13th World Conference on Earthquake Engineering

Vancouver, B.C., Canada

August 1-6, 2004
Paper No. 793

PROCEDURE FOR MAKING PROBABILISTIC SEISMIC HAZARD MAP

AND UNDERSTANDING OF THE EVALUATED HAZARD
NAKAO Yoshihiro1 and KUSAKABE Takaaki2

SUMMARY

Past earthquake records have been primarily used for producing probabilistic seismic hazard maps. In the
hazard analyses earthquakes with various magnitudes are assumed to occur randomly in time and space
based on past earthquake records. However, earthquakes with particular magnitudes and recurrence
intervals occur repeatedly on active faults and subduction zones. In the present paper we discuss a
procedure to evaluate seismic hazards based on past earthquake records, active faults and inter-plate
earthquakes. Resultant seismic hazard map is shown for all Japan. Japan Meteorological Agency Seismic
intensity records, which have been observed at various sites in Japan for long periods, are employed for
understanding of the evaluated seismic hazard.

INTRODUCTION

Past earthquake records have been primarily used for producing probabilistic seismic hazard maps (e.g.
Arakawa [1] and Cornell [2]). Those maps have been applied to form regional classification maps in
seismic design codes for various civil infrastructures in Japan so that regional seismicity is incorporated
into seismic design motions. Past earthquake records used for producing seismic hazard maps are based
on instrumental observation and historical descriptions. Though these records date back as long as one
thousand and hundreds years, they are still insufficient to evaluate seismic hazard due to active faults,
because recurrence intervals of active faults are generally longer than thousands of years. Besides,
earthquakes with various magnitudes are assumed to occur randomly in both time and space in the
analysis based on past earthquake records, however earthquakes with particular magnitudes and
recurrence intervals occur repeatedly on active faults and subduction zones. Considering insufficiency of
past earthquake records and the events repeatedly occurring on active faults and subduction zones, active
faults and inter-plate earthquakes should be taken into consideration in seismic hazard analysis separately
from past earthquake records. In the present paper we discuss a procedure to evaluate seismic hazards
based on past earthquake records, active faults and inter-plate earthquakes (Nakao et al.[3]). Resultant
seismic hazard map is shown for all Japan.

1
Researcher, Earthquake Disaster Prevention Division, Research Center for Disaster Risk Management, National Institute for Land and
Infrastructure Management, Ministry of Land, Infrastructure and Transport, Japan, Email: [email protected]
2
Head, Earthquake Disaster Prevention Division, Research Center for Disaster Risk Management, National Institute for Land and
Infrastructure Management, Ministry of Land, Infrastructure and Transport, Japan, Email: [email protected]
In Japan seismic intensities have been recorded at various sites for long periods by Japan Meteorological
Agency. We employ the seismic intensity records to compute observational seismic hazard. Note that
almost all the seismic intensity records are attributed to the earthquakes that are assumed to occur
randomly in time and space in the present procedure. Therefore, the seismic hazard based on past
earthquake records is compared with the observational hazard in order to understand the evaluated hazard.
Through the comparison, it is recognized that regional differences of the evaluated hazard roughly agree
with those of the observational hazard. In the present procedure characteristics of the earthquakes that
occur repeatedly on active faults and subduction zones are evaluated based on the latest seismological and
geological researches, and they are reflected in hazard evaluation. The seismic hazard based on past
earthquake records is complemented by the hazard based on active faults and inter-plate earthquakes.

PROCEDURE FOR MAKING PROBABILISTIC SEISMIC HAZARD MAP

Seismic Hazard Based On Past Earthquake Records

Earthquake Catalogs
Earthquake catalogs adopted in the present study are as follows:
-Usami Catalog [4] for 416-1884
-Utsu Catalog [5] for 1885-1925
-Japan Meteorological Agency (JMA) Catalog [6] for 1/1926- 7/1996
Figs.1(a) and (b) show cumulative numbers of earthquake records against time for events with Mj<6.0 and
Mj≥6.0 (Mj is Japan Meteorological Agency Magnitude), respectively. Since evident accumulation of
earthquake records can be found from 1926 and 1885 for Mj<6.0 and Mj≥6.0, respectively, we
incorporate the following catalogs into analysis:
- JMA Catalog for Mj<6.0
- Utsu and JMA Catalogs for Mj≥6.0
3000 500
C um ulative N um ber of Earthquakes

C um ulative N um ber of Earthquakes

5.0≦Ｍ＜5.5 6.0≦Ｍ＜6.5
2500 5.5≦Ｍ＜6.0 6.5≦Ｍ＜7.0
400 7.0≦Ｍ＜7.5
7.5≦Ｍ＜8.0
2000
300
1500

200
1000

100
500

0 0
1500 1600 1700 1800 1900 2000 1500 1600 1700 1800 1900 2000
Tim e（
year） Tim e（year）
Fig.1 (a) Mj<6.0 (b) Mj≥6.0
Fig.1 Cumulative Number of Earthquake Records

Fig.2 shows the epicenters of past earthquakes included in the catalogs used for analysis. Note that we
exclude records of the earthquakes attributed to active faults and the inter-plate earthquakes from the
catalogs so that these earthquakes would not be doubly considered in hazard evaluations based on past
earthquake records. The excluded earthquakes are considered in analysis based on active faults and inter-
plate earthquakes separately from past earthquake records.
Based on seismotectonics around Japan after Hagiwara [7], we develop background zones in which
uniform seismicity is assumed individually. The background zones are shown in fig.3. The largest
magnitude of past earthquakes in each background zone is adopted as the maximum magnitude. The
earthquake records used for assuming maximum magnitudes are included in the three catalogs (after
Usami [4], Utsu [5] and JMA [6]). If magnitude of earthquake attributed to active fault or that of inter-
plate earthquake is adopted as maximum magnitude, we adopt the largest one of the other events as
maximum magnitude [8]. Maximum magnitude in each background zone is shown in Table 1. We do not
employ magnitudes of earthquakes which occur repeatedly on active faults and subduction zones as
maximum magnitudes because these earthquakes are considered in analysis based on active faults and
inter-plate earthquakes separately from past earthquake records.

1 08

1 10
10 7 10 4

10 1

10 5
1 12 10 2
1 09

11 1

1 06
1 26 10 3
1 24
1 15
12 1
12 5
1 18 11 4
12 2
1 23 11 9 11 6
131 1 20 11 3
11 7
132

1 33 12 9
12 7

1 34
1 30
12 8

Fig. 2 Epicenters of Past Earthquakes Fig.3 Background Zones

Table1 Earthquake Occurrence Modeling for Each Background Zone

Upper layer of background zone Lower layer of background zone
Upper and Lower layers of background zone
Background Number Minimum Maximum （0 to 30 (km) forcal depth） （30 to 100 (km) forcal depth）
Zone of events Magnitude Magnitude Average focal Mean numbers of events Average focal Mean number of events Average focal Mean number of events
a value b value
depth(Km) per year (numbers/year) depth(Km) per year (number/year) depth(Km) per year (number/year)
101 527 5.0 8.1 4.76 0.84 36.7 3.62 14.5 1.64 55.1 1.98
102 771 5.0 8.5 4.62 0.77 24.4 5.82 10.0 3.82 53.6 2.00
103 272 5.0 8.0 4.37 0.80 33.1 2.34 13.7 1.18 52.8 1.16
104 411 5.0 7.8 5.06 0.90 48.6 3.68 14.6 0.85 58.8 2.83
105 432 5.0 8.5 5.02 0.88 37.5 4.34 17.2 1.85 52.6 2.49
106 628 5.0 8.0 4.58 0.79 33.7 4.50 12.1 1.98 50.6 2.52
107 83 5.0 7.3 4.41 0.89 49.8 0.92 14.7 0.34 70.6 0.58
108 3 5.0 7.3 1.36 0.56 20.0 0.04 10.0 0.02 40.0 0.01
109 292 5.0 7.5 5.39 0.97 52.6 3.34 12.1 0.70 63.3 2.64
110 18 5.0 7.3 5.05 1.13 16.7 0.26 10.0 0.22 66.7 0.04
111 73 5.0 7.6 4.74 0.96 7.0 0.88 10.0 0.86 90.0 0.02
112 165 5.0 7.8 6.37 1.14 19.7 4.66 12.6 3.62 44.4 1.05
113 174 5.0 7.4 5.47 1.02 42.4 2.33 13.6 0.90 60.4 1.43
114 72 5.0 7.5 3.04 0.64 49.0 0.65 12.3 0.20 65.2 0.45
115 212 5.0 7.3 4.93 0.92 46.1 2.14 12.9 0.59 59.0 1.54
116 157 5.0 7.3 5.11 0.97 18.9 1.86 10.5 1.48 51.6 0.38
117 71 5.0 7.8 3.66 0.76 18.6 0.69 10.3 0.56 55.5 0.13
118 37 5.0 8.0 4.45 0.97 14.5 0.41 10.0 0.35 41.0 0.07
119 39 5.0 8.0 4.02 0.88 32.4 0.40 10.0 0.20 56.0 0.19
120 41 5.0 8.0 6.57 1.36 25.7 0.58 10.0 0.33 46.3 0.26
121 18 5.0 7.5 4.95 1.13 12.8 0.21 10.0 0.18 43.3 0.04
122 13 5.0 7.8 4.84 1.10 16.3 0.23 10.0 0.19 52.5 0.04
123 20 5.0 7.8 2.62 0.67 42.7 0.19 15.0 0.07 61.1 0.11
124 49 5.0 7.3 4.50 0.94 7.0 0.65 10.0 0.63 37.0 0.01
125 107 5.0 7.4 4.71 0.93 8.4 1.12 10.0 1.09 41.3 0.03
126 17 5.0 7.3 5.03 1.11 22.3 0.29 16.9 0.22 40.0 0.07
127 232 5.0 7.8 5.08 0.92 31.1 3.05 11.3 1.59 52.6 1.46
128 127 5.0 8.0 4.03 0.77 46.9 1.52 13.7 0.45 61.1 1.06
129 70 5.0 8.0 4.64 0.96 29.0 0.69 10.0 0.43 66.7 0.26
130 87 5.0 8.0 4.85 0.98 30.3 0.84 10.0 0.46 58.9 0.38
131 10 5.0 7.3 3.22 0.82 18.8 0.13 10.0 0.11 100.0 0.01
132 34 5.0 7.3 4.82 1.03 5.9 0.47 10.0 0.47 - 0.00
133 17 5.0 7.3 4.15 0.98 20.7 0.18 10.0 0.14 57.5 0.04
134 33 5.0 7.3 3.19 0.72 28.9 0.37 10.0 0.23 67.6 0.13
It is possible that causative faults do not appear clearly on the ground surface even in the case of large-
scale earthquakes such as The Western Tottori prefecture earthquake in 2000(Mj=7.3). The earthquakes
attributed to concealed active faults are not considered in analysis based on active faults. Besides, it is
possible that the large-scale earthquakes were not recorded in historical descriptions if they occurred on
concealed active faults long years ago. These earthquakes are not considered also in analysis based on past
earthquake records if magnitude of the events are larger than the adopted maximum magnitude in
background zone. Therefore, Mj=7.3 of The Western Tottori prefecture earthquake is employed as lower
limit of maximum magnitude in each background zone in order to consider concealed active faults. There
are nine background zones that adopt the under limit as maximum magnitude. Mj=5 is employed as
minimum magnitude in every background zone.

Gutenberg-Richter relationship given by eq.(1) is assumed to represent frequency distribution of

earthquake magnitude for each background zone.

log Ni [M>m] = ai− bi m (1)

where,
Ni: Number of earthquakes per year with magnitude greater than m within i-th background zone
ai, bi: Coefficients for i-th background zone

Based on earthquake records within i-th background zone, ai and bi–values are determined by method of
maximum likelihood as shown in Table 1. Mean earthquake occurrence rate per year and area in i-th
background zone is computed by eq.(2).

(10a i −bi MiL − 10 a i −bi MiU )

νi = (2)
Ai
where,
Ai: Area of i-th background zone
MiL: Minimum magnitude of earthquakes considered in analysis. (Mj = 5.0)
MiU: Maximum magnitude in i-th background zone

Figs.4 (a) and (b) show epicenters with focal depths of 0 to 30 (km) and 30 to 100 (km), respectively.
According to figs.4 it is evident that almost all events in north side of Japan and Japan Sea occur at the
depth of 0 to 30(km). In order to consider the seismicity and ground motion intensity depending on focal
depth we divide background zones into upper and lower layers. Upper layer of background zone is 30km
thick surface part of the zone and lower layer of zone is 70km thick part underneath the upper layer.
Earthquakes are assumed to occur in both layers in analysis. Mean earthquake occurrence rate per year
and area in upper and lower layers are computed by eq. (3) and (4), respectively. Uniform earthquake
occurrence rate per year and area is assumed in each layer.

N iupper
ν iupper = ν i (3)
N iupper + N ilower
N lower
ν lower
i = νi i
(4)
N iupper + N lower
i

where,
N iupper : Number of earthquake records in upper layer of i-th background zone
N ilower : Number of earthquake records in lower layer of i-th background zone
νi: Mean earthquake occurrence rate per year and area in i-th background zone
 ν iupper : Mean earthquake occurrence rate per year and area in upper layer of i-th background zone
ν ilower : Mean earthquake occurrence rate per year and area in lower layer of i-th background zone

Probability density function of magnitude is derived from eq.(1) as

b i exp[−b i (m − Mi L )]
fi M (m) = (5)
1 − exp[−b i (Mi U − Mi L )]
where,
fiM(m): Probability density function of magnitude

10 8 108

110 110
10 7 1 04 107 104

101 101

10 5 105
112 1 02 11 2 102
10 9 109

111 1 11

106 1 06
12 6 10 3 126 103
1 24 12 4
115 1 15
121 12 1
12 5 125
1 18 11 4 11 8 114
122 12 2
123 119 116 123 11 9 1 16
131 12 0 113 131 120 1 13
117 11 7
132 132

133 129 1 33 1 29
127 127

1 34 13 4
130 1 30
128 128

(a) 0≤h<30 (km) (b) 30≤h<100 (km)

Figs.4 Epicenters of Past Earthquakes with Different Focal Depth h

Ground Motion Attenuation Relation with Distance

In the present study peak ground acceleration (PGA) is estimated by attenuation relation after Annaka et
al.[9], which is given by eq.(6).

log X = 0.606M + 0.00459H − 2.136 log( R + 0.334e0.653 M ) + 1.730 (6)

where,
X: Peak ground acceleration [gal]
M: Japan Meteorological Agency magnitude
H: Focal depth [km]
R: The shortest distance between site and fault plane [km]

PGA on the outcropped bedrock whose shear wave velocity is larger than 300 to 600m/sec is estimated by
eq. (6). Although they use the shortest distance to a fault plane as a distance parameter, we substitute
hypocentral distance for the shortest distance. Average depth of focuses in each layer of background zones
is used as focal depth parameter in the attenuation relation. 10km is given to focal depth parameter as
lower limit. For incorporating the scatter of ground motion estimated by attenuation relation into analysis,
±2σ variation around mean value is considered, where σ represents a standard variation of attenuation
equation. Suppose an earthquake with magnitude m occurs at a distance r from the site, probability that
PGA X exceeds a specific level x is expressed as
 ∞

∫
P x i [X > x | m, r, h ] = fx i (X | m, r, h )dX
X
(7)

where,
fxi(X |m,r,h): Probability density function of PGA generated by an earthquake with magnitude m at a
distance r and focal depth h.
m: Japan Meteorological Agency magnitude
r: The shortest distance between site and fault plane [km]
h: Focal depth [km]

Hazard Evaluation Based on Past Earthquake Records

Combining eqs.(3), (4) ,(5) and (7), probability that PGA X exceeds x during a period of TD can be given
by eq.(8).

Ph [X > x | TD ] = 1 − exp(−λTD ) (8)

Mi U
λ= ∫ ∫ (ν > x | m, r , H iupper ]+ ν ilower fi M ( m) P x i [ X > x | m, r, H ilower ])dmds
upper
i fi M ( m) P x i [ X (9)
A Mi L
where,
Ph[X＞x|TD]: Probability that PGA X exceeds x during a period of TD
TD: Period [year]
λ: Probability that PGA X exceeds x during a year
H iupper : Average depth of focuses in upper layer of i-th background zone
H lower
i : Average depth of focuses in lower layer of i-th background zone
A: Area in which all the assumed earthquakes are considered in analysis

Seismic Hazard Based of Active Faults

Active Faults for Analysis

We employ the following two kinds of active faults for analysis. Active faults categorized in (2) are also
assumed to generate independent earthquakes in the same way as the faults in (1). Fig.5 shows locations
of seismogenic and active faults.

(1)Seismogenic faults after Matsuda[10]: Active faults or groups of active faults that may produce
independent large earthquakes
(2)Active faults with length of 10km or longer, which are not categorized as seismogenic faults
(Research group for active faults[11])

Magnitudes and Occurrence rates

Matsuda [12] derived relationships among fault length, dislocation and earthquake magnitude as eqs. (10)
and (11). Introducing an average slip rate of fault, we can evaluate mean recurrence interval by eq. (12).
When a wide range of average slip rate is given for an active fault, we use the middle value of the range;
i.e. when 1 to 10[mm/year], 0.1 to 1[mm/year] and 0.01 to 0.1[mm/year] are given as ranges, we use
5[mm/year], 0.5[mm/year] and 0.05[mm/year] as an average slip rate, respectively. When a number of
average slip rates are obtained along fault line, we use the largest rate of them.

Mj = ( log( Lj ) + 2.9 ) / 0.6 (10)

Mj =(log( Dj )+4.0) / 0.6 (11)
log(TRj) = log(Lj / νj)+1.9 (12)
where,
Mj: Magnitude
Lj: Fault length [km]
Dj: Dislocation of fault rupture [m]
TRj: Mean recurrence interval [year]
ν j: Average slip rate [m/year]

Fig.5 Seismogenic Faults and Active Faults

The Headquarters for Earthquake Research Promotion [13] was installed by Prime Minister’s office after
the 1995 Kobe Earthquake. The Headquarters is promoting survey at 98 major active faults. Ministry of
Education, Culture, Sports, Science and Technology [14] and Geological Survey of Japan[15] also
conduct surveys on active faults. We incorporate newly obtained information by those surveys into
analysis. When the occurrence time of the latest event is known, we assume a time-dependent stochastic
process model for earthquake occurrence, which is given by eq.(13). We use a Brownian passage time
distribution for this model [13]. When the occurrence time of the latest event is unknown, we employ
stationary Poisson process for earthquake occurrence as shown by eq.(14)[13].
Fj (t0j + TD) − Fj(t0j )
Pj[TD] = (13)
1− Fj(t0j)
TD
TRj
Pj[TD ] =1− e (14)
where,
TD: Time interval to calculate probability of earthquake occurrence [year]
TRj: Mean recurrence interval [year]
Fj(t): Brownian Passage Time distribution function of recurrence interval
t0j: Elapsed time from the latest event [year]
Pj[TD]: Probability of earthquake occurrence

Ground Motion Attenuation Relation with Distance

We employ ground motion attenuation relation after Annaka et al.[9] for analysis based on active faults
the same as the case of past earthquake records. The shortest distance to a fault plane is used as distance
parameter. In eq.(6) focal depth is defined as depth of central point on the fault plane. Assuming all the
fault widths to be 13km and all the fault dip to be 90˚, we adopt 6.5km as focal depth. The assumption of
fault width is based on the followings. As for more than 70% of active faults earthquake magnitudes are
estimated to be larger than or equal to Mj6.8. According to the relationship between magnitudes and fault
widths after Takemura [16], fault widths of earthquakes with Mjs6.8 saturate at 13[km].

In the same manner as eq.(7), probability that PGA X exceeds a specific level x due to an earthquake
generated by j-th active fault is written as eq. (15).
∞

∫
P x j [X > x | M j , r,6.5] = fx j (X | M j , r,6.5)dX
X
(15)

Hazard Evaluation Based on Active Faults

Using eqs. (13)-(15), we can calculate probability that PGA X exceeds x during TD years due to j-th active
fault by eq.(16). Probability that PGA X exceeds x during TD years due to all active faults is computed by
eq.(17).

Pf j [X > x, TD ] = Pj [TD ]Px j [X > x | M j , r ] (16)

Pf [ X > x , TD ] = 1 − ∏ {1 − P
j
f j[X > x , TD ] } (17)

Seismic Hazard Based on Inter-plate Earthquakes

Inter-Plate Earthquakes for Analysis

Large-scale earthquakes that occur repeatedly in subduction zones are considered as inter-plate
earthquakes. Inter-plate earthquakes introduced in analysis are shown in fig.6 and table2. In the followings
we discuss fault planes, magnitudes, and recurrence intervals established for each inter-plate earthquake.

46

44
The Northen Etrofu Earthquake(M8.1)
The Southern Etrofu Earthquake(M8.1)
42 The Hokkaido Toho-oki Earthquake(M7.8)
The Nemuro-oki Earthquake(M7.4)
The Northern Tokachi-oki Earthquake(M8.2)
40
The Southern Tokachi-oki Earthquake※
（Thin line：The 1856 Southern Tokachi-oki Earthquake (M7.8) ,
Thick line：The 1968 Southern Tokachi-oki Earthquake (M7.9))
38
The Miyagiken-oki Earthquake(M7.5)

36

34 The Kanto Earthquake※

(Thin Line：The 1703 Kanto Earthquake(M 8.2), Thick line：
The 1923 Kanto Earthquake(M7.9 ))
The Tokai Earthquake(M8.0)
32 The Tonankai Earthquake(M8.1)
The Nankai Earthquake(M8.4)
※Thin and Thick lines show patterns of earthquake occurrences
30
128 130 132 134 136 138 140 142 144 146 148 150 152
Fig.6 Inter-plate Earthquakes
 Table 2 Inter-plate Earthquakes Considered in Analysis
Mean The Latest
α for BPT
Recurrence Earthquake
Inter-plate Earthquakes Magnitude Distributio
Interval Occurrence
n Model
[year] Time
The Nankai Earthquake M8.4 90.1 12/21/1946 0.20
The Tonankai Earthquake M8.1 86.4 12/7/1944 0.18
The Tokai Earthquake M8.0 118.8 12/23/1854 0.24
The Kanto Earthquake M7.9 or M8.2 219.8 9/1/1923 0.24
The Miyagiken-oki Earthquake M7.5 37.1 6/12/1978 0.18
The Southern Tokachi-oki Earthquake M7.8 or M7.9
The Northern Tokachi-oki
M8.2
Earthquake
The Nemuro-oki Earthquake M7.4
The Hokkaido Toho-oki
M7.8 57.0 6/30/1962 0.18
Earthquake
The Southern Etorofu
M8.1
Earthquake
The Northern Etorofu
M8.1
Earthquake

Fault Planes, Magnitudes, Recurrence Intervals, The Latest Event Time

-Earthquakes in the Pacific Ocean off the Coast of Hokkaido and North Tohoku Region areas
The Headquarters for Earthquake Research Promotion [17] suggests that inter-plate earthquakes in the
Pacific Ocean off the coast of Hokkaido and North Tohoku Region areas occur within a relatively short
period of time, and their source regions do not overlap as shown in fig.7. These earthquakes can be
explained in the following way. Strain accumulates over a period of several decades to 100 years in the
area adjoining the Chishima (Kuril) Trench due to the subducting Pacific Plate. After this strain has
approached its limit, it is released by a series of inter-plate earthquakes. As a result, a series of large-scale
earthquakes occurs along the trench in a short time with no overlapping of focal regions. Past earthquakes
in this region are categorized into different sets of event series by Utsu[18] as shown in fig.8. We employ
time interval of median years in last two sets of event series as a mean recurrence interval in common for
inter-plate earthquakes along the Chishima Trench. The median year of the latest set of event series is
employed as the latest event time in common for all the inter-plate earthquakes. We suppose source
locations and magnitudes of earthquakes in this region based on past events [22]. Note that we suppose
two patterns with same probability to occur as for magnitudes and source locations of the Southern
Tokachi-oki Earthquake as shown in fig.6 and table2.

ch
r en
aT
i shim
Ch

Fig.7 Earthquakes in the Pacific Ocean off the Coast of Hokkaido and North Tohoku Region Areas
 A B C D E F Time Interval
1763 unclear 1780 Time Interval of a Series
M8 M8 17 years
of Earthquake Occurrences
59 years
1856 1839 1843 Time Interval of a Series
unclear 17 years
M8 M7.5 M8+ of Earthquake Occurrences
37 years
1894 1893 1918 * 1918 Time Interval of a Series
M7.9 M7.75 M7.7 M8.0 25 years of Earthquake Occurrences
34 years Mean Recurrenc e Interval 57(years)
1968 1952 1973 1969 1958 1963 Time Interval of a Series
M7.9 M8.2 M7.4 M7.8 M8.1 M8.1 21 years of Earthquake Occurrences
# Source P lace of the 1896 Sanriku-oki Earthquake o verlapped.
* The 1918 Earthquake may have occurred in F zone.
Fig.8 Past Earthquakes in the Pacific Ocean off the Coast of Hokkaido and North Tohoku
Region Areas

-The Miyagi-oki Earthquake, the Tonankai Earthquake and the Nankai Earthquake
We suppose source locations, magnitudes, mean recurrence intervals, and the latest event time according
to The Headquarters for Earthquake Research Promotion [13].

-The Kanto Earthquake

Mean recurrence interval of the Kanto earthquake is evaluated as 220 years based on past Kanto
earthquakes in 1703 and 1923. The year 1923 of the last Kanto earthquake is employed as the latest event
time. Similar to the case of the South Tokachi-oki earthquake, we suppose two patterns with same
probability to occur as for magnitudes and source locations as shown in fig.6 ([19] and [20]).

-The Tokai Earthquake

We suppose source location and magnitude for the Tokai Earthquake according to Central Disaster
Prevention Council[21]. Mean recurrence interval and the latest event time are supposed as shown in
table2 based on series of events in Z region shown in fig.9 and table 4.

For evaluating event rates of the inter-plate earthquake Brownian Passage Time distribution function for
recurrence interval is employed [13].

Ground Motion Attenuation Relation

Eq.(6) is used for ground motion estimation from inter-plate earthquakes. The shortest distance to fault
planes is employed as distance parameter. Depth of the center point on a fault plane is used as focal depth
in the attenuation relation. For incorporating the scatter of ground motion estimated by attenuation relation
into analysis, ±2σ variation around mean value is considered, where σ represents a standard variation of
attenuation equation.

Z region

Y region

X region

Fig.9 Source Regions for The Nankai Earthquake, The Tonankai Earthquake, and The Tokai
Earthquake
 Table 4 Source Regions
Event Time Earthquake X Region Y Region Z Region
9/20/1498 The Meio Tokai Earthquake ∗∗ ∗
2/3/1605 The Keicho Earthquake ∗∗ ∗
10/28/1707 The Hoei Earthquake ∗∗ ∗∗ ∗~∗∗
12/23/1854 The Ansei Tokai Earthquake ∗∗ ∗∗
12/24/1854 The Ansei Nankai Earthquake ∗∗
12/7/1944 The Showa Tonankai Earthquake ∗∗
12/21/1946 The Showa Nankai Earthquake ∗∗
（∗∗: Source zone overlaps almost all the region, ∗: Source zone overlaps part of the region）

Hazard Evaluation Based on Inter-plate Earthquakes

Recurrence intervals of inter-plate earthquakes are shorter than those of active faults. Therefore, we
consider three times occurrences for each inter-plate earthquake in hazard analysis [22]. Probability that
PGA X exceeds x during TD years due to i-th inter-plate earthquake is computed by eq.(18). Probability
that PGA X exceeds x during TD years due to all inter-plate earthquakes is evaluated by eq.(20).
3
Pp i [X > x , TD ] = 1 − ∏ (1 − P [T
n
i D , n ]Px i [X > x | m i , r, h ]) (18)
∞

∫
P x i [X > x | M i , r ] = fx i (X | M i , r, h )dX
X
(19)

N
Pp [ X > x, TD ] = 1 − ∏ (1 − P
i
p i [X > x, TD ]) (20)

where,
Pi [TD , n ] : Probability that i-th inter-plate earthquake occurs once more following to n-1 times
occurrences of it during TD years
Pp i [ X > x , TD ] : Probability that PGA X exceeds x during TD years due to i-th inter-plate earthquake
P xi[X > x | M i , r, h ] : Probability that PGA X exceeds a specific level x due to an event of magnitude Mi
Pp [X > x, TD ] : Probability that PGA X exceeds x during TD years due to all inter-plate earthquakes

Seismic Hazard Based on Past Earthquake Records, Active Faults and Inter-plate Earthquakes

We calculate comprehensive seismic hazard due to past earthquakes, active faults and inter-plate
earthquakes by eq.(21) on the assumption that these three earthquake sources are independent each other.
P[ X > x , TD ] = 1 − (1 − Ph [ X > x | TD ])(1 − Pf [ X > x , TD ])(1 − Pp [ X > x , TD ]) (21)
where,
Ph[X>x, TD]: Probability that ground motion X from background zones exceeds x during TD years
Pf[X>x, TD]: Probability that ground motion X from active faults exceeds x during TD years
Pp[X>x, TD]: Probability that ground motion X from inter-plate earthquakes exceeds x during TD years.

Resultant Seismic Hazard Map

Figs.10(a) and (b) present seismic hazard maps in which past earthquakes, active faults and inter-plate
earthquakes are incorporated. PGA with 63% and 5% probability of exceedance during 100 years from the
year 2003 are shown in figs.10(a) and (b), respectively.
According to the numerical results in figs.10, the active faults with high event rates and inter-plate
earthquakes affect seismic hazards nearby. Strong ground motions are estimated widely around areas with
high dense distributions of active faults.
 0
20
40
60
80
100
120
140
160
180
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
950
1000
1100
1200
1300

(c m /s 2 )

(a) 63% probability of exceedance (b) 5% probability of exceedance

Fig.10 PGA with 63% and 5% Probabilities of Exceedance in 100 Years from The Year 2003

UNDERSTANDING OF THE EVALUATED SEISMIC HAZARD

Evaluated Seismic Hazard According to the Present Procedure

According to the present procedure seismic hazards are evaluated at the sites, where Japan Meteorological
Agency (JMA) has announced the JMA seismic intensities for longer than 50 years. The evaluated seismic
hazards are compared with observational seismic hazards that are computed from the announced seismic
intensity records. The seismic intensity, which provides a measure of the strength of seismic motion, is
divided into 10 scales. The relationship between seismic intensity on JMA scale and Modified Mercalli
intensity is shown in fig.11 for reference.
5 Upper 6 Upper
5 Lower 6 Lower
JMA Seismic Intensity 0 1 2 3 4 7
JMA Measured 4.5 5.0 5.5 6.0
Seismic Intensity ＜0.5 0.5-1.5 1.5-2.5 2.5-3.5 3.5-4.5 -5.0 -5.5 -6.0 -6.5 6.5＜
Modified Mercalli 1 2 3 4 5 6 7 8 9 10 11 12
Seismic Intensity
Fig.11 JMA Seismic Intensity and MM Intensity [23]

Note that we evaluate seismic hazards based on only past earthquake records for the comparison with the
observational ones because almost all the observed seismic intensity records are from the earthquakes that
are assumed to occur randomly in time and space in the present procedure. In the hazard evaluation
probability that JMA seismic intensity x is larger than or equal to intensity 4 during a year is computed.

In hazard analysis Seismic intensities are estimated by attenuation relation after Shabestari, K. et al.[24],
which is given by eq.(22). As shown in fig.11 seismic intensity is divided into 10 scales based on
measured seismic intensity, which is continuous value measured by seismic intensity meters. The
measured seismic intensity is estimated by eq.(22) and fig.11 yields seismic intensity based on the
estimation. We determine station coefficient in eq.(22) so that the sum of squares of the residual that is
defined as differences between estimated seismic intensities and observed ones at each site is minimized.
For incorporating the scatter of measured seismic intensity estimated by attenuation relation into analysis,
±2σ variation around mean value is considered, where σ represents a standard variation of attenuation
equation.
I = −0.087 + 1.053M − 0.00256r − 1.89 log r + 0.00496h + c (22)
where,
I : Measured seismic intensity
 M: Japan Meteorological Agency magnitude
r: The closest distance to fault rupture
h: Source depth
c: Station coefficient

Observational Seismic Hazard

All the seismic intensity records due to active faults and inter-plate earthquakes, which are considered
separately from background zones in the present procedure, are excluded for observational computation of
seismic hazard. As shown in fig.12 the sites for the hazard computation are chosen so that they cover
almost all Japan. At the sites seismic intensity has been announced for no shorter than 50 years. Based on
the long-term seismic intensity records we compute probability that JMA seismic intensity x is larger than
or equal to intensity 4 during a year.
W akkanai City

Sapporo City

Urakawa Town

Sendai City
Iwaki City
Yonago City Matsumoto City
Matsue City Tokyo
Okayama City
Shimonoseki City
Omaezaki
Kushimoto Town

Kagoshima City

Fig.12 Sites Chosen for Hazard Evaluation

Note that JMA Seismic intensity has been measured automatically with seismic intensity meters since
1996, while former JMA seismic intensity had been determined from human response or observation of
damage. However, it is shown that former seismic intensity agrees well with instrumentally measured one
[25]. Therefore, we employ seismic intensity records before 1996 in addition to instrumentally measured
records for the computation of observational seismic hazard.

Understanding of Evaluated Seismic Hazard

Seismic hazards evaluated at various sites are compared with the observational ones in figs.11. Because
maximum magnitudes assumed for background zones in the hazard analysis are larger than the
magnitudes of earthquakes that occurred actually in limited periods of time, probabilities of exceedance
tend to be evaluated conservatively. According to figs.11, regional differences of exceedance probabilities
obtained from the hazard analysis based on past earthquake records agree roughly with the observationally
computed ones.

In the present procedure recurrence intervals, source locations, and magnitudes of the earthquakes that
occur repeatedly on active faults and subduction zones are evaluated based on the latest researches on
active faults and inter-plate earthquakes, and those characteristics of earthquakes are reflected in hazard
evaluation. The seismic hazard based on past earthquake records, whose regional differences agree with
the observations, is complemented by the hazard based on active faults and inter-plate earthquakes.
 1.00
0.6
O bservationalS eism ic H azard
0.90
Evaluated S eism ic H azard

Evaluated Probability of Exceedance

Tokyo
0.80 0.5
Exceedance Probability

0.70 Sendai City Urakawa Town

0.60 0.4
Wakkanai City Iwaki City
0.50 Matsumoto City
0.3
0.40 Sapporo City
Kagoshima City
0.30 0.2
Okayama City
0.20 Omaezaki
0.10 0.1 Matsue City
Kushimoto Town
0.00 Shimonoseki City
Yonago City
0.0
ue i

i ty
o
ty

sh City

aw City
i ty

or y

a g ty
i ty

k
m ty

Ur nda y
ya wn

wn

ky
pp Cit

t
i
i

Se Ci

iC
C
su i C

Yo o C

on o C

iC

0.0 0.1 0.2 0.3 0.4 0.5 0.6

o

To

To
ae

i
O oT
S a to

ak
na

ek

im
o

a
Ku ats

Iw
os
ka

ot
O
n

Obsevational Probability of Exceedance

im
M
ak

ka

ak
go
at

sh
m
W

Ka
M

Si

(a) Sorted by Observational Probability of Exceedance (b) Relationship Between Evaluated

Probability and Observational One
Fig.11 Probability that seismic intensity x is larger than or equal to Intensity 4 During a Year

CONCLUSION

The following conclusions are deduced from the present study.

1) We incorporate three kinds of earthquakes, i.e., earthquakes that occur randomly in both time and space
within background zones, earthquakes from active faults and inter-plate earthquakes, into seismic
hazard analysis. Assuming that each kind of earthquake occurs independently, a joint seismic hazard
due to three sources is estimated. Results are shown for PGA with 63% and 5% probabilities of
exceedance during 100 years from the year 2003.
2) According to the results active faults with high event rates and inter-plate earthquakes affect seismic
hazards nearby. Besides, strong ground motions are estimated widely around areas with high dense
distributions of active faults.
3)JMA seismic intensity records, which have been observed for long periods of time at various sites in
Japan, are employed for understanding of the evaluated seismic hazard. It is recognized that the seismic
hazard based on past earthquake records tend to be evaluated conservatively, and regional differences
of the seismic hazard agree roughly with the observational ones. In the present procedure the seismic
hazard based on past earthquake records is complemented by the hazard based on active faults and
inter-plate earthquakes.

ACKNOWLEDGMENTS

In order to develop the procedure for making seismic hazard map, “Technical Committee on Methodology
for Formulating Level 2 Seismic Design Motions (Chairman: Prof. Tatsuo Ohmachi, Tokyo Institute of
Technology)” had been established at the National Institute for Land and Infrastructure Management. The
authors would like to express their cordial thanks to all members of the committee.

REFERENCES

1. Arakawa, T. (1983). “Probabilistic Procedure in Selecting Input Motions for Dynamic Response
Analysis”, Technical Memorandum, No.1992, Public Works Research Institute, Ministry of
Construction (in Japanese)
2. Cornell, C. A.(1968). “Engineering Seismic Risk Analysis”, Bull. Seism. Soc. Am. Vol.58
3. Nakao, Y., Kusakabe, T., Murakoshi, J., and Tamura, K. (2003). “ Procedure for Producing
Probabilistic Seismic Hazard Maps”, Research Report of National Institute for Land and Infrastructure
Management, No.16 (in Japanese)
4. Usami, T. (1996). “ Materials for Comprehensive List of Destructive Earthquake in Japan”, University
of Tokyo Press (in Japanese)
5. Utsu, T. (1987). “ Encyclopedia of Earthquake ”, Asakura Publishing Co., Ltd. (in Japanese)
6. Japan Weather Association (1995), ”Catalog of Major Earthquakes Which Occurred in and around
Japan
7. Hagiwara, T. (1991),” Earthquake Engineering and Seismotectonics”, Kajima Institute Publishing Co.,
Ltd. (in Japanese)
8. Property and Casualty Insurance Rating Organization of Japan(2000). “Study on Seismic Hazard
Considering Active Faults and Historical Earthquakes”, Research Report for Earthquake Insurance
No.47 (in Japanese)
9. Annaka, T. et al (1997). “ Proposal of Attenuation Equation of Peak Ground Motion and Response
Spectrum Based on JMA-87 type Accelerogram”, Proc. of 24th JSCE Symposium on Earthquake
Engineering (in Japanese)
10. Matsuda, T. et al (2000) “ Distribution of M≧5 shallow earthquakes and active faults on Japanese
islands”, Active Fault Research, Vol. 20(in Japanese)
11. Research Group for Active Faults (1991), ” Active Faults in Japan, Sheet Maps and Inventories,
Revised Edition”, The Univ. of Tokyo Press (in Japanese)
12. Matsuda, T. (1975) “ Magnitude and Recurrence Interval of Earthquakes from a Fault, Jisin Journal,
Seism. Soc. Japan, Vol. 28
13. Headquarters for Earthquake Research Promotion, “Long Term Evaluation”,
http://www.jishin.go.jp/main/index.html
14. Science and Technology Agency (1995-1999), Proceedings of Active Faults Research (in Japanese)
Ministry of Education, Culture, Sports, Science and Technology (2001), Proceedings of Active Faults
Research (in Japanese)
15. Geological Survey of Japan (1998-1999), Interim Reports on Active Fault and Paleoearthquake
Researches in the 1998 and 1999 Fiscal Year
16. Takemura, M. (1990) “ Magnitude-seismic moment relations for the shallow earthquakes in and
around Japan”, Journal of the Seismological Society of Japan, Vol.43
17. Headquarters for Earthquake Research Promotion (2001), Seismic activity in Japan -- regional
perspectives on the characteristics of destructive earthquakes
18. Utsu, T. (1984).,“ Seismology”, Kyoritsu Shuppan Co., Ltd.
19. Sato, R. et al. (1989),“ Source Parameters for Earthquakes in Japan”, Kajima Publishing Co., Ltd.
20. Wald, D. J. and Somerville, P. G. : Variable Slip Rupture Model of the Great 1923 Kanto Earthquake,
Bull. Seism. Soc. Am. Vol.85, 1995
21. Central Disaster Prevention Council (2001), “ Paper in 7th Technical Committee on The Tokai
Earthquake”
22. Ishikawa, H. (2001), “ Producing Seismic Hazard Map and Problems”, 2nd Workshop on Earthquake
Research and Earthquake Disaster Prevention Engineering (in Japanese)
23. Japan Meteorological Agency (1996), ”Understanding of Seismic Intensity”, Gyosei Co., Ltd.
24. Shabestari, K. and Yamazaki, F.(1997): “Attenuation of JMA Intensity Based on JMA-87-type
Accelerometer Records”, Proceedings of the 2nd Symposium on the disaster by urban area
earthquakes
25. Midorikawa, S. et al(1999).,”Correlation of New J.M.A. Instrumental Seismic Intensity with Former
J.M.A. Seismic Intensity and Ground Motion Parameters”, Journal of Institute of Social Safety
Science, Vol.1