order to interpreta Ryukyus,Izu-Bonin,Marianas,andJava)arelikelyto have
comparison
of g• and gre•in termsof coupling,we must aseismic
slipthroughout
the contactzone,but regionswith
moderate
valuesof c• mayhavevariousinterpretations
of
motionof plateson the scaleof hundreds
of years,although the natureof the aseismic
slip:
this is uncertain.
1. In thesezones,thearcmaybesegmented
alongstrike
Note that Vre•is not included for the Philippine subduc- into regions
with highseismic
slipratesand regionswith
assume that Vre•gives an accurate measure of the actual
tion zone in Table 2.
This is because subduction is occur-
low rates.
10,242
PETERSON
ANDSENO:
SEISMIC
MOMENT
RELEASE
IN SUBDUCTION
ZONES
TABLE
Zone
Cs
Cc
Cn
Ps
Pn
Chile, south
Chile, central
Chile, north
Peru, south
Peru, north
Co
CA
COlombia
Central America
M,e
Mexico
3.
Moment
Release Rates and Subduction
Zone Parameters
MRR,
Vslal•, Vuppe,
r,
Vconv
,
Age, zmax' Lslab'
1016
N-m/km/yr
mm/yr
mm/yr
mm/yr
10
6yr km km
165.8
15.3
0.6
12.2
1.5
62
57
61
53
53
26
28
30
26
32
88
84
91
79
84
15
45
51
42
30
180
190
320
200
200
530
760
700
630
950
22.8
3.6
35
73
26
8
61
81
15
30
240
200
610
280
AI
Alaska
60.5
11.5
42
21
63
68
40
250
150
250
At
Aw
Aleutians, east
Aleutians, west
46.9
18.9
63
22
9
10
71
32
60
60
260
250
325
300
Ka
Ku
Kamchatka
Kurile
36.3
21.6
87
88
3
2
89
90
80
110
300
610
440
780
1200
61
8
10
350
Ja
Japan
17.7
103
2
105
125
530
Na
Nankai
15.0
39
0
39
25
70
170
Ry
Ryukyu
2.0
60
290
450
Ph
Philippines
7.8
47
IB
Ma
Su
Jv
NH
Izu-Bonin
Marianas
Sumatra
Java
New Hebrides
0.7
0.2
2.7
0.0
8.0
To
Tonga
6.3
99
Ke
Kermadec
8.4
92
57
- 1
56
58
ND*
ND*
93
107
6'2
77
30
- 37
-64
- 2
- 3
60
200
400
56
43
59
74
90
140
160
55
130
37
540
700
250
660
290
810
810
500
870
350
- 10
89
100
700
970
- 19
73
100
660
920
MRR:moment
release
rate; V.•la
b: absolute
velocity
of thesubducting
platenormalto thetrench(subduction
rate),
averaged
along
thearc;Vupper:
'fi•solute
velocity
of theupper
platenormal
to thetrench,
averaged
along
thearc;
Vconv:
relative
platevelocity
normal
tothetrench
(convergence
rate),averaged
along
theare;Age:ageof thesubductinglithosphere,
averaged
alongthearc; Zmax:
maximum
depthof seismicity;
Lslab:
down-dip
lengthof theWadatiBenioff
zone.
* ND: Vuppe
r forthePhilippine
subduction
zone
isnotwelldefined.
2. The degree of coupling may vary with depth in the
contact zone [Sykes and Quittrneyer, 1981; Ruff and
Kanarnori, 1983].
meters, we representthe degree of coupling by the seismic
moment release rate per kilometer per year (hereafter
referred to as the moment release rate). This will give us a
parameter which can be used to compare the amount of
seismicity in the various subduction zones. The moment
releaserate is the correctedtotal moment value (Table 1) for
3. Seismicand aseismicslip may vary in time at a given
place along the arc, with aseismicslip occurring in the time
interval between earthquakes. In cases 1 and 2, the values
of c• for the whole subductionzone would representa spa- the time interval 1904-1980,divided by the length of the
tial average of the coupling strength, while in case 3 it time interval T, and the horizontal length of the subduction
would represent a time average. The seismic/aseismicslip zone arc L•c:
in some zones may be a combination of all three possibili(6)
momentrelease
rate = •MoL
ties.
We compare our values of c• with those obtained by
Sykes and Quittrneyer [1981] in Table 2. They estimated
seismicslip rates from the coseismicdisplacementdivided by
the time interval to the next large event in a given region.
In general our values are in good agreement with theirs,
consideringthe errors inv,olvedin the two methods.
In regions with well-defined repeat times and accurate
measurements of the slip in large events, the method of
Sykes and Quittmeyer may be more appropriate. However,
for the zones where these quantities are not available, or for
zones where the seismicityis not dominated by large events,
our method of summing the moment of all events will be a
more appropriatemethod to obtain the seismicslip rates.
MOMENT RELEASE RATES
In our study of the relationship of seismiccoupling in
subduction zones to the various subduction zone para-
The moment release rates are listed in Table 3.
We use this
quantity, rather than the seismicslip rate, to representthe
coupling becauseit involves less uncertainty. The moment
releaserate is essentiallythe same as V•, exceptthat it does
not involve the variables d (depth of the contact zone), /5
(dip angle of the contact zone), and /z (rigidity). Thus we
avoid the uncertainties involved in these variables by using
the moment
release rate.
AGE AND MOMENT
RELEASE RATE RELATION
In Figures 4a and 4b, we show the relationship between
the moment release rate and the age of subducting lithosphere for each subduction zone. The age of subducting
lithosphere is the age of the oceanic plate at the trench,
averagedalong the length of the trench (Table 3). The ages
are based on magnetic lineations from the Plate Tectonic
Map of the Circum-Pacific Region [Addicott and Richards,
PETERSON
AND
SENO:
SEISMIC
MOMENT
RELEASE
INSUBDUCTION
ZONES
Ryukyus,
andPhilippine
zonesj,
theCocos
plate(Mexico
Age of SubduclhgLithospice
I
and Central American zones), and the Indian plate (Suma-
I
60
_
o
t0,243
Pacific
•:) cocos
/•
PhilOpine
[]
Indlen
tra, Java, and New Hebrides, zones), the moment release
rate strictly decreaseswith increasing age. The age and
moment releaserate relation for each plate lies in a different
place on the plot. This indicates that the absolute value of
the coupling is influencedby some characteristicof the plate
as a whole, while among the zones of a single plate, the relative degree of coupling is influencedby the age of the subducting lithosphere. Each individual plate seemsto have a
40
characteristic
momentreleasebudgei.
The zones in which the Pacific plate is subducting
(Alaska, easternAleutians, westernAleutians, Kamchatka,
Kuriles, Japan, Izu-Bonin, Marianas, Tonga, and
20
,.
Kermadec) also show a remarkably linear decrease 0¾
moment
release
ratewithage,withtheexception
of th?•&e
Age (my)
zones. The western Aleutians have a moment release rate
AgeofSubducting
Lithosphe•'e
i
150
100
50
lessthan expected
from the generaltrend. The relativeplate
motion direction in this zone is subparallelto the trend of
I
the trench, resultingin a large componentof strike-slip
motion. The relative velocity averagedover the zone is 79
mm/yr, while the relative velocity normal to the trench is
-
-
only 32 mm/yr, much lessthan in the other Pacific plate
zones. This is probablythe causeof the relativelylow total
moment releaserate from thrust earthquakesin this zone.
-
The Tonga and Kermadec arcs also have moment release
rates below the general trend for the zones in which the
Pacific plate is subducting. This may be due to the absolute
velocityof the overridingplate. For the sevenzoneson the
main trend the velocityof the overridingplate normal to the
trenchis 10 mm/yr or lesstowardthe subduCt•ing
plate.
However, for Tonga and Kermadecthis velocityis 10 mm/
yr and 19 mm/yr,respectively,
awayfromthe subducting
20
40
60
plate. So the seismiccoupling may be reduced in subduc-
Age (my)
tion zoneswhichhaveretreatingupperplates,[e.g., Uyeda
Fig. 4. Moment
release
rateversus
ageofth•subducting
litho- and Kanarnori,1979]. This generalization
is further supsphere. The moment releaserate is definedin (6) and has units of portedin the discussion
of the effectof the upperplatevelo1016N-m/km/yr:values
arelistedfor eachzonein Table3. Each city on the momentreleaserate (Figure6).
two-letter symbol correspondsto a subduction•'zonein Table 3.
The subductingslab in each zone belongsto the oceanicplate repre-
The zonesin whichthe Nazcaplateis subducting
(southsented
by thesurrounding
shape:
hexagon,
Pa•iOcplate;diamond, ern Chile, centralChile, southernPeru, northernPeru, and
Cocosplate;triangl
e, Philippine
platei'square,
Ifidianplate;circle,
Nazca plate. The ages used are listed in Table 3. (a) The zones of
the Pacific, PhilippineSea, Cocos,and Indian platesare connected
by lines to illustrate the relationship for zones belongingto one
plate. These plates have a decreasein the moment releaserate with
increasing age. (b) The zones of rthe Nazca plate do not show a
simplerelationship
between
the momentrelease
r•te andtheage.
Alpha vs. Age
I
I
0.8
•
coco.
0.6
198i],andagesgivenbyMolnarandAtwater[1978].
When all subductionzonesare considered,there doesnot
Seemto be a simplerelationshipbetweenthe momentrelease
0.4
rate and the age. The only observationwe can make is that
zones,with agesgreater than 100 million years tend to have
low momentreleaserates, but amongthe zoneswith ages
less than 100 million years there are both high and low
0.2
moment release rates. However, when the subduction zones
related to a single oceanicplate are considered,there is a
nearly•linear relationship between the moment releaserate
and the age, with the moment decreasingas the age
increases.This is shownin Figure 4a by connectingthe
50
100
150
Age (my):
Fig. 5. The relationship
between
ot = Fs / •a andthe ageof
the subductinglithosphere.The value of ot decreases
with increas-
points for all zonesin which the slab belongsto the same ing agefor tF•e.
subduction
zonesbelonging
to a singleplate. (For
oceanicplate. For the PhilippineSeaplate(Nankai Trough, otherdetails;•eeFigure4.)
10,244
PETERSON
AND SENO:SEISMICMOMENTRELEASEIN SUBDUCTION
ZONES
Velocity of Upper Plate
I
I
I
I
I
6O
¸
i
Plciflc
•> coc.
i
/•
and $eno [1977]. These velocities were averaged along the
lengthof eachzoneto obtainthe parameterVupper
listedin
I
philippine
[] Indlon
Table 3. As discussed earlier, the Philippine subduction
zone doesnot have a well-definedupper plate velocity and is
excluded from the analysis. Figure 6 shows the relation of
the momentreleaserate to Vuppe
r. Negativevaluesof Vuppe
r
correspond to upper plates which are retreating from the
oceanic plate in an absolute reference frame. Those zones
4O
with negativevaluesof Vuppe
r tend to have low moment
releaserates. The 12 zones with the highest moment release
2O
ratesall have positivevaluesof Vuppe
r (upperplatesmoving
-
-6
-4
-2
2
4
vetoc•y (cm/yr)
Fig. 6. Moment release rate versus absolute velocity of the
upper plate normal to the trench. Zones with retreating upper
plates (negativevalues) tend to have much lower moment release
rates than thosewith advancingupper plates. The dashedline separates zones with retreating and advancing upper plates. Southern
Chile is not shown in Figures 6, 7, 9, 10, and 11 becauseits large
moment releaserate is off scale. (Values of the velocity are listed in
Table 3. For other details, see Figure 4.)
toward the oceanic plate). Thus the degree of seismiccoupling is related to the direction of motion of the upper
plate, with retreating plates causing weak coupling, and
advancing plates resulting in strong coupling. This supports
the ideas of Uyeda and Kanamori [1979] that retreating
upper plates result in weak coupling, and this is elaborated
on in the discussion section.
The absolute velocity of the subducting plate normal to
the trench (subduction rate) was calculated and averaged
along each zone to obtain the parameter Vs•ab,with values
given in Table 3. Figure 7 showsthe relationshipbetween
Vs•ab
and the moment releaserate. From this figure we see
that the relationship between the subduction rate and the
moment release rate is similar to the relationship between
the age and the moment releaserate (Figure 4a). That is,
Colombia) do not exhibit the linear relation between among the subductionzoneswhich belong to a single plate,
moment releaserate and age, as shownin Figure 4b. This the moment release rate decreases as the subduction rate
may be dueto the dii•culty in assigning
an ageto thesesub- increases. However, we feel that the more basic relationship
duction zones. The magneticlineationsare not parallel to is the one between the age and the moment release rate.
the trench, so the age of subductinglithospherevaries This is becausethe subduction rate is dependent on the age
greatlyfrom one end of a zoneto the other. The lineations of subductinglithosphere. Figure 8 shows the subduction
are also offsetin severalplacesby transform faults, and an rate and age for each zone, and we seethat the subduction
averageage along the trenchmay not have any physical rate increasesas the age increases. Carlson et al. [1983]
significance.The Nazca plate zonesalso have subducting have studied this relationshipin greater detail, and they find
aseismicridgeswhich may alter the momentreleasevalues. that the subductionrate is highly correlated with the square
Note that northern Chile is not shown on the plot due root of the age of subducting lithosphere. Therefore the
to the uncertaintyin its moment releaserate, as discussed relationship observed between the subduction rate and the
earlier.
The Philippine Sea, Cocos, Indian, and Pacific plates
show a definiterelationshipbetweenthe momentreleaserate
and the age for their subductionzones. The Pacific plate
trend hasa muchhighermomentreleasefor mostof the age
rangethan the other plates. Possiblereasonsfor thesetwo
observationsare givenin the discussion
section.
Figure5 showsthe relation betweena (= seismicslip
rate / relativeplate velocity)and the age. Again, there is a
Subduction Rate
• 60
•
releaserates in Figure 4a.
•- Indlen
(• No=co
4o
n-
20
_
'\
E
o
-
MOMENT RELEASE RELATION TO PLATE VELOCITIES
•_• philippine
•
consistentdecreasein the degreeof seismiccoupling, representedthis time by a, with increasingage of the subducting
lithospherefor the zones of the Philippine Sea, Cocos,
Indian, and Pacificplates. The valuesof a for the Pacific
plate are consistently
larger than those for the zonesof
other plates with comparableages, as were the moment
•> coco.
\
2
4
•
6
8
10
Velodty (cm/yr)
In the previous section it was mentioned that the absolute
Fig. 7. Moment release rate versusabsolute velocity of the subvelocity of the upper plate may influencethe degreeof seisducting plate normal to the trench (subductionrate). The moment
mic coupling. We calculated the absolute velocity of the releaserate clearly decreaseswith increasingsubductionrate for the
upper plate in the direction normal to the strike of the
trench from the rotation polesof Minster and Jordan [1978]
zones of the Pacific, Cocos, and Indian plates. (Values of the velocity are listed in Table 3. For other details, see Figure 4.)
PETERSON
AND SENO:SEISMICMOMENT RELEASEIN SUBDUCTION
ZONES
moment releaserate seemsto result from the dependenceof
the subduction rate on the age.
The relative plate velocity normal to the trench (conver-
ConvergenceRate
60
Pucl/Ic
gencevelocity Veonv)
was also calculatedand averagedalong
Cocoo
each arc (Table 3). We expected that the moment release
rate would increase with V•onv,but the convergence velocities do not show a simple relation to the moment release
rate (Figure 9). There may even be a slight tendencywithin
a single plate for the moment release rate to decreasewith
increasing convergencerate. Even for zones in a small age
range we do not find a simplerelationshipbetween V½onv
and
the moment
release rate.
This
differs
from
Indian
Nazca
40
the results of
Ruff and Kanamori [1980], who representthe degreeof coupling by the maximum earthquake magnitude in a
subduction zone rather than the moment releaserate. They
found that there is a relationship between the convergence
rate and the maximum earthquake magnitude for zones with
similar ages. Reasonsfor this differencewill be given in the
discussion section.
MOMENT
10,245
RELEASE AND OTHER SUBDUCTION ZONE
2
4
6
8
10
Vek)city (cm/yr)
Fig. 9. Moment release rate versus relative velocity normal to
the trench (convergencevelocity). (Values of the velocity are listed
in Table 3. For other details, seeFigure 4.) There appearsto be no
relationshipbetweentwo parameters,even when zones of a single
plate are considered.
PARAMETERS
rately. The moment release rate decreasesconsistentlywith
increasingZmaxfor only the Cocos and Philippine Sea plates.
The Nazca and Pacific plates show a lessconsistentrelationship betweenthe moment releaserate and Zmax
or Lslab.
It has been shown that zones with older subducting lithosphere have seismicityat greater depths and longer WadatiBenioff zones [Vlaar and Wortel, 1976; Molnar et al., 1979].
In Figures 12 and 13, we see a fairly linear increasein Zmax
and Ls•abwith the age of the subductinglithosphere. This is
due to the fact that older lithosphere is cooler and takes
longer to heat up to the critical point where earthquakesno
longer occur. Therefore the relationship observed between
Lslaband the moment releaserate is a side effect of the relashould be noted that there are other zones with short
tion between Ls•aband age. The same is true for Zmax.The
Wadati-Benioff
zones which have small seismic moment
release rates. For the zones of the Cocos, Philippine Sea, dip angle of the slab was also consideredas a possible facand Indian plates, the moment releaserate strictly decreases tor affecting the coupling, but no correlation was found
with increasing Lslabwhen each plate is considered sepa- between the dip angles and the moment releaserates.
The maximum depth of seismicity, Zmax,and the downdip length of the slab, Lslab,for each zone are listed in
Table 3. These values are taken from Isacks and Barazangi
[1977], Yokokura [1981], and other sources and represent
the extent of the slab as evidencedby seismicity. Therefore
the slab may extend deeper and be longer than these values
indicate. The relationship of these two parameters to the
moment release rate is shown in Figures 10 and 11. The
four zones with the largest moment release rates (southern
Chile, Alaska, eastern Aleutians, and Kamchatka) have
maximum depths of seismicity less than 300 km and
Wadati-Benioff zone lengths less than 550 km. However, it
DISCUSSION AND CONCLUSIONS
SubductionRate vs. Age
lO
We have calculated
moment
release rates for 24 subduc-
tion zones. This parameter enables us to compare the
amount of seismicity and the degree of seismic coupling
among the subduction zones. The moment release rates
vary by several orders of magnitude, which indicates that
the mode of plate interaction varies greatly from zone to
zone.
Ruff and Kanamori [1980] representedthe coupling by
the magnitude of the largest earthquake in each zone. We
Puclhc
feel that the moment release rate gives a better representaCOCOS
tion
of the seismic coupling for some subduction zones,
Phlllj)plne
Indian
becausethe moment releasedby earthquakesother than the
largest event can also provide information on the coupling.
For example, for 12 of the 24 subductionzones studied, the
moment of the largest earthquake from 1904 to 1980 was
I
I
I__
50
100
150
lessthan 40% of the total moment releasedduring that time
Age (my)
period (Table 1). The magnitude of the largest shock is less
Fig. 8. Subduction
rate versusage of the subducting
litho- sensitiveto the variation in seismicity among subduction
sphere.The subduction
ratetendsto increase
with the age. (The zonesthan the moment releaserate. For example, two subvaluesof the two parametersare listedin Table 3.)
duction zones may both have a maximum earthquake mag-
10,246
PETERSONAND SENO:SEISMICMOMENT RELEASEIN SUBDUCTION
ZONES
Maxiaxrn Depth of Seismicity
Depthof Seismicity vs. Age
I
I
•
60
pacific
Cocos
600
Phlllpplnn
pacific
•
cocos
i
/• Philippine
[] Indian
I.dla.
0 Nazca
Nazca
40
0
_
400
•
20
200
4••-'•-•
-
E
o
i
i
50
100
Depth (kin)
i
150
Age (my)
Fig. 10. Moment release rate versus the maximum depth of
seismicity. The moment releaserate decreaseswith increasingmaximum depth of seismicityfor the zones of the Philippine Sea,
Cocos, and Pacific plates. Zones of the other plates do not show a
simplerelationship. The four zoneswith the largestmomentrelease
rates all have maximum depths less than 300 km (Southern Chile
not shown. Values of the depth are listed in Table 3. For other
details, seeFigure 4.)
Fig. 12. Maximum depthof seismicityversusage of subducting
lithosphere. The depth tendsto increasewith the age. (Valuesof
the two parametersare listed in Table 3.)
moment. Therefore large moment release rates are the
result of single large events, rather than an abundance of
moderate size events. These four zones also have the highest seismicslip rates and valuesof c•, indicating that there is
nitude of 8.0, but if the distributionof eventsversusmagni- a high degree of seismic coupling and very little aseismic
tude is different for the two zones, we will get different slip. Another characteristicthese regions have in common
moment release rates.
is the age of subducting lithosphere. Alaska, the eastern
There are many similaritiesbetweenthe four subduction Aleutians, and Kamchatka are the zones in which the younzoneswith the largestmomentreleaserates(southernChile, gest Pacific plate lithosphere is subducting, and southern
Alaska, eastern Aleutians, and Kamchatka). These zones Chile also has very young lithospheresubducting.
are the locationsof the four largestearthquakesof this cenThe age of subductinglithospherehas an important effect
tury. The moment releasecurvesfrom 1904 to 1980 for on the degree of seismiccoupling. As was shown in Figure
each zone are dominated by one large event, with smaller 4a, there is a consistent decreasein the moment release rate
earthquakescontributingonly a few percentof the total with increasing age among the subduction zones with slabs
belonging to the same plate. This is observed for each of
the Cocos, Philippine Sea, Indian, and Pacific plates and
implies that an older slab causesa reduction in the seismic
Down-Dip Length of Benioff Zone
•,
60
\
0 ..c,,,c
\
•
\
E
Length of Benioff Zone vs. Age
Cocos
•'x PhlllpMn©
<• \ \
[] i.d,.n
C) #oaca
I
I
I
1200
\
40
\
\
\
\
\
800
\
ß
20
E
/•
o
400
800
phlfipMne
1200
Length (km)
Fig. 11. Moment releaserate versusthe down-dip length of the
Wadati-Benioff zone. The zones belonging to the same plate have
been connectedby lines as shown. The moment releaserate strictly
decreaseswith increasinglength for the Philippine Sea, Cocos, and
Indian plates. The Pacific plate also has a similar trend. The four
zones with the largest moment releaserates all have Wadati-Benioff
zone lengths less than 550 km (Southern Chile not shown. Values
of the length are listed in Table 3. For other details, seeFigure 4.)
i
50
i
100
i
150
Age (my)
Fig. 13. Down-dip length of the Wadati-Benioff zone versusage
of the subductinglithosphere. If the zones of the Nazca plate (circles) are excluded, the length of the Wadati-Benioff zone tends to
increasewith the age. (Values of the two parameters are listed in
Table 3.)
PETERSON
AND SENO:SEISMICMOMENT RELEASEIN SUBDUCTIONZONES
coupling. An old slab is cooler, denser, and has a greater
gravitational force pulling it downward into the mantle
[Molnar and Atwater, 1978]. This may result in a reduced
coupling acrossthe contact zone betweenthe two plates.
The lack of correlation
between the moment release rate
and the age of subductinglithosphere for the Nazca plate
(Figure 4b) may be due to the widely varying agesand many
fracture zones offsetting the magnetic lineations, which
make it difficult to define a physically meaningful age for
these zones. It is also possible that the lack of correlation
indicates that the subduction processis being affected by
some other, as yet undetermined, parameter. However, we
do note that the southern Chile zone has the highest
moment release rate of all zones consideredin this study,
and it is the zone which is closestto a spreadingridge.
Within a singleplate, the age of subductinglithosphereis
the dominant factor affecting the moment release rates.
However, the question remains as to what causesthe trends
for the various platesto lie in different placeson the plot of
moment release rate versus age (Figure 4a). For example,
the Pacific plate zones have much higher moment release
rates than zones of other plates with similar ages of subducting lithosphere. It appearsthat each plate has a characteristic seismicmoment releasebudget and that some feature
of the plate as a whole determines the magnitude of the
moment release. Note that the Pacific plate, which has the
10,247
relationshipbetweenthesetwo parameters(Figure 9). This
is in contrastwith the result of Ruff and Kanamori [1980],
in which the degreeof coupling tends to increasewith convergencevelocity. There are several possiblereasons for
this discrepancy.The definitionof couplingin our studyis
the moment releaserate, while in Ruff and Kanamori's work
it is the maximumearthquakemagnitude. Examplesof how
these two parameters may differ were described above.
There is uncertaintyinvolvedin defininga maximum earthquake size and in calculatingthe moment releaserates, and
this may affect the correlation. Finally, in some casesthe
location of subduction zone boundaries differs between the
two papers.
The momentreleaserate dependsmostclearlyon the age
of subductinglithosphereand the upper plate velocity. If
we divide all the factors consideredinto dependent and
independentparameters,we find that the age and the upper
plate velocity are the two independent parameters. For
example,the length and maximum depth of seismicityboth
dependon the age and the convergence
rate [Molnar et al.,
1979]. The subductionrate dependson the age of subducting lithosphere[Carlsonet al., 1983]. The convergence
rate
is a combinationof the subductionrate and the upperplate
velocity. The age and the upper plate velocity are the basic
"given" parameters on which the others depend. We conclude that the observed weaker dependenceof the moment
largestmomentreleaserates,alsohasthe largestareaof the release rate on the subduction rate, length of slab, and
platesconsidered.An interesting
possibility
is that the large depth of seismicityresults from the strong dependenceof
area of the Pacific plate affectsthe mantle flow pattern the moment releaserate on the age.
beneaththe plate. The mantleflow modelsof Hager and
Acknowledgments We would like to thank Hitoshi Kawakatsu
O'Connell [1979] show that this large size inducesa relatively smoothflow patternbeneaththe Pacificplate from and Stuart Stephensfor helpful discussions,and Bob Geller for a
ridge to trench with flow parallelingthe plate motion. careful review of this paper. This work was supportedby NSF
grants PFRAL-06471 and EAR 81-08718.
Neighboringsmallerplates have their flow patternsdisturbed by the dominating Pacific flow, and the flow does
not parallel the plate velocity. The mantle convection
REFERENCES
modelsof Richterand McKenzie[1978]showthat the pres- Abe, K., Reliableestimationof the seismicmomentof largeearthsureperturbationsdue to the flow patternscontrolthe stress quakes,J. Phys. Earth, 23, 381-390, 1975.
state within the slab. The questionremains whether these Abe, K., Magnitudesof large shallowearthquakesfrom 1904 to
1980, Phys. Earth Planet. Inter., 27, 72-92, 1981.
flow patternscan affectinterplatestresses
in the thrust zone.
The coherentmantleflow approaching
the Pacificplatesub- Addicott, W. O., and P. W. Richards,Plate tectonicmap of the
ductionzonesin a directionnormalto the trenchmay exert
pressure on the slab and result in an increase in the normal
stressesacross the contact zone. This would in turn cause
circum-Pacific region, Am. Assoc. of Pet. Geol. Circum-Pacific
Map Project, Circum-PacificCouncil for Energy and Mineral
Resour., Tulsa, Okla., 1981.
Carlson,R. L., T. W. C. Hilde, and S. Uyeda, The drivingmechanism of plate tectonics: relation to age of the lithosphereat
an increasein the seismiccoupling. If the mantle flow is
trenches,Geophys.Res. Lett., 10, 297-300, 1983.
not parallel to the plate velocityfor the smallerplates, as
Engdahl, E. R., Seismicityand plate subductionin the centralAleusuggestedby Hager and O'Connell's models,then this effect
tians, in Island Arcs, Deep Sea Trenches,and Back-Arc Basins,
would not occur, and the couplingwould be reduced. We
wish to emphasizethat this is a speculative
argumentand
has not been tested. More fluid mechanical models are
neededto study the effectsof mantle flow patternson
interplatestresses
and couplingin subductionzones.
The observation that the moment release rates are small
Maurice Ewing $er., vol. 1, edited by M. Talwani and W. C.
Pitman III, pp. 259-271, AGU, Washington,D.C., 1977.
Geller, R. J., Scalingrelationsfor earthquakesourceparameters
and magnitudes,
Bull. Seismol.$oc. Am., 66, 1501-1523, 1976.
Geller, R. J., and H. Kanamori,Magnitudesof greatshallowearthquakes from 1904 to 1952, Bull. Seismol. Soc. Am., 67,
587-598,
1977.
for zoneswith retreatingupperplatesmay be explainedby Gutenburg,B., and C. F. Richter,Magnitudeand energyof earthquakes,Ann. Geofis.Rome, 9, 1-15, 1956.
the "anchoredslab" modelof UyedaandKanamori[1979].
In this model, the slab cannotmove easilyin a horizontal
directionnormalto the trenchdue to its largearea being
Hager, B., and R. O'Connell, Kinematicmodelsof large-scale
flow
in the earth'smantle,J. Geophys.Res., 84, 1031-1048, 1979.
Hasegawa,A., N. Umino, and A. Takagi, Double-planedstructure
of the deep seismic zone in the northeasternJapan arc,
anchoredin the mantle. A retreatingupperplate will tend
to separatefrom the slaband causea reductionin coupling Tectonophysics,47, 43-58, 1978.
House,L. S., and K. H. Jacob,Earthquakes,plate subduction,and
and will sometimes
resultin back-arcspreading.
We had expected that the moment release rate would
stressreversalsin the easternAleutianArc, J. Geophys.Res., 88,
9347-9373,
1983.
increasewith convergence
velocity, but we found no clear Isacks,B. L., and M. Barazangi,Geometryof Benloftzones:lateral
10,248
PETERSON
ANDSENO:
SEISMIC
MOMENT
RELEASE
IN SUBDUCTION
ZONES
segmentationand downwards bending of the subducted lithosphere, in Island Arcs, Deep Sea Trenches,and Back-Arc Basins,
Maurice Ewing Ser., vol. 1, edited by M. Talwani and W. C.
Pitman III, pp. 99-114, AGU, Washington, D.C., 1977.
Isacks, B., and P. Molnar, Distribution of stressesin the descending lithospherefrom a global survey of focal mechanismsolutions of mantle earthquakes, Rev. Geophys. Space Phys., 9,
103-174, 1971.
Kanamori, H., Seismic and aseismic slip along subduction zones
and their tectonic implications, in Island Arcs, Deep Sea
Trenches, and Back-Arc Basins, Maurice Ewing Ser., vol. 1,
edited by M. Talwani and W.C. Pitman III, pp. 163-174, AGU,
Washington, D.C., 1977a.
Kanamori, H., The energy release in great earthquakes, J.
Geophys. Res., 82, 2981-2987, 1977b.
Kanamori, H., and D. L. Anderson, Theoretical basis of some
empirical relations in seismology, Bull. Seismol. Soc. Am., 65,
1073-1095, 1975.
Kawakatsu, H., and T. Seno, Triple seismiczone and the regional
variation of seismicity along the northern Honshu Arc, J.
Geophys. Res., 88, 4215-4230, 1983.
Kelleher, J., and W. McCann, Buoyant zones, great earthquakes,
and unstable boundaries of subduction, J. Geophys. Res., 81,
4885-4896, 1976.
Kelleher, J., J. Savino, H. Rowlett, and W. McCann, Why and
where great thrust earthquakes occur along island arcs, J.
Geophys. Res., 79, 4889-4899, 1974.
Lay, T., H. Kanamori, and L. Ruff, The asperity model and the
nature of large subduction zone earthquakes,Earthquake Predict. Res., 1, 3 - 71, 1982.
McCann, W. R., S. P. Nishenko, L. R. Sykes, and J. Krause, Seismic gapsand plate tectonics:seismicpotential for major boundaries, Pure Appl. Geophys., 117, 1082-1147, 1979.
Minster, J. B., and T. H. Jordan, Present-day plate motions, J.
Geophys. Res., 83, 5331-5354, 1978.
Mogi, K., Seismicityin western Japan and long-term earthquake
forecasting, in Earthquake Prediction: An International Review,
Maurice Ewing Ser., vol. 4, edited by D. W. Simpsonand P. G.
Richards, pp. 43-51, AGU, Washington, D.C., 1981.
Molnar, P., and T. Atwater, Interarc spreadingand Cordilleran
tectonics as alternates related to the age of subductedoceanic
lithosphere,Earth Planet. Sci. Lett., 41, 330-340, 1978.
Molnar, P., D. Freedman, and J. Shih, Lengths of intermediate
and deep seismiczones and temperaturesin downgoing slabs of
lithosphere,Geophys. J. R. Astron. Soc., 56, 41-54, 1979.
Newcomb, K. R., and W. R. McCann, Seismicevidencefor varying
modesof subductionalong the Indonesianarc, Eos Trans. A GU,
63, 381, 1982.
Richter, F., and D. McKenzie, Simple models of mantle convection,
J. Geophys., 44, 441-471, 1978.
Ruff, L., and H. Kanamori, Seismicityand the subductionprocess
Phys. Earth Planet. Inter., 23, 240-252, 1980.
Ruff, L., and H. Kanamori, Seismiccoupling and uncouplingat
subductionzones, Tectonophysics,99, 99-117, 1983.
Seno, T., The instantaneousrotation vector of the Philippine Sea
plate relative to the Eurasian plate, Tectonophysics, 42,
209-226, 1977.
Seno, T., and T. Eguchi, Seismotectonicsof the western Pacific
region, in Geodynamics of the Western Pacific-Indonesian
Region, Geodyn. Ser., vol. 11, edited by T. Hilde and S. Uyeda,
pp. 5-40, AGU, Washington, D.C., 1983.
Seno, T., and B. Pongsawat,A triple-planedstructureof seismicity
and earthquake mechanismsat the subductionzone off Miyagi
Prefecture, northern Honshu, Japan, Earth Planet. Sci. Lett.,
55, 25- 36, 1981.
Shimazaki, K., and T. Nakata, Time-predictable recurrencemodel
for large earthquakes,Geophys.Res. Lett., 7, 279-282, 1980.
Sykes, L. R., and R. C. Quittmeyer, Repeat times of great earthquakes along simple plate boundaries, in Earthquake Prediction:
An International Review, Maurice Ewing Ser., vol. 4, edited by
D. W. Simpsonand P. G. Richards, pp. 217-247, AGU, Washington, D.C., 1981.
Thatcher, W., and T. C. Hanks, Source parameters of southern
California earthquakes,J. Geophys.Res., 78, 8547-8576, 1973.
Utsu, T., Seismicactivity in Hokkaido and its vicinity, Geophys.
Bull. Hokkaido Univ., 20, 51-75, 1978.
Uyeda, S., and H. Kanamori, Back-arc opening and the mode of
subduction,J. Geophys.Res., 84, 1049-1061, 1979.
Veith, K. F., The relationshipof island arc seismicityto plate tectonics, Ph.D. thesis, Southern Methodist Univ., Dallas, Tex.,
1974.
Vlaar, N., and M. Wortel, Lithosphericaging, instability, and subduction, Tectonophysics,32, 331-351, 1976.
Wang, S., Tectonic implicationsof global seismicitystudies,Ph.D.
thesis, Stanford Univ., Stanford, Calif., 1981.
Yokokura, T., On subduction dip angles, Tectonophysics,77,
63-77, 1981.
Yoshii, T., A detailed cross-section of the deep seismic zone
beneath northeastern Honshu, Japan, Tectonophysics, 55,
349- 360, 1979.
E. T. Peterson,Departmentof Geophysics,Stanford University,
Stanford, CA 94305.
T. Seno, International Institute of Seisinologyand Earthquake
Engineering,BuildingResearchInstitute, Ministry of Construction,
Tsukuba, Ibaraki Prefecture, Japan 305.
(Received October 7, 1982;
revisedApril 4, 1984;
acceptedApril 10, 1984.)