13th World Conference on Earthquake Engineering

Vancouver, B.C., Canada

August 1-6, 2004
Paper No. 2188

SEISMIC RISK ANALYSIS FOR CITY GAS NETWORK IN SOUTHERN

KANTO AREA, JAPAN

Wataru NAKAYAMA1, Yoshihisa SHIMIZU2, Fumio YAMAZAKI3, Junya FUKUOKA4, Ryoji

ISOYAMA5, and Eisuke ISHIDA6

SUMMARY

We constructed a detailed seismic hazard model which shows possibility of ground motion during specific
length of time, on every 50m square mesh based on recent knowledge of seismology. The superposition of
seismic hazard data and damage estimation system on the GIS map makes it possible to calculate seismic
risk of large-scale networks such as city gas supply networks. This paper introduces the recent
development of GIS-based seismic hazard assessment to evaluate seismic risks for city gas networks, and
the examples of its application are demonstrated. The result shows that 1) preventive measure for low
pressure which have enormous length should be avoided because they constitute overspending. 2) In
contrast, emergency response measure for the low-pressure networks is cost-effective.

INTRODUCTION

Several measures to reduce seismic risks had been planned and executed by most companies which are
concerned about risk management. However, the effect of these measures cannot be easily estimated by
companies which have large-scale networks. The reason for the difficulty is that there has not been any
procedure to quantify seismic risks for complicated networks. Consequently, there has not been an answer
to the question “How much risk has been removed by the measure?” or “Which measure for same purpose
is more effective?” Although some researchers have studied earthquake risk analysis since the 1980's, a
large-scale city gas network has not been a subject of such research.

Recently, Tokyo Gas Co. constructed a geographic information system (GIS), which contains data from
60,000 bore holes. It also developed a new disaster mitigation system “SUPREME” (Shimizu et al., 2002)
[1] which includes a damage assessment system capable of calculating damage to low-pressure gas

1
Assistant manager, Tokyo Gas Co. Ltd.., Tokyo, Japan. Email: wataru@[Link]
2
Director, Tokyo Gas Co. Ltd.., Tokyo, Japan. Email: yshimizu@[Link]
3
Professor, Chiba University, Chiba, Japan. Email: yamazaki@[Link]
4
Assistant manager, Japan Engineering Consultants, Tokyo, Japan, Email: fukuokajy@[Link]
5
Director, Japan Engineering Consultants, Tokyo, Japan, Email: Isoyama@[Link]
6
Manager, Japan Engineering Consultants, Tokyo, Japan, Email: isidae@[Link]
pipelines in every 50m square mesh. Superposition of Seismic hazard data on GIS map makes it possible
to calculate seismic risk of large-scale networks such as city gas supply networks. This paper introduces
the recent development of GIS-based seismic hazard assessment to evaluate seismic risk for city gas
networks, and examples of its application, by Tokyo Gas Company.

BASIC FLOWCHART OF RISK ANALYSIS

Fig. 1 shows flowcharts which explain how the process of seismic risk analysis is applied for prioritization
for seismic retrofit. Seismic risk is estimated by direct and indirect expected loss of total gas supply
system, that is,

TLj= Σi(DLj(i)+ILj(i))*Pj(i)j . (1)

Here, TLj :: expected loss at j-th point, DLj(i), ILj(i):direct and indirect loss at j-th point due to ground
motion with the intensity of i, Pj(i): probability of a specific seismic ground motion on the ground surface
at the j-th point caused by i-th earthquake. Direct loss consists of repair cost, work expense for resumption
of gas supply, and the decrease of an income by supply stop / supply reduction. Indirect damage may be
exemplified by secondary damage due to the outbreak of fire, the spread of a fire, etc. resulting from the
damage to pipes.

In this study, seismic hazard is evaluated and applied to complicated city gas networks, and cost
effectiveness of several disaster prevention measures are estimated. This paper profiles a case of
performance for rational earthquake measure investment through prioritization using estimated cost
effectiveness.

Source model Seismicity at each area

Life of facilities Required earthquake

resistance
Seismic hazard for the
facilities
Earthquake resistance of Several kind of risk
the facilities mitigation plan

Expected direct loss

Expected indirect loss Expected mitigation of Cost estimations of each

risks plan

Execution of plans Estimation of cost effectiveness

Choice or prioritization of plans
Possession of remaining risks

Fig. 1: Flow chart of the seismic risk assessment

BASIC FLOWCHART OF HAZARD ANALYSIS

Total process of hazard analysis is shown in Fig. 2.

 Zoning of earthquake activity Source model
Historical earthquake data
a) Area source model
Gutenberg-Richter’s equation

Matsuda’s catalogue (2000)

b) Active fault model
New catalogue of Japan (1991)

Record of seismic activity c) Inter-plate earthquake

Fault parameters model

Seismic activity change between d) Seismicity trend model

1703 and 1923 Kanto earthquake

Attenuation function on engineering base

Seismic hazard on engineering base

SI value amplification map on GIS

Seismic hazard on ground surface

Fig. 2: Flow chart of the seismic hazard analysis

Four types of source model are constructed to calculate seismic hazard: a) area source using Gutenberg-
Richter’s equation calculated from historical earthquake database; b) active fault model taking into
account change in probability, which depends on the time lapse since the latest activity and average
interval of the occurrence of earthquake on each fault; c) inter-plate earthquake model for major recurrent
earthquakes; and d) seismicity trend model for special features of activity change before major
earthquakes in the Southern Kanto area.

As an index of ground motion strength for hazard models, it was decided to employ the SI value, which is
generally used as the criterion for city gas supply suspension or damage presumption. Seismic hazard
evaluation by each source model was performed using the attenuation function for SI value, which is
thought to have a high correlation with damage on the engineering base of Vs=600 (m/s). Next, total
seismic hazard on the engineering base was compounded on the assumption that seismic activity of each
source does not have mutual interaction, in other words, they are statistically summed up independently.
Finally, a fine (50m mesh square) seismic hazard map and hazard curve on the ground surface are
constructed using an amplification map (50m mesh square) made by processing data from 60,000
boreholes [2]. Seismic risk assessment was done using the calculated seismic hazard values. The most
suitable seismic retrofit program or development of a new emergency response system are chosen to
reduce estimated seismic risks.

BASIC IDEA OF SEISMIC HAZARD DEFINED BY “PROBABILITY OF EXCEEDANCE (PE)”

For any given site on the map, the ground motion effect (e.g. SI value, peak ground acceleration) is
calculated at the site for all the earthquake locations and magnitudes believed possible in the vicinity of
the site. Each of these magnitude-location pairs is believed to happen at some average probability per year.
Small ground motions are relatively likely; large ground motions are very unlikely.
Beginning with the largest ground motions and proceeding to smaller ones, we add up probabilities until
we arrive at a total probability corresponding to a given probability, P, in a particular period of time, T. The
probability P comes from ground motions larger than the ground motion at which we stopped adding. The
corresponding ground motion is said to have a P probability of exceedance (PE) in T years. Once many
pairs of PE and ground motion are calculated, seismic risks of any facilities are calculated based on PE at
the same site. In this study, magnitude-location pairs on different sites are categorized as four types of
source model to calculate PE.

CONSTRUCTION OF SOURCE MODEL

Historical seismicity model

We construct an area source model based on historical seismicity data (JMA magnitude +5.0 and depth -
100km earthquakes) from JMA monthly reports (1926 – July 1996) that contain enough data for the
construction. Since source parameters are unreliable and uncertain, we excluded pre-instrumental
seismicity data before 1926, which were constructed on the basis of seismic damage reports. Distribution
of instrumental earthquakes in and near Japan recorded by JMA is shown in Fig.3.

Fig. 3: Distribution of instrumental earthquakes (1926-1996)

Area source models for the Philippine Sea, Pacific, and ‘Continental’ (Eurasia and North-America) plates
are partitioned in accordance with Annaka et al. (2001) [3]. In consideration of the subduction direction
and the depth, the model for the Philippine Sea (Fig. 4), Pacific (Fig.5), and Continental plates (Fig.6) was
prepared in three dimensions. These area source models projected on historical seismicity data, rates and
magnitudes are summarized as Gutenberg-Richter equations,

log( N i [M > m ] = a i − bi m , (2)

for each partition of area source. Here, Ni: cumulative number of earthquakes whose magnitude (M) is
greater than m in i-th area. The values ai and bi are determined by the least square fit to the observed data.
We assume that an occurrence of earthquakes on this model is a random (Poisson) process in time
described as,
 p (τ ) = 1 − exp(ν 0τ ) . (3)

Here, p (τ ) : probability of earthquake occurrence, τ : calculation period, ν 0 : average annual number of

earthquakes.

Fig. 4: Area source model for the Philippine Sea Plate

Fig. 5: Area source model for the Pacific Plate

Fig. 6: Area source model for the North America Plate and Eurasia Plate
Active fault model
To estimate seismic hazard, we employed the maximum magnitude model based on characteristic
earthquake model (Schwarz and Coppersmith, 1984) [4], which proposes that the earthquake of maximum
magnitude occurs repeatedly at almost the same interval. When the latest activity is unknown, the rate of
earthquake occurrence is calculated using Poisson process. If the latest activity can be predicted from the
geological dating of the previous event, we assume the log normal distribution, where probability of
earthquake occurrence increases after the last previous event. We assumed 0.23 for the standard deviation
of the log normal distribution according to the report by Headquarters for Earthquake Research Promotion
(1999) [5]. The rate of earthquake occurrence for time-dependent source is estimated for 100 years
starting from 2001. The configurations and locations of active faults are modeled based on “Active Fault
in Japan –Sheet Maps and Inventories” [6], and fault parameters are based on the latest information in
addition to that of Matsuda et al. (2000) [7]. The distribution of active faults used in this study is shown in
Fig.7, and fault parameters for them are presented in Table 1.

La t e s t Ev e n t i s k n o wn
La t e s t Ev e n t i s un k o wn

Fig. 7: Distribution of active faults in the Kanto Region

 Table 1: Parameters of active faults in the Kanto Region

（ ）
Active fault Length Magnitude Activity Average Recurrence Elapsed
km slip rate interval time since
(m/ky) (ky) latest event
(ky)
Sekiya 40 7.5 A 5.00 0.6 0.3
North Kamogawa 29 7.3 AB 1.00 2.3 10.0
South Kamogawa 26 7.2 AB 1.00 2.1 6.0
Hirai-Kushibiki 20 7.0 B 0.25 7.9 10.0
Motoarakawa 16 6.8 B 0.30 6.9 5.0
Arakawa 20 7.0 B 0.20 8.0 -
Tachikawa 25 7.1 B 0.50 5.0 1.4
Kitatake 15.2 6.8 A 0.40 1.0 1.4
Isehara 13 7.0 B 0.50 3.3 1.1
Kannawa-Kohdu 25 8.0 A 3.00 3.0 3.0
Matsuda
Tanna 30 7.3 A 2.00 1.2 0.07
Fujigawa Kakou 20 8.0 A 7.00 1.1 2.1
Shinanogawa 106 8.0 A 5.00 1.0 0.15
Itigawa-Shizuoka 100 8.0 A 14.00 1.0 1.2
Chubu
Kirigamine 20 7.0 A 5.00 0.3 -
Asama Nishi 10 6.5 AB 1 0.8 -
Fukaya 10 6.5 B 0.3 2.6 -
Konan Imaichi- 16 6.8 C 0.1 12.7 -
Sugaya
Ogose 13 6.7 C 0.05 20.7 -
Tsurukawa 30 7.3 C 0.05 47.7 -
Ougiyama 18 6.9 BC 0.1 14.3 -
Choujagoya 10 6.5 B 0.5 1.6 -
Doushigawa 10 6.5 BC 0.1 7.9 -
Kurokura 10 6.5 B 0.5 1.6 -
Daibosatsu-Touge 14 6.7 C 0.05 22.2 -
Nishi
Sone-Kyuryou 15 6.8 B 0.5 2.4 -
Darumagawa 12 6.6 B 0.5 1.9 -
Sekiguchi-Kuroiso 10 6.5 D 0.005 158.9 -
Sekiguchi-Yonehira 10 6.5 D 0.005 158.9 -
Tokyowan Hokubu 25 7.2 B 0.1 19.9 -
Itsukaichi 10 6.5 C 0.05 15.9 -
Oonajika 10 6.5 B 0.5 1.6 -
Inter-plate earthquake model
Four (Kanto, Kanagawa-ken Seibu, Toukai and Tounankai) earthquakes are chosen to make an interplate
earthquake model. Probability of earthquake occurrence is calculated in the same way as in the active fault
model, based on information about latest activity. The magnitude is assumed to follow the maximum
magnitude model. The Kanto Earthquake, which is recognized as the largest earthquake to occur in the
Kanto area, is of two types: the greater Genroku Kanto (1703) type and the smaller Taisho Kanto (1923)
type. The seismic source of the Genroku Kanto type is off Chiba in addition to a fault below Tokyo Bay,
which slipped during the Taisho Kanto earthquake. In this study, these two faults are assumed to slip
independently. The locations and fault parameters are based on “Fault Parameter Handbook” (1989) [8].
The source models for interplate earthquakes are shown in Fig. 8, and fault parameters, in Table2.

Tokai-Tonankai

Kanto (1703)
Kanagawa Ken
Seibu

Kanto (1923)

Fig. 8: Distribution of the source models for interplate earthquakes

Table 2: Parameters for interplate earthquakes

Earthquake Magnitude Latest event(Yrs.) Interval of events
(Mj) (Yrs.)
Kanto (1703)
(Off Chiba fault) 7.9 1703 440
Kanto (1923) 7.9 1923 220
Tokai-Tonankai 8.1 1944 119
Kanagawa Ken Seibu 7.0 1923 73

Seismic model for increasing probability between major Kanto earthquakes

Seismic activity along the Sagami Trough below the Kanto Region is time-dependent, and increases from
the last large interplate earthquake (Kanto Earthquake) to the next one. Usually, the activity in that region
stops soon after the previous large earthquake, except for aftershocks. Since the area source model based
on historical seismicity (1926-1996) mentioned above was constructed based on data from the dormant
term, the model will underestimate the seismic hazard without any correction.

In this study, the data of seismic activity increase were extracted from a pre-instrumental earthquake
catalogue which records major earthquakes that occurred during the 220 years between the Genroku
Kanto Earthquake (1703) and the Taisho Kanto Earthquake. Earthquakes were extracted selecting only
large earthquake with an SI value larger than 20 cm/sec in a circle with a radius [centering on the Tokyo
Gas’ head office] of 40km. Attenuation function is applied to this extraction process. The cumulative
number of earthquakes selected is plotted in Fig.9. Twenty earthquakes have occurred during the 220
years between two Kanto earthquakes. It is particularly notable that occurrence was concentrated just
before the Taisho Kanto Earthquake. Only two earthquakes remained when the aftershocks of 1923 Kanto
Earthquake were excluded. This change of seismic activity between large inter-plate earthquakes is
quantitatively modeled and used for the analysis to evaluate future increase of seismic risk in Kanto area.
In the above modeling process, aftershocks were excluded by the procedure proposed by the Public Works
Research Institute, Ministry of Construction (1983) [9]. To predict future increase in seismic risk,
exponential shaped function is applied to the data which show the cumulative number of earthquakes, and
the frequency of earthquakes is estimated, as follows.

ν=(0.52exp(0.0163(x+78))-2)/ x (4)
x: Particular period of time to calculate seismic hazard
ν: Frequency of earthquakes in x years (earthquakes/year)
The magnitude is presumed to be Mj = 6.6 as calculated from the average of the seismic moment of the
earthquakes shown in Fig. 9, using the magnitude-moment relationship (Kanamori, 1977) [10]. The
location of seismic source was modeled for distribution at random in the rectangular area (Solid line in Fig.
10) assumed from the distribution of earthquake (Dots in Fig. 10). For example, the rate of earthquake
occurrence in this area is estimated to be about 0.075 (occurrence per year) using Equation (4) applied in
the calculation period of 100 years.

Fig. 9: Relationship between the Kanto Earthquake and the number of earthquakes in the Southern Kanto
Region (SI>20kine)

Fig.10: Area source model of the Southern Kanto Region considering time-dependent seismicity
 SEISMIC HAZARD ON ENGINEERING BASE

The four seismic source models constructed are compounded independently from each other, and the total
seismic hazard on the engineering base (Vs=600m/sec) is calculated. Seismic hazard in each source model
is estimated using the attenuation function (Fig. 11) proposed by Shabestari and Yamazaki (1999) [11].
Fig.12 shows an example of a seismic hazard map obtained on the engineering base level (Vs=600m/s) for
a 9.5% probability of exceedance in 100 years. This probability corresponds to ground motion with a
recurrence period of 1,000 years. Higher seismic hazard is estimated to exist in the southern coastal area,
area around the source of the Tokai (A) and Kanagawaken-Seibu (B) earthquakes, and the Itoigawa-
Shizuoka tectonic line(C).

Log(SI)=-0.785+0.491M-0.00146r-log(r)+0.00359h+ci
M: Magnitude
r: Hypocentral distance
h: Source depth
ci: Local amplification factor
(ci=-0.251 for Vs=600m/sec)

Fig.11: Attenuation function for the SI-value used in the analyses

(C)

(B)
(A)

Fig.12: Seismic hazard map on the base rock level (Vs=600m/s)

(the probability of exceedance in 100 years: 9.5%)
Fig.13 shows the hazard (PE: probability of exceedance) curve estimated on the engineering base level
below the head office of Tokyo Gas Co. In a 100-year period, a value of 30kine has a 14% probability of
exceedance, and 60kine, one of 1.3%. At low ground motion, the effect of the historical earthquake model
and the time-dependent model is larger than that of other models. At a ground motion greater than 30kine,
the effect of the interplate earthquake model is dominant. The analysis did not find any significant effect
with the active fault model. This result shows that there is no active fault whose next earthquake is
expected to occur in near future in the Kanto Region.

SEISMIC HAZARD ON GROUND SURFACE

To take into account the characteristics of surface foundation, SI value amplification factors on GIS
(Fig.14) estimated for every 50m mesh square using data from 60,000 bore holes are multiplied by the
seismic hazard values obtained on the engineering base level for calculation of the seismic hazard at
ground surface. SI value amplification is estimated to be high in the eastern part of the Tokyo Area (A) in
Fig. 14.

Fig.15 shows the SI value distribution at ground surface for the 39.5% probability of exceedance in 100
years. This probability corresponds to ground motion with a recurrence period of 200 years. There is
estimated to be a high seismic hazard in the eastern part of Tokyo (A), where SI amplification is large, in
addition to the southern part (B) where a high seismic hazard is obtained on the engineering base level.

1.000

ec
and
ee
cx
fE
0.100
o
yit
li
ba
b
or Area Sour ce
P0.010
'sr Active Fault
ae Int erplate Ea rthquake
0y
01 Incr easing pr obabilit y
Composed Result
0.001
1 10 100
SI- value( cm/s)

Fig.13: Seismic hazard curve on the base rock (Vs=600m/s)

(A)

Fig.14: Distribution of the amplification ratio of SI-value

 (A)

(B)

Fig.15: Seismic hazard map on the ground surface

(the probability of exceedance in 100 years: 39.5%)

(A)

(B)

Fig.16: Seismic hazard map on the ground surface

(the probability of exceedance in 100 years: 9.5%)
1. 000

e
cn
a
d
e
e
cx
E0. 100
fo
yt
lii
b
a
b
o
r Area Source
P0. 010
s' Act iv e Faul t
ra
e
y0
Int erplat e Eart h quake
Increasin g p robabil it y
0
1 Com po sed Result
0. 001
1 10 1 00
SI-value(cm/s)

Fig.17: Seismic hazard curves on the ground surface

Fig.16 shows the SI value distribution at ground surface for the 9.5% probability of exceedance in 100
years (corresponding to Fig.12). In this map, the SI value is estimated to be larger than 60 kine in the
whole supply area of Tokyo Gas Co. and 100 kine at area (A) and (B). From a different viewpoint, hazard
(PE: probability of exceedance) may also be expressed as curves (Fig. 17) on the ground surface at the
head office of Tokyo Gas Co. A value of 30 kine has 81% probability of exceedance, and 60 kine has 29%
PE in 100 years.

RISK ANALYSIS FOR LOW PRESSURE NETWORK

Estimation of damage probability to low-pressure pipes

Expected values of breaks for low-pressure pipes in a 100-year period are calculated by the damage
estimation function of SUPREME (Shimizu et al., 2002) [1] based on estimated seismic hazard. The
seismic hazard obtained is probability of exceedance (PE), so that the probability P corresponding to a
specific SI value should be determined by differentiating PE with respect to the SI value. The projected
number of pipe breaks (NDLj) at j-th point due to ground motion with the intensity of SIi is calculated
using the following formula.

NDL ij = Pj ( SI i ) × DL j ( SI i ) . (5)

Here, DLj(SIi) is a fragility curve with respect to the SI value, based on the pipe damage experience in
past earthquakes. The total damage probability of the network is obtained by adding i and j. Fig. 18
presents the projected number of pipe breaks in each section (block) of the gas supply network. The total
number of projected pipe breaks in Tokyo Gas supply area was 1,050 for 15 years and 8,800 for 100 years.
These results show pipe breaks comparable to those in the Great Hanshin Earthquake (Kobe) (5,190
breaks in distribution pipes) may be expected to occur in its supply area within the next 100 years.

Fig.18: Expected value of breaks for low-pressure pipes in 15-year period

Fig.19: Expected value of breaks for low-pressure pipes in 100-year period

Choice of measure
In the Kobe Earthquake, numerous breaks in distribution and service pipes in low-pressure networks were
reported, but little damage occurred to pipes with higher earthquake resistance. Thus, damage will be
prevented all pipes with low earthquake resistance are replaced before the next large earthquake. However,
such a preventive measure would require an enormous cost because the extended length of low-pressure
pipes is more than 40,000km.

Another option is the new seismograph (new SI sensor) developed (Koganemaru et al, 2000) [12] to shut
off district regulator (equipment which lowers gas pressure from medium to low) to stop supply of only
low-pressure gas. SUPREME can be used as a real-time damage mitigation system which employs the
new SI sensors with remote shut-off equipment installed at 3,700 district regulators in the Tokyo Gas
supply area. Quick shut-off of gas supply by SUPREME will be able to prevent secondary disaster caused
by numerous leaks from low-pressure pipes. Generally, this kind of emergency measure entails a much
lower cost than the preventive measures mentioned above.

Estimation of cost effectiveness

Here, the cost effectiveness of two methods to prevent secondary disaster due to low-pressure gas is
estimated and compared. While preventive measures such as replacement of low-pressure pipes have not
been carried out by Tokyo Gas, the cost effectiveness is calculated for comparison with that of the
emergency response measures already carried out (see Table 3).

Table 3: Cost effectiveness of measures for low-pressure network

The choice of measure Type of measure Cost effectiveness
Replacement of all low-pressure pipes Preventive < 100%
Installation of New SI sensors at 3700 district Emergency > 100%
regulators
SUPREME system in addition to New SI sensors Emergency > 100%

It was confirmed that the preventive measure, which entails an enormous construction cost, is overly
expensive and to be avoided. In contrast, the cost effectiveness of the emergency response measure is
sufficient.
 CONCLUSION

The new method of assessing seismic hazard for large-scale city gas networks enabled quantitative
estimation of cost effectiveness of plan to prevent secondary disaster By checking cost effectiveness,
companies can avoid overspending for earthquake disaster prevention. The major conclusions of this study
are as follows,
1. New detailed (every 50m mesh square) seismic hazard assessment which consists of four
source models; historical earthquakes, active fault earthquakes, inter-plate earthquakes, and
time-dependent seismicity between great Kanto earthquakes.
2. The new seismic hazard assessment showed that the southern part of Kanagawa and eastern
part of Tokyo have greater possibility of large ground motion than other areas.
3. Damage probability for complicated and vast amount of city gas network can be estimated by
the new seismic hazard assessment.
4. Preventive measures for the low-pressure network should be avoided because they constitute
overspending.
5. In contrast, emergency response measures for the low-pressure network are cost-effective

REFERENCES

1. Shimizu, Y., Yamazaki, F., Nakayama, W., Koganemaru, K., Ishida, E., and Isoyama, R., (2002):
“Development of Super High-density Real-time Disaster Mitigation System for Gas Supply Networks”,
12th European Conference on Earthquake Engineering, Paper No.858, 10p.
2. Ishida, E., Isoyama, R., Yamazaki, F., Shimizu, Y. and Nakayama, W., (2001.8): “Preparation of
Spatial Distribution of Site Amplification Factor of SI Value using GIS”, 26th JSCE Earthquake
Engineering Symposium, Vol.1, pp.421-424.
3. Tadashi ANNAKA, Masayoshi SHIMADA, Tomohiko HIROSHIGE,(2001.8): “Model of estimating
uncertainty for seismic hazard curves of around Kanto region based on Monte Carlo approach”,
Proceedings of the 26th JSCE Earthquake Engineering Symposium, pp.133-136, (in Japanese)
4. Schwarz, D.P. and Coppersmith, K.J. (1984): “Fault behavior and characteristic earthquakes:
Examples from the Wasatch and San Andreas fault zones.”, J. Geophys. Res., 89, 5681―5698.
5. Headquarters for Earthquake Research Promotion (1999): “Regarding the revised proposal ‘About
methods for evaluating long-term probability of earthquake occurrence’” (in Japanese).
6. The Research Group for Active Faults of Japan (ed.), (1991): “Active Fault in Japan-Sheet Maps and
Inventories”, University of Tokyo Press (in Japanese).
7. Tokihiko MATSUDA, Tomomi TSUKAZAKI and Mari HAGINOYA, (2000): “Distribution of active
faults and historical shallow earthquakes of Japanese Islands, with a catalog of on-land seismogenic
faults and earthquakes”, Active Fault Research, No.19, pp.33-54 (in Japanese).
8. Ryosuke SATO (ed.), (1989): “Fault Parameter Handbook”, Kajima Institute Publishing, (in Japanese)
9. Earthquake Disaster Prevention Department, Public Works Research Institute (PWRI), Ministry of
Construction, (1983): “Investigations of the number and scale of foreshocks and aftershocks”, PWRI
document, Nos. 1995 (in Japanese).
10. Kanamori, H., (1977): “The energy release in great earthquakes”, J. Geophys. Res., 82, 2981-2987.
11. Shabestari, T.K., and F. Yamazaki, (1999): “Attenuation relation of strong ground motion indices
using K-NET records”, 25th JSCE Earthquake Engineering Symposium, pp. 137-140.
12. Koganemaru, K., Shimizu Y., Nakayama W., Yanada T., Furukawa H., and Takubo K., (2000):
“Development of a New SI Sensor”, 12th World Conference on Earthquake Engineering.