GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 17, PAGES 3389-3392, SEPTEMBER 1,2001
Sediment effect on tsunami generation of the 1896 Sanriku
tsunami earthquake
Yuichiro
Tanioka
Seismology
andVolcanologyResearch
Department,
Meteorological
ResearchInstitute,Tsukuba305-0052,Japan
Tetsuzo Seno
Earthquake
Research
Institute,Universityof Tokyo,Tokyo113-0032,Japan
Abstract. The 1896 Sanrikuearthquakewas one of the most
devastating tsunami earthquakes, which generated an
anomalouslylarger tsunamithan expectedfrom its seismic
waves.Previousstudiesindicatethat the earthquakeoccurred
beneaththe accretionary
wedgenearthe trenchaxis. It was
horizontal
movement
on thedecollement
mighthavecauseda
large additional uplift. The horizontal trench-ward
displacement
of the backstop
leadsto overallfoldingof the
accretionary
prism.In this paper,we try to answera key
question:whetheror not suchan additionaluplif• near the
We computethe tsunamifrom
pointedout recentlythat sediments
neara toe of an inner trenchcausesa largetsunami?
trenchslope with a large horizontalmovementdue to the the 1896 Sanriku event for ocean bottom deformation
earthquakemight have causedan additionaluplift. In this includingelastic deformationdue to faulting and the
paper,the effectof the additionaluplift to tsunamigeneration additionalsedimentuplift. Then, the effect of the additional
is discussed.
of the 1896 Sanfiku tsunamiearthquakeis quantified.We uplif•onthetsunamigeneration
estimatethe slip of the earthquake
by numericallycomputing
tsunamisandcomparingtheirwaveformswith thoserecorded
at threefide gauges.The estimatedslip for the modelwithout 2. Three Models for Additional Uplift
the additionaluplift is 10.4 m, and thosewith the additional
The mechanismof the additionaluplift is represented
by
uplit• are 5.9-6.7 m. This indicatesthatthe additionaluplift of the horizontal movement of the backstopscrapingthe
thesediments
nearthetrenchhasa largeeffectonthe tsunami sedimentsin front of it [Seno,2000] as shownin Figure 1.
generation.
1. Introduction
Thismechanism
wasoriginallysuggested
for the formationof
theaccretionary
prism[Daviset al., 1983;Byrneet al., 1988],
althoughthe time scaleis muchlongerthan considered
here
for tsunamigeneration.We use three simplifiedmodels,
Models A, B, and C, shown in Figure 2. In Model A,
Most large, shallow earthquakesin subductionzonesare
horizontal
movementof the backstopupliftssediments
only
tsunamigenic.Among them, unusual earthquakesthat
abovetheslopeof thebackstop.In thismodel,theuplif• of the
generatemuch larger tsunamisthan expectedfrom their
by us=untan19
where un is the
seismicwavesare called "tsunamiearthquakes"[Kanamori, sediments,u•, is represented
and t9is the dip
1972]. The 1896 Sanrikutsunamiearthquakethat occurred horizontalmovementdue to the earthquake,
along the Japan trench was one of the most anomalous angleof theslope.Taniokaand Satake[ 1996a]usedthe same
of a slope.
earthquakes;
the groundshakingwas relativelyweak,but the modelto uplift waterby thehorizontalmovement
In
Model
B,
horizontal
movement
of
the
slope
causes
the
followingtsunamisweredevastating.
Abe [1994] assigned
the
pile. In thismodel,with
surfacewave magnitudeas Ms=7.2.The tsunamimagnitude uniformuplif• of thewholesediments
us,is represented
was determinedas 8.6 from globaldata [Abe, 1979] and 8.2 themassbalance,theuplif•of thesediments,
by
us
=uh/-//W
where
H
is
the
height
of
the
backstop
slope,W is
from local data [Abe, 1981]. Recently,Taniokaand Satake
the
width
of
the
sediments,
and
un
is
the
same
as
in
Model A.
[1996b] determinedthe fault parametersusing the tsunami
waveformsand estimatedthe momentmagnitudeas 8.0. The Model A produceslarger uplif• in a narrowerarea, while
largediscrepancy
betweenMs and Mt categorizesthis event Model B producesa smalleruplif• in a wider area. These
modelsare the two end members.In Model C, we assumethat
asa tsunamiearthquake.
the sediment block behaves like rubber with an effective
Satake and Tanioka [1999] showedthat most moment
Poisson's
ratioof 0.49. The surfacedeformation
is computed
releaseof tsunamiearthquakes,.
includingthe 1896 Sanriku
event,occursin a narrowregionnearthetrench.In thisregion,
a large amount of unconsolidatedor semiconsolidated
sediments
exist.The low anglethrustfaultingnearthe trench
on the decollementcauses large horizontal movement
perpendicularto the trenchaxis. Seno [2000] inferredfrom
abnormaluplif• associatedwith the horizontalmovementof
the 1999 Chi-Chi earthquakein Taiwan that sedimentswith
fikrn
•
Copyright2001 by the AmericanGeophysical
Union.
Papernumber2001GL013149.
0094-8276/01/2001GL013149505.00
Figure1. A schematic
cross-section
of thelowertrenchslope
andtrenchwedgeseenin theJapanTrench.
3389
3390
TANIOKA
AND SENO:
SEDIMENT
EFFECT ON TSUNAMI
uplift
1)Model
AusII--"- usadditonal
ß
I
Us--'-Uhtan0
Uh:horizontal
deformation
dueto faulting
2) Model B
us;•!
[usadditonal
uplift
GENERATION
tsunami waveforms. The location of the fault is shown in
Figure3. We usethe sameparameters
exceptthe dip angle.
Umino et al. [1995] relocated micro-earthquakes
off the
Tohokuregionusingthe sP depthphaseandindicatedthatthe
dip angleof theplate interfacenearthe sourceregionis less
than 10ø. Therefore,we usea dip angleof 10ø insteadof 20ø.
The elastic ocean bottom
deformation
due to the
1896
earthquake
is computedfrom thosefaultparameters
usingthe
equations of Okada [1985]. The sum of this elastic
deformationand the additionaluplift causedby the above
models(A, B, or C) is usedas the oceanbottomdeformation
dueto theearthquake.
4. Tsunami Computation
3)Model
C
In order to compute tsunami propagation,initial water
surfacedeformation
mustbe estimated.In general,the water
surfacedeformation
due to faultingof a largeearthquake
is
•
•
i
[
assumedto be the same as the ocean bottom deformation,
us
computed
additonal
uplift
becausethe wavelengthof the oceanbottomdeformationis
fixed
'••
.......
"'"--'-'
"•.
'"•'
ß.- -• -'•
"• Poisson's
ratio---0.49
Figure 2. Three models for additional uplifts causedby
sediments
with horizontalmovementof thebackstop.
muchlargerthanthe oceandepth.However,thisassumption
cannotbe applied in this study,becausethe width of the
additionaldeformation(Figure 2) is about 7 kin, which is
similarto the oceandepthnear the sourceregion.Kajiura
[1963]showedthatthewatersurfacedeformation,
rl(x,y),due
to theoceanbottomdeformation,
HB(xoYo),
is expressed
by the
followingequation:
r/(x,y)
=$lHB(xo,Yo)Rdxodyo
(1)
usingthestructural
analysis
software,
MSC/NASTRAN,
in
which a finite element method is used. In this model, the
bottom of the sediment block is fixed and the horizontal
movementof the slope,un,dueto the faultingis appliedalong
theslopeof thebackstop.
In and aroundthe northernpart of the sourceregionof the
1896 Sanriku earthquake,a detailed survey combining
Seabeam mapping, single-channel seismic reflection
observations,
andgravityandgeomagnetic
measurements
was
conductedduring Leg 3 of the French-Japanese
KAIKO
project[Cadetet al., 1987]. The resultof the mappingshows
that thereis a linear escarpment
about7 km landwardof the
trench axis and parallel to the axis. Cadet et al. [1987]
indicatesthat,in front of the escarpment,
thereis a thick pile
of sub-horizontalshortreflectionswhich representreworked
slump debris. From the reprocessedmultichannelseismic
record shown by Cadet at al. [1987], the height of the
backstop(/-/) canbe as largeas 4 km and the dip angleof the
scarp is about 50ø (Figure 1). The sea beam survey by
Hakuho-marucruiseKH90-1 [Kobayashiet al., 1991] shows
that the escarpment
continuesthroughoutthe sourceregionof
the 1896 Sanrikuearthquakeand the averagedistanceof the
escarpment
from the trenchaxis is about7 km. Therefore,we
assumethat the width of the slumpsedimentsis 7 km, the
heightof the backstopslopeis 4 km, and the dip angleof the
slopeis 50ø throughoutthe sourceregionof the 1896 Sanriku
earthquake.
3. Ocean Bottom
Deformation
Sanriku Earthquake
due to the 1896
$
where
R=(1/•:)•
(-1)n(2n
+1)•2n
+1)
2+(X-Xo)
2+(yy0)2•3/2(
n-O
4• ø
41 o
'• 1896Sanriku
Tsunami
EQ•
-A•yuka
•w,•."
•(•I • - ,
35 ø
140 ø
142 ø
144 ø
146 ø
Figure 3. The tsunami computationarea and the fault
Taniokaand Satake[ 1996b]estimatedthe faultparameters parameters of the 1896 Sanriku tsunami earthquake.
(length,210 km; width, 50 km; strike, 190ø;dip angle,20ø; Rectangleshows the location of the fault. Stars show the
rake, 90ø) of the 1896 Sanrikutsunamiearthquakeby using locationof tidegauges.
TANIOKA
AND SENO:
SEDIMENT
EFFECT ON TSUNAMI
GENERATION
3391
1896 Sanriku
observed
synthetic
withoutadditionaluplifts
slip
=10.4
rn
Model A
slip
•.6.7
m,.
-, ,'
Model B
Model C
slip = 6.6 rn
slip = 5.9 rn
Hanas
'aki..•.••.,.,,f•j
wa•'
';
Ayuka
",,
Choshi
....
50cmF
20 '' 40
,,'
,•,.
....
•
F
.... 60
80 20 ' 40
time, min
__. •"w.•.. _
r'
r-
'' 20
80
0 ' 6'0'''80 20 '
time, min
time, min
i
i
40
i
i
60
80
time, min
Figure 4. Comparison
of observedand computedtsunamiwaveformsfrom four models,the verticaldeformationwithoutadditional
uplifts(left), andthatwith theadditionalupliftsof ModelA, B, andC.
In the above equations,physical variablesare written in
non-dimensional
form as follows:x=x'/d, y=y'/d, and r/=r/'/d
where a prime indicatesa dimensionalvariableand d is the
water depth. We numericallycompute the water surface
estimatedslip due to the 1896 Sanrikuearthquakebecomes
5.9-6.7 m by addingan additionaluplift; significantlyless
than 10.4 m from the purely elasticmodel.The uplift due to
the sedimentsnear the trenchhas thereforea large effect on
deformation,
•l(x,y),usingtheaboveequations
anduseasthe __thegeneration
ofthetsunami.
initialconditionof thetsunamicomputation.
The tsunami propagationis computedby numerically
solving the linear Boussinesqequation in a spherical
coordinatesystem.The methodof the numericalcomputation
1) Elasticdeformationdue to faulting
is the sameasthatusedby Tanioka[2000].The computational
area is shownin Figure3. The grid spacingis 20 secof arc
(about600 m).
surfac•
ocean •--- --•
ocean
bottom
•
5. TsunamiWaveformsand EstimatedSlips
The observedtsunamiwaveformsat three tide gauges,
Hanasaki,Ayukawa,and Choshi,are used for this analysis
(Figure4). However,considering
thepooraccuracyof thetide
gaugeclocksin 1896 [Taniokaand$atake,1996b],we usethe
waveformswithoutabsolutetiming.We computethe tsunami
waveformsat the threetide gaugesusingthreeoceanbottom
deformationmodels;one is the elasticdeformationonly, the
other three are the elastic deformation
with the additional
uplift causedby Models A, B, and C as detailed above.
Comparisonsof the observed and computed tsunami
waveformsare shownin Figure 4. The first arrivalsfor all
stationsareadjustedto match.Slip on the fault for eachmodel
is estimated
by comparingtheamplitudesof the firstpulseof
theobserved
andcomputedtsunamiwaveforms.
The estimatedslips are 10.4 m for the model using the
elasticdeformationonly, and 6.7 m, 6.6 m, and 5.9 m for the
elastic deformationswith the additionaluplift causedby
ModelsA, B and C, respectively.
The estimatedslip for the
modelusingthe elasticdeformationonly is largerthan that
estimated
by Taniokaand Satake[ 1996b].The mainreasonis
that the dip angle, 10ø, for the fault model in this studyis
smallerthanthat, 20ø, usedby Taniokaand $atake [ 1996b].
The estimatedslips from the two additionaluplift models,
ModelsA and B, are almostthe same,but thoseare slightly
largerthanthat from Model C. Theseresultsindicatethat the
I
0
20krn 40km
I
•
I
I
60krn 80krn 100km
2) Additionaluplift
Model A
Model B
Model C
ocean bottom
Figure 5. Comparisonof oceanbottom and oceansurface
deformations
for (1) the elasticdeformationdue to faulting,
and (2) the additionaluplift causedby Model A, B, and C.
Cross-sections
of initial oceansurfacedeformations(top),
oceanbottomdeformations
(middle),and the locationof the
faultor thesediment
block(bottom)areshownfor eachmodel.
The same scales are used for all cases.
3392
TANIOKAAND SENO: SEDIMENTEFFECTON TSUNAMIGENERATION
Figure 5 showsthe cross-sectionof the oceanbottom and
the ocean surface deformations for the elastic deformation and
threeadditionaluplifts causedby Model A, B, and C. The
cross-section
wastakenalonga line perpendicular
to the fault
Acknowledgments.
We thankDrs. K. KanjoandA. Yoshidafor
helpful
discussions.
Thismanuscript
wasg•eatly
improved
by the
comments
from two reviewers,Dr. E. L. GeistandDr. K. Satake.We
alsothankDr. J. M. Johnson
for helpfulcomments
to polishthe
manuscript
andDr. T. Yamamoto
for technical
supports
to usethe
strikeat a positionof themiddleof the fault.The slipon the MSC/NASTRAN
software.
fault is 1 m for all casesin Figure 5. The oceansurface
deformationis computedfrom the oceanbottomdeformation
References
by applyingequation(1). The oceansurfacedeformationsfor
Models A, B, and C becomesimilar althoughthe ocean Abe, K., Size of greatearthquakes
of 1873-1974inferredfrom
tsunami
data,J. Geophys.
Res.84, 1561-1568,1979.
bottomdeformations
fromthethreemodelsarequitedifferent.
earthquakes
of the
Therefore,the tsunamiwaveformscomputedfromthe elastic Abe, K, Physicalsize of tsunamigenic
northwestern
Pacific.Phys.EarthPlanet.Inter. 27, 194-205,
deformation
withtheadditional
upliftscaused
by ModelsA, B,
1981.
andC becomesimilarasshownin Figure4.
Abe, K., Instrumental
magnitudes
of historicalearthquakes,
In general,the observedtsunamiwaveformsare equally
1892-1898,
Bull.Seisin.Soc.Am.84,415-425,1994.
size
explainedby the computedtsunamiwaveformsfrom the three Byrne,D. E., D. M. Davis,andL. R. Sykes,Lociandmaximum
of thrust
earthquakes
andthemechanics
of theshallow
region
of
models. However, in detail, a small differenceexists in the
subduction
zones,Tectonics,
7, 833-857,1988.
initial downpulseat Ayukawa.The computeddownpulses
Cardet,
J.-P.,K. Kobayashi,
J. Aubouin,
J. Boulegue,
C. Deplus,
J.
fromthe threemodelswith the additionalupliftsare smaller
Dubois,R. vonHuene,L. Jolivet,T. Kanazawa,
J. Kasahara,
K.
and closer to the observed waveform than that from the model
Koizumi,
S.Lallemand,
Y. Nakamura,
G. Pautot,
K. Suyehiro,
S.
Tani,H. Tokuyama,
andT. Yamazaki,
TheJapantrenchandits
withouttheadditionaluplift.The additional
upliftsaddednear
juncture
withtheKuriltrench:
cruise
results
of theKaikoproject,
the trench did not affect the subsidence located toward the
Leg3, EarthandPlanet.Science
Lett.83,267-284,1987.
coastand away from the trench(Figure5). The initial down Davis,D., J. Suppe,andF. A. Dahlen,Mechanicsof fold-and-thrust
pulse was generatedby this subsidence.Hence, the smaller
beltsandaccretionary
wedges,
J. Geophys.
Res.88, 1153-1172,
1983.
slip estimatedfor the threemodelswith additionaluplifts
earthquakes
andsubduction
processes
near
generatea smallerinitial downpulsethan the modelwithout Fukao,Y., Tsunami
deep-sea
trenches,
J.
Geophys.
Res.,
84,
2303-2314,
1979.
additional
uplift.Thismayindicatethattheadditional
uplifts Kajiura,K., The leadingwaveof a tsunami,
Bull.Earth.Res.Inst.
betterexplainthetsunamiwaveformat Ayukawa.
Univ.Tokyo 41,535-571, 1963.
Kanamori,
H., Mechanism
of tsunami
earthquake,
Phys.
EarthPlanet.
6. Discussion and Conclusion
Inter. 6, 246-259, 1972.
Kobayashi,
K., K. Tamaki,
H. Fujimoto,
M. Nakanishi,
J.Oshida,
D.
C. P. Masalu,
Y. Ogawa,
H. Iino,T. Tsukioka,
K. Kashihara,
T.
The additionaluplift of the sediments
nearthe trenchhasa
Matsumoto,
T. TanakaandT. Funam.,
Results
of bathymetric
largeeffectonthetsunamigeneration.
To explainthetsunami
measurement,
Prelim.Rep.HakuhoMaru CruiseKH90-1,10-11,
1991.
waveformsof the 1896 Sanrikutsunamiearthquake
recorded
due to shearandtensilefaultsin a
at thethreetidegauges,thesliponthefaultwasreducedfrom Okada,Y.. Surfacedeformation
half-space,
Bull.Seism.
Soc.Am.75,1135-1154,
1985.
10.4m to 5.9-6.7m by includingthe additionaluplift of the
E. A., Seismicparameters
controlling
far-fieldtsunami
sediments.
The observed
waveforms
are alsowell explained Okal,
amplitudes:
a review,NaturalHazards,1, 67-96,1988.
by thecomputedwaveforms.The seismicmomentof the 1896 Pelayo,A.M., and D.A. Wiens,Tsunamiearthquakes:
slow
Sanriku
tsunami
earthquake
is computed
to be 12-14x 1020
thrust-faulting
events
intheaccretionary
wedge,
J. Geophys.
Res.,
Nm (Mw=8.0-8.1)by assuming
thattherigidityaroundsucha
shallowfaultis 2 x 10•øN/m2.
97, 15321-15337, 1992.
associatedwith some events,such as the 1896 Sanriku and
Geophys.
Res.Lett.23, 1549-1552,1996b.
Satake,
K. andY. Tanioka,
Source
of tsunami
andtsunamigenic
earthquakes
in
subduction
zones,
Pure
and
Appl.
Geophys.
154,
Themechanism
of tsunamiearthquakes
wasfirstproposed 467-483, 1999.
by Kanamori[ 1972]whoindicated
thatthelargediscrepancySeno,T., The21 September,
1999Chichiearthquake
in Taiwan:
betweenseismicandtsunamiwavesof tsunamiearthquakes
is
implications
fortsunami
earthquakes,
Terr.Atmos.
Ocean
Sci.11,
701-708, 2000.
caused
by slowandlongruptureprocesses.
Palayoand Weins
T. andK. Satake,Tsunami
generation
by horizontal
[1992],usingseismicwaveanalysis,similarlyconcluded
that Tanioka,
displacement
of ocean
bottom,
Geophys.
Res.Lett.23, 861-864,
tsunami
earthquakes
areslowthrusteventsin theaccretionary 1996a.
wedge.However,it hasbeennoticedthatslowslipsaloneare Tanioka,T. andK. Satake,Faultparameters
of the 1896Sanriku
not enough to explain anomalousexcitation of tsunamis
tsunami
earthquake
estimated
fromtsunami
numerical
modeling,
T., Numerical
simulation
of far-fieldtsunami
usingthe
1946Aleutianearthquakes.
Fukao [1979] and Okal [1988] Tanioka,
linearBoussinesq
equation,
Pap.Meteorol.
Geophys.,
51, 17-25,
then proposedthat an earthquakesourcewithin a shallow
2000.
sedimentary
layer,mightexcitemuchlargertsunamis
thanin Umino,
N.,A. Hasegawa,
andT. Matsuzawa,
sPdepth
phase
atsmall
solidrock,and be responsible
for suchanomalous
tsunamis. epicentral
distances
andestimated
subducting
plateboundary,
J. Int. 120,356-366,1995.
SatakeandTanioka[1999],usingtsunami
waveform
analysis, Geophys.
concluded
thatmostmomentreleaseof tsunamiearthquakes T. Seno,Earthquake
Research
Institute,
University
of Tokyo,
occursin a narrowregionnearthetrenchandtheconcentratedTokyo113-0032,
Japan.
(e-mail:
seno•eri.u-tokyo.ac.jp)
slip is responsible
for the largetsunamis.
In thispaper,we
Y. Tanioka,
Seismology
andVolcanology
Research
Department,
Research
Institute,
1-1Nagamine,
Tsukuba
305-0052,
indicatethattsunamiearthquakes
are causednot onlyby a Meteorological
(e-mail:ytanioka•mri-jma.go.jp)
slowrupture
process
ora concentrated
slipin theaccretionaryJapan.
wedge,but alsoby an additionaluplift of thesediments
near
the trenchdue to a largecoseismichorizontalmovementof
thebackstop.
The horizontal
trench-ward
displacement
of the (ReceivedMarch9, 2001;revisedJune1,2001;
backstop
leadsto overallfoldingof theaccretionary
prism.
acceptedJune28, 2001.)