=
( 1)
where V is the input earthquake ground motion on surface in velocity (cm/s), is
the normal distribution function, and i represents the damage state. These values are
obtained after Hyogoken-Nambu earthquake for houses with grade 1, which is the basic
performance grade as specified by the Law for Quality Assurance of Performance of
Houses. The repair cost in Table 1 i s obtained al so from the Hyogoken-Numbu
Earthquake as surveyed by Kanda and Hirakawa, 1998 .
Seismic performance for newly built houses in J apan can be categorized into three
grades as the regulation specifies. Grade 1 is the ordinary level satisfying the criteria
specified in the seismic regulation according to the Building Standard Law, Grade 2 is
the level where 1.25 times seismic load of the regulation is applied and Grade 3 is the
level where 1.5 times is applied.
Table 1 Parameters for log-normal fragility curve
Damage state i
Repair cost (%)
Slight ln50 0.4 -
Minor ln100 0.4 10
Moderate ln150 0.4 15
Severe ln200 0.4 30
Collapse ln250 0.4 100
According to the conventional contract of the earthquake insurance, the insurance is
paid when the following criteria are met,
I f the monetary loss of elements of the relevant structure falls between 3 % and
20 %, or the house is affected by flood above floor level, 5 % of insured value is paid.
(level 1)
I f the monetary loss of elements of the relevant structure falls between 20 % and
50 %, or 20 % to 70% of total floor is lost due to fire, tsunami, or flood, 50 % of
insured value is paid. (level 2)
I f the monetary loss of elements of the relevant structure exceeds 50 %, or 70 % or
more of total floor is lost due to fire, tsunami, or flood, 100 % of insured value is
paid. (level 3)
Compared with the loss due to the direct damage from earthquake motion, the
probability that the house is damaged by fire, tsunami, or flood is so much smaller that
only damage due directly to the earthquake motion is considered hereafter.
I n order to estimate some statistics of earthquake insurance, we need to relate the
criteria of earthquake insurance to damage state of structure in Table 1. I t seems
rational to relate as level 1 to the slight and minor damage, level 2 to the severe and
moderate damage, level 3 to the collapse. As a consequence, the insured loss,
i
L , where
i indicates the damage level, is obtained as the sum of the following,
1
ln ln50 ln ln150
0.05
0.4 0.4
V V
L
=
( 2a)
2
ln ln150 ln ln250
0.5
0.4 0.4
V V
L
=
( 2b)
3
ln ln250
1
0.4
V
L
=
( 2c)
Figure 1 shows the diagram of damage ratio for each stage of loss level. These loss
functions can be interpreted as the ones for portfolio. The original fragility curve
represents the average loss ratio so that the local loss ratios may differ from average
loss ratio. However, when portfolio loss is considered, the loss ratio can be rationally
assumed to converge to average loss ratio by the law of large number. The portfolio
loss can be used for insurance premium calculation. From this reason, thereafter we
adopt the above loss functions as the ones for portfolio in a deterministic manner.
SEI SMI C HAZARD ESTI MATI ON
For illustrative purposes, 8 major cities in J apan are selected, i.e. Sapporo, Sendai,
Tokyo, Nagoya, Osaka, Hiroshima, Takamatsu, and Kitakyushu as shown in Figure 2.
Seismic hazard analysis is conducted utilizing data and procedures proposed by Ochi et
al 2002. Fi gure 3 shows earthquake hazard at each city. Major acti ve faul ts are
considered with respective occurrence probability according to a report provided by the
Ministry of Education, Culture, Sports, Science and Technology, 2001 and the
Gutenberg-Richer relationship for the earthquake occurrence is adopted where active
fault information is not available based on the past earthquake statistics. Soil
amplification factor at each site is assumed to be 2.0, which is a reference value for a
relatively firm soil condition.
Hazard values at Osaka, Nagoya and Takamatsu are relatively high as shown in
Figure 2 as the consequence of To-kai and To-nankai Earthquakes with their fairly high
occurrence probabilities. The uncertainty of attenuation formula used in the study is
assumed as 0.5 in terms of the logarithmic standard deviation. Although this value is
consistent to the statistics available for many earthquakes, it could be considered to be
extremely high when the fault source model can be estimated and soil characteristics
and path information are provided. The combination of high occurrence probability
and high uncertainty of attenuation formula cause significantly high hazard results and
could lead to the conservative loss estimation. Thi s conservatism may be eliminated
by applying scenario earthquakes to probabilistic hazard estimation such as proposed
by Kanda et al, 2004.
I NSURANCE PREMI UM AND EXPECTED LOSS
Table 2 shows basic insurance premium for each site for a house of reinforced concrete
or steel structure. Even higher premium is determined f or wooden houses. The
seismic activity is considered based on the hazard evaluation and premium is
differentiated for sites as listed in Table 2.
Table 2 Annual basic premium for RC or S
Sapporo Sendai Tokyo Nagoya Osaka Hiroshima Takamatsu Kitakyushu
Premium
(%)
0.05 0.07 0.175 0.135 0.135 0.05 0.05 0.05
Figure 4 shows the comparison between the present insurance premiums at each site
and the mean insured loss estimated from engineering calculation. I f a house is
categorized to grade 1, 2 or 3, the insurance premium is reduced by 10 %, 20 %, 30 %
respectively. The above mentioned fragility curve corresponds to the one for grade 1.
From the engineering point of view it can be assumed that the resistance of the house
with grade 2, 3 is stronger in terms of seismic load considered by 25 %, 50 % than the
above mentioned fragility curve. Therefore the fragility curves for houses with grade 2,
3 are obtained by shifting the original fragility curves accordingly and so are insured
loss functions.
Table 3 shows the ratio of present premium, for each site and each grade, to expected
insured loss obtained by hazard anal yses and l oss functi ons. Resul ts of Tabl e 3
indicates that present insurance premium in Table 2 does not fully reflect the local
earthquake hazard, and what is more, the reduction of premiums by 20% or 30% for
grade 2 or 3 respectively does not compensate at all for strengthened resistance of
house.
Table 3 (insurance premium at present) / (expected insured loss)
Sapporo Sendai Tokyo Nagoya Osaka Hiroshima Takamatsu Kitakyushu
grade1 466.6 13.7 9.0 1.5 1.9 6.3 1.8 17.5
grade2 999.0 25.6 15.7 2.8 3.7 13.5 3.8 38.5
grade3 2054.4 42.0 25.4 4.0 5.0 16.5 4.7 44.8
Probable maximum loss (PML) curves for are shown in Figure 5. 90 percentile loss
estimation was used as i n the conventi onal PML estimation. 475 year is often
considered for PML as the return period for the maximum earthquake ground motion,
but the return period is treated as the variable in the figure to examine the tendency of
the return period of input motion on the probabilistic loss estimation. As the tendency
of the seismic activity differs from site to site, PML varies with the return period rather
differently.
I n sites where the slope shallow as in Tokyo or Nagoya in Figure 5, damages are
expected for relatively high frequency events which corresponds to the lower level
damages. Then this tendency can also pointed out in Figure 4, where the expected
level 1 damage is relatively high in Tokyo or Nagoya.
CONCLUSI ONS
Sei smi c i nsurance premi um for houses i s consi dered to be too hi gh i n J apan.
Although insurance companies explain that the premium is determined according to the
past earthquake damage statistics, the strength of present houses is not considered
sufficiently for the premium determination. The present study describes numerical
examples by showing the present premium and the expected loss, then the
inconsistency of the local hazard and also the individual strength of houses are
discussed.
ACKNOWLEDGEMENT
This project is partly supported by the grant from the J SPS for young scientists.
REFERENCES
Kanda, J . and Hirakawa, T. Failure cost estimation equivalent to multi-level limit state
failure, Structural Safety and Reliability, Balkema, pp295-300, 1998.
Kanda, J . et al., Probabilistic design earthquake ground motions considering scenario
earthquakes, Proc. 13
th
World Conf. Earthquake Eng., Vancouver, Paper No. 2120,
2004.
Ministry of Education, Culture, Sports, Science and Technology, Report on Survey of
Active Faults, 1997-2001. (in J apanese)
Murao, O. and Yamazaki, F. Development of fragility curves for buildings based on
damage survey data of a local government after the 1995 Hyogoiken-Numbu
Earthquake, J . Struct. Constr. Eng., A.I .J . No.527, pp189-196, 2000. (in J apanese)
Ochi, S., Sakamoto, S., Takada, T. and Kanda, J . An internet-based system for seismic
performance evaluation of existing buildings, 1
st
ASRANet I nternational Colloquium,
Glasgow, 2002.
Fi gure 1 Diagram of damage ratio for each level
Figure 2 Locations of 8 sites in J apan
Figure 3 Hazard curves at 8 sites
Figure 4 Comparison between present insurance premium and expected insured loss
from our analysis at each site
Figure 5 PML curve at each site