Content uploaded by Haruo Sato
Author content
All content in this area was uploaded by Haruo Sato on Jan 10, 2015
Content may be subject to copyright.
TOhoku Geophys. Journ. (Sci. Rep. TOhoku Univ., Ser. 5), Vol. 34, No. 4, pp. 121-134, 1996
Source Process of the 1994 Far East Off Sanriku
Earthquake, Japan, as Inferred from a
Broad-Band Seismogram
TAKESHI NISHIMURA, HISASHI NAKAHARA,
and MASAKAZU OHTAKE
HARUO SATO
Department of Geophysics, Graduate School of Science,
Tohoku University, Sendai 980-77
(Received November 7, 1995)
Abstract : The 1994 far east off Sanriku earthquake (Mw= 7.7 ; the 1994 Sanriku-Oki
earthquake) took place near the Japan trench on December 28, 1994. The earthquake was
a reverse fault type, and the rupture proceeded westward to the Sanriku coast along the
boundary between the Pacific plate and the landward plate, overlapping the southern part
of the source region of the 1968 Tokachi-Oki earthquake (Ms=7.9). The Sanriku-Oki
earthquake was well recorded without saturation by a broad-band seismometer (STS-2) at
a station 300 km far from the epicenter. Fitting synthetic seismograms to the low-fre-
quency component of the observed three-component waveform, we estimate slip distribu-
tion of the rupture on the main fault with an area of approximately 170 km x 84 km. The
fault mainly slipped in the region from the middle to the western (landward) part of the
fault plane, having the maximum slip of 1.2 m, whereas a small slip of less than 0.4 m slip
is observed in the region close to the initial break point. The total seismic moment
amounts to 3.2 x 1020 Nm. The spatial distribution of fault slip indicates that the 1994
Sanriku-Oki earthquake ruptured the region of incomplete slip at the time of the 1968
Tokachi-Oki earthquake. From the analysis of the high-frequency component of the
observed seismogram, we find that the high-frequency energy is also strongly radiated from
the middle to the western part, with the highest intensity at the western edge of the fault
plane. Most of the aftershock are observed outside the region of large slip at the time of
the main shock.
1. Introduction
At 12h19m of 28 December 1994 (UT), a large earthquake of Mw= 7.7 occurred far
east off the Sanriku coast, northeast Japan. The Harvard CMT solution shows a focal
mechanism of reverse fault type with a strike of 184°, a dip of 15° and a rake of 70° with
a centroid at 143.12°E, 40.41°N, 34 km in the depth, suggesting that the earthquake
ruptured along the boundary between the Pacific plate and the landward one. The
epicenter of the main shock is located by the Japan Meteorological Agency at 143.43°E,
40.27N, close to the Japan Trench, and most part of the aftershocks distribute in a
region of about 170 km x 84 km to the west of the initial break of the main event as
shown in Figure 1. Umino et al. (1995) re-determined the focal depths for selected
aftershocks accurately by using arrival times of depth phases (sP), and revealed that
these events distribute along the boundary of the two plates. Matsuzawa et al. (1995)
determined the focal mechanisms of the aftershocks from the initial motion of P-waves,
122 Source Process of the 1994 Sanriku-Oki earthquake
and observed larger dip angles (45°) for the aftershocks occurring close to the Sanriku
coast. These results indicate that the rupture of the main shock proceeded westward
from the trench axis along the plate boundary.
Seismogram of the main shock was completely recorded with a broad-band seis-
mometer (STS-2) at the Tuyama station of the Department of Geophysics, Tohoku
University, which is located about 300 km to the southwest of the epicenter (Figure I).
The seismic signals are digitized with a resolution of 24 bits and a sampling frequency of
80 Hz by using the Q680 system of QUAN TERRA Inc. Figures 2 and 3 show the
displacement record for three components, and filtered velocity seismograms for vertical
component, respectively. Broad-band seismograms at other stations in the northeastern
part of Japan were all saturated at the maximum amplitude (Matsuzawa, personal
communication), and Tuyama is the closest station that recorded full broad-band
waveform of the main shock.
The 1994 Sanriku-Oki earthquake took place within the area enclosed by the rupture
zone of the 1968 Tokachi-Oki earthquake (Ms =7.9), whose source process was studied
by Mori and Shimazaki (1985) and others. Hence, quantitative study on the source
process of the 1994 Sanriku-Oki earthquake is important not only to understand the
rupture process of this event but also to clarify its relation to the Tokachi-Oki earth-
quake of 26 years before. In the present study, we reveal the spatial distributions of
fault slip and intensity of high-frequency wave radiation based on the broad-band
seismogram observed at the Tuyama station. We further discuss the result in compari-
son with the aftershock distribution and the focal process of the 1968 Tokachi-Oki
earthquake.
2. Slip Distribution
2.1. Fault Model and Inversion Method
We estimate the rupture velocity, rise time and spatial distribution of moment
release for the main shock by fitting the synthetic displacement waveform with the
observed one. The rectangle of 170 km x 84 km in Fig. 1, which includes most part of the
aftershocks, presents the horizontal projection of the rupture plane we assumed. The
plane dips to the west following the distribution of well-relocated aftershock
hypocenters (Umino et al., 1995). We divide the rupture area into eight subfaults of
roughly 40 km in the linear dimension so that each subfault is approximated by a point
source at frequencies lower than 0.05 Hz. For all the subfaults, we fix the angles of
strike and rake at 184° and 70°, respectively, following the CMT solution. The dip angle
is changed in accordance with the location of subfaults referring to the spatial distribu-
tion of relocated aftershocks (Umino et al., 1995) and the focal mechanisms of after-
shocks (Matsuzawa et al., 1995). The dip angles and depths of subfault at the center for
the eight subfaults are summarized in Table I.
We postulate that the rupture proceeds unilatelaly from the east to the west after the
initial break at the subfault 8. We further assume a constant rupture velocity VT, along
T. NISHIMURA, H. NAKAHARA, H. SATO AND M. OH TAKE 123
42°
41°
40°
39°
38°
140° 141° 142° 143° 144° 145°
CMT
tc;
0
(.) •
fI
2' • •
• -;•!
. • "
1
•
6/
•..
3
^••
8..
. . . .
Main
0
TUYAMA. 1^11k)
ct;
crsq
0
20
—
-c 40
60
80
. • .
•
• . ' •
„. •
•
,••
r
•
, .
•
• •
.
. .•
Fig. 1. Hypocenters of the 1994 Sanriku-Oki earthquake (diamond) and aftershoks for
the period of Dec. 28, 1994-Jan. 10, 1995 (small dots), the fault area, the CMT solution
by Harverd University (solid square is the centroid). Solid circles in the lower panel
are re-determined hypocenters of selected aftershocks by Umino et al. (1995).
Tuyama station is shown by open square. Rupture area of the main shock (approxi-
mately 170 km x 84 km) is divided into eight subfaults for the analysis. Hypocenter
locations are provided from the Observation Center for Prediction of Earthquakes and
Volcanic Eruptions (OCPEV), Tohoku University.
124 Source Process of the 1994 Sanriku-Oki earthquake
Fig. 2. Displacement seismogram of three components recorded by the STS-2 seismome-
ter at the Tuyama station. Arrow indicates the data span used for the analysis of slip
distribution.
the fault plane. The starting time of rupture for the k-th subfault, a, is calculated from
the rupture velocity and the distance between the k-th subfault and subfault 8. A ramp
function having a rise time r is assumed as the source time function for all of the
subfaults. The i-th component of the synthetic displacement seismogram, C,(t), can be
expressed as
8 ( Ci(; vr,S(6 ; vr, r),1)
where m, represents the amount of moment release from the k-th subfault, i the
component of seismogram, j the discrete time, G,,k(b) the Green's function for the i-th
component by the k-th subfault, and S(6) the source time function. Asterisk represents
T. NISHIMURA, H. NAKAHARA, H. SATO AND M. OHTAKE
^
original
2.62E+07 nm/s
1.63E+07 nm/s
1.25E+07 nm/s
9.54E+06 nm/s
5.59E+06 nm/s
16-32Hz
I it I I I I I I I 1 I I 1 I I I I 1 I I I li
1.93E+06 nm/s
5.59E+05 nm/s
5 SEC/DEV
Fig. 3. Filtered velocity seismograms for vertical component recorded by the STS-2
seismometer at the Tuyama station. Arrow denotes the data span for the analysis of
the radiation intensity of high-frequency waves.
Table 1. Elements of subfaults for the 1994 Sanriku-Oki earthquake
Subfault No. Depth
(km) Dip
(deg.) Moment
( x 1020 Nm) Slip
(m) Resolution
1
2
3
4
5
6
7
8
51
51
32
32
20
20
16
16
32
32
25
25
19
19
12
12
0.25
-0 .04
0.78
1.20
0.25
0.59
0.04
0.18
0.2
0.0
0.8
1.2
0.3
0.6
0.0
0.2
0.69
0.88
0.76
0.40
0.69
0.85
0.77
0.96
125
126 Source Process of the 1994 Sanriku-Oki earthquake
the convolution integral. We determine mk (k=1, , 8) by minimizing a squared error,
defined by
E2(vr,r)=ti[WO- CP.); yr, r)]2 (2)
,=1i=1
where Oi(t) is the i-th component of the observed displacement and n is the total number
of waveform data for each component.
We calculate the Green's functions for each subfault by using the discrete wavenum-
ber method (Bouchon, 1981) and the reflection and transmission matrixes (Kennett and
Kerry, 1979), assuming a horizontally layered structure with four layers as shown in
Table 2 referring to Nishizawa et al. (1992). For the inversion, we use the three-
component waveform data of 180 s (see Fig. 2) by resampling the data with a sampling
frequency of 1 Hz. For smoothing the waveform, the second-order butterworth filter
with a pass band of 0.05-0.01 Hz is applied both to the observed displacement and the
synthetic one. Giving the rupture velocity yr ranging from 2.0 to 3.8 km/s and the rise
time r from 4 to 16 s, which are common to all the subfaults, we estimate eight unknown
parameters (mk) from 540 points of waveform data. After the inversion for each yr and
r, we choose the solution having the minimum residual as the best-fit solution.
2.2. Result
Figure 4 shows the distribution of the residual for different yr and r values, in which
the residual is normalized by the minimum value. It is found from the figure that the
rupture velocity of 3.0 km/s and the rise time of 10 s best fit the observed data. Figure
5 compares the observed waveforms with the best-fit synthetics. The synthetics well
explain the characteristics of the main phases.
In Table 1, we summarize the result for each subfault with diagonal components of
the resolution matrix in the inversion. Total moment release is estimated at 3.2 x 1020
Nm, which is in good agreement with 4.0 x 1020 Nm of the CMT solution by Harvard
University. Slip dk for each subfault is estimated from the relation mk--=fidksk, where
sk is the area of the k-th subfault, and a rigidity of 5.0 X 1010 N/m2 is assumed.
Average slip of the fault is estimated at 0.4 m, and the maximum one amounts to 1.2 m,
which appears at the subfault 4. The regions having larger slip (subfault 3, 4, 6) are
Table 2. Horizontally layered structure for calculat-
ing the Green's function
Layer No. Vp
(km/s) Vs
(km/s) Density
(g/cm3) Thickness
(km)
1
2
3
4
5
3.8
4.8
5.7
6.8
8.1
2.2
2.8
3.3
3.9
4.7
2.2
2.5
3.1
3.2
3-3
4
4
6
5
T. NISHIMURA, H. NAKAHARA, H. SATO AND M. OHTAKE 127
16
14
Cl) 12
10
Lil
C')
E 8
6
4 1,r
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8
RUPTURE VELOCITY(km/s)
Fig. 4. Contour map of residuals as defined in eq. (2) for various sets of rupture velocity
and rise time of the source time function.
observed in the middle to the landward part of the fault plane, which is consistent with
the location of the CMT centroid. On the other hand, the region around the initial break
point (subfault 7 and 8) and the western edge of the fault (subfault 1 and 2) slipped less
than 0.4 m. We note that the portions of larger slip coincide with the turning point of
the curvature of the plate boundary (see Fig. 1).
3. Spatial Distribution of High Frequency Wave Radiation
3.1. Method
We estimate spatial distribution of high-frequency wave radiation on the fault plane
from the envelope of squared ground velocity. First, we assume that the high-frequency
waves are mainly composed of far-field S-waves. We further assume the fault geome-
try and the rupture velocity as estimated in the preceding section, and calculate iso-
chrones of the arrival time of the observed envelope. The isochrone shows which area
of the fault plane contributes to each time on the observed seismogram (e.g., Spudich and
128 Source Process of the 1994 Sanriku-Oki earthquake
Fig. 5. Comparison between the observed (solid curves) and the synthetic (broken
curves) seismograms of the 1994 Sanriku-Oki earthquake.
Frazer, 1984), and the envelope seismogram consists of waves radiated from subareas on
the isochrone. We can express it as
ti2(t)= i(Rrs))2 eia(t — to, (3)
ri
where u2(t) is the summation of three-component squared velocity amplitude and 8 is
a delta function. / represents the subarea number, each of which have a unit area on the
fault. Energy radiation intensity ei shows the radiation intensity on the fault. t1 is the
time when an isochrone intersects the l-th subarea, Rfs the radiation pattern of S-waves,
and r i the hypocentral distance, all of which are given from the fault geometry, the focal
mechanism, and the rupture velocity that were estimated in the preceding section. N
represents the total number of subarea, and ra the arbitrary reference distance. Since
T NISHIMURA, H NAKAHARA, H SATO AND M OHTAKE
40.8
40.4
40.0
40.8
40.4
40.0
40.8
40.4
40.0
Z40.8
0
P 40.4
• *
-1=174
CO 40 0
40.P
40.4
40.0
40.8
40.4
40.0
142
Longi
143
tude(E) 144
1-32Hz
1-2Hz
2-4Hz
4-8Hz
8-16Hz
16-32Hz
2.0
-2 .0
Fig 6. Contour maps of the relative energy radiation intensity, RERI , for six frequency
ranges Darker shade represents stronger intensities Scale is logarithmic, and the
intensities are normalized by the average intensity for each frequency range .
129
130 Source Process of the 1994 Sanriku-Oki earthquake
the envelope consist of the waves radiated from plural subareas, we take the following
steps to obtain a spatial distribution of the high-frequency wave radiation. First, we
calculate squared velocity amplitude at a time and distribute them equally to the
subareas where the isochrone intersects. Then, we estimate r,ge, in eq.(3) correcting the
effects of the hypocentral distance and the focal mechanism. We call r,22ei as the relative
energy radiation intensity (RERI). The spatial distribution of RERI shows the fault area
where strong high-frequency waves are radiated from.
3.2. Result
By using the above method, we inverted the spatial distribution of RERI for six
frequency ranges of 1-32, 1-2, 2-4, 4-8, 8-16, 16-32 Hz. We used the waveform data for
50 s, which corresponds to the time from the initial break to the rupture stopping, and
calculated the isochrones and envelope seismograms at every 1 s. The results are shown
in Figure 6, where dark shade represents high energy radiation and light gray does low
radiation. For all the frequency ranges, it is found that radiation of high-frequency
energy is strong in the western part of the fault. This means that the radiation of high
frequency waves, initially having been weak, was accelerated with the westward propa-
gation of fault rupture. The maximum intensity in the western region of the fault is as
much as 100 times of that in and around the initial break. The patterns of spatial
variation of the intensity seem to be smoother for the higher frequency ranges than lower
frequency ranges, which may be due to the effect of source process and/or wave
propagation in heterogeneous media. However, the resolution of our analysis is not high
enough to distinguish those two possibilities.
4. Discussion and Conclusion
In Figure 7, we compare the intensity of high-frequency wave radiation for each
subfault, slip distribution, and aftershock distribution for the first two weeks. The
intensity of high-frequency waves is shown by RERI for the frequency range of 1-32 Hz.
The amount of slip and the intensity of high-frequency waves are well correlated for the
subfaults 3-8 ; the subfaults 3,4,6 have larger slips and stronger intensities, and the
subfaults 5,7,8 smaller slips and weaker radiation. However, the subfaults 1 and 2,
which are located at the western edge where the rupture stopped, indicate smaller slips
and stronger radiation. Kosuga et al. (1995) also reported a strong energy radiation
near the terminal of the fault from the analysis of arrival times of large amplitude waves
in strong-motion records. The strong radiation may be a result of the high-frequency
waves radiated from rupture stopping area or places of large slip variation as suggested
by Zeng et al. (1993) for the Loma Prieta earthquake. However, coda waves and
reflection phases may partly affect the spatial distribution of high-frequency wave
radiation, therefore, it would be necessary to adopt a method taking heterogeneity of the
structure into account for estimating more realistic distribution.
The largest aftershock (M =6.9) occurred in the subfault 1, where the slip was
T. NISHIMURA, H. NAKAHARA, H. SATO AND M. OHTAKE
Slip(m)
0L
142°E
Relative
4
Energy Radiation
144°E
! !
oL
142°E
Aftershock
N
41.0°
40.5°
40.0°
Intensity
40.8°N
40.0°N
L.
,
Distribution
40.8°N
40.0°N
144°E
M
0
<a° leo
:AP
%
6.9
,o
cif
go co°
0.04,0
ki 0??,serl'"
risetliA
0'90° 0
i;
° _
"6")-74C°
,2,
8,0, oc,8
o
r P °
008.?
06)
a
cp°0 o
!' 0 00
0 o
M=7.5
00 '
8A
- 00
a° 0 0
0 °
142.0° 142.5° 143.0° 143.5° 144.0° E
Fig. 7. Comparison of the aftershock distribution (Dec. 28, 1994-Jan . 10, 1995 ; after
OCPEV) with the spatial distributions of the fault slip and relative energy radiation
intensity, RERI.
131
132 Source Process of the 1994 Sanriku-Oki earthquake
43°
42°
141°
41°
40°
142° 143° 144° 145°
6.)
21
"1
. 2
1968 Tokachi-Oki
1994
, Sanriku-Oki
Fig. 8. Slip distribution of the 1968 Tokachi-Oki earthquake. Amount of slip is shown
in each subfault with a unit of m. The solid circle denotes the midwestern subfault
where larger siip is observed (see text). The data are read from Figure lib of Mori
and Shimazaki (1985). Shaded area represents the fault plane of the 1994 Sanriku-
Oki earthquake.
smallest, 10 days after the main shock. Most of the aftershocks also took place in the
region of small slip of the main rupture. On the contrary, a relatively small number of
aftershocks were observed in the region of subfaults 4 and 6. Similar contrast is often
seen in the source regions of other large earthquakes (e.g., Takeo, 1988).
The 1994 Sanriku-Oki earthquake took place in the area that was ruptured by the
1968 Tokachi-Oki earthquake (Ms=7.9). Figure 8 shows the slip distribution for the
1968 Tokachi-Oki earthquake after Mori and Shimazaki (1985) together with the source
region of the 1994 Sanriku-Oki earthquake (see Fig. 1). Mori and Shimazaki (1985),
analyzing Rayleigh waves with short period of 10-25 s, estimated the slip distribution by
superposing the result on the averaged long-period slip by Kanamori (1971). Hence, it
is necessary to note that the actual contrast of the slip distribution might be stronger
T. NISHIMURA, H. NAKAHARA, H. SATO AND M. OHTAKE 133
than that shown in the figure. In the figure, we recognize that the moment of the 1968
Tokachi-Oki earthquake was mainly released at two regions ; the midwestern edge and
the northern edge of the fault. Comparing the spatial distribution of slip for the two
large earthquakes, it is found that the fault area of the Sanriku-Oki earthquake shares
southern part of the focal region of the Tokachi-Oki earthquake, where slip was
relatively small. We, therefore, conclude that the 1994 Sanriku earthquake ruptured the
portion of plate boundary where slip was incomplete at the time of the 1968 Tokachi-Oki
earthquake of 26 years before. Figure 8 indicates a large slip (2.5 m) for the midwestern
subfault of the Tokachi-Oki earthquake fault. This area roughly corresponds to the
subfault 2 of the present study, where the moment release was nearly zero at the time
of the 1994 Sanriku-Oki earthquake (see Fig. 7 and Table 1).
Mori and Shimazaki (1984, 1985) reported a good correlation between the spatial
distributions of moment release and stress drop for the fault of the 1968 Tokachi-Oki
earthquake. This is in good agreement with our result that the subfaults with large slip
indicate strong radiation of high frequency waves for the 1994 Sanriku-Oki earthquake.
At the subfault 2, however, the strongest radiation of high frequency waves is observed
in spite of small slip of the fault.
The main result is summarized as follows :
(1) The fault of 170 km x 84 km slipped mainly in the region from the middle to the
western (landward) part, having the maximum slip of 1.2 m, whereas a small slip of less
than 0.4 m was observed near the initial break point. The total seismic moment is
estimated at 3.2 x 1020 Nm. The region of larger slip is characterized by low activity of
aftershocks.
(2) The region of larger slip coincides with the region of low seismic moment
release at the time of the 1968 Tokachi-Oki earthquake. This implies that the 1994
Sanriku-Oki earthquake completed the incomplete fault slip in the source region of the
Tokachi-Oki earthquake.
(3) High-frequency wave energy was mainly radiated at the middle to western part
of the fault rather than the initial break region, which is well correlated with the slip
distribution. However, the western edge where the rupture stopped indicates strongest
radiation of high-frequency waves but smaller low-frequency waves, suggesting genera-
tion of high-frequency waves associated with the termination of fault rupture.
Acknowledgments : We wish to thank the staff of the Observation Center for Prediction
of Earthquakes and Volcanic Eruptions, Tohoku University, for providing us with
hypocenter data of the main shock and aftershocks. We thank the Harvard University
group for making available us the CMT solution through computer network. We are
very grateful to Mr. and Mrs. Abe who kindly offered a part of their yard for the seismic
observation.
134 Source Process of the 1994 Sanriku-Oki earthquake
References
Bouchon, M., 1981 : A simple method to calculate Green's functions for elastic layered media, Bull.
seism. Soc. Am., 71, 959-971.
Kanamori, H., 1971 : Focal mechanism of the Tokachi-Oki earthquake of May 16, 1968: Contor-
tion at a junction of two trenches, Tectonophysics, 12, 1-13.
Kennett, B.L.N. and N.J. Kerry, 1979 : Seismic waves in a stratified half space, Geophys. J. R. astr.
Soc., 57, 557-583.
Kosuga, M., K. Imanishi, T. Sato, K. Tanaka and H. Sato, 1995: Focal mechanism and rupture
process of Sanriku-Haruka-Oki (off Sanriku) earthquake, Investigation on the 1994
Sanriku-harukaoki earthquake and the earthquake disaster caused by it, 79-88.
Matsuzawa, T., N. Umino, S. Horiuchi and A. Hasegawa, 1995 : Aftershock activity of the 1994 far
east off Sanriku earthquake observed by a microearthquake observation network, Abstract
of 1995 Japan earth and planetary science joint meeting, A31-08.
Mori, J. and K. Shimazaki, 1984: High stress drops of short-period subevents from the 1968
Tokachi-Oki earthquake as observed on strong-motion records, Bull. seism. Soc. Am. , 74,
1529-1544.
Mori, J. and K. Shimazaki, 1985: Inversion of intermediate-period Rayleigh waves for source
characteristics of the 1968 Tokachi-Oki earthquake, J. Geophys. Res., 90, 11374-11382.
Nishizawa, A., T. Kanazawa, T. Iwasaki and H. Shimamura, 1992 : Spatial distribution of earth-
quakes associated with the Pacific plate subduction off northeastern Japan revealed by
ocean bottom and land observation, Phys. Earth Planet. Interiors, 75, 165-175.
Spudich, P. and L.N. Frazer, 1984 : Use of ray theory to calculate high-frequency radiation from
earthquake sources having spatially variable rupture velocity and stress drop, Bull . seism.
Soc. Am., 74, 2061-2082.
Takeo, M. 1988: Rupture process of the 1980 Izu-Hanto-Toho-Oki earthquake deduced from
strong motion seismograms, Bull. Seism. Soc. Am., 78, 1074-1091.
Umino, N., T. Matsuzawa and A. Hasegawa, 1995 : Focal depth distribution of aftershocks of the
1994 far east off Sanriku earthquake estimated from sP phases, Abstract of 1995 Japan
earth and planetary science joint meeting, A31-09.
Zeng, Y., K. Aki and T. Teng, 1993 : Mapping of the high-frequency source radiation for the Loma
Prieta earthquake, California, J. Geophys. Res., 98, 11981-11983.
