RISK COMPARISON OF NATURAL HAZARDS IN JAPAN

Transcription

1 4th International Conference on Earthquake Engineering Taipei, Taiwan October 12-13, 2006 Paper No. 248 RISK COMPARISON OF NATURAL HAZARDS IN JAPAN Tsuyoshi Takada 1 and Yoshito Horiuchi 2 ABSTRACT Japan has various natural hazards that have been great threat to building safety so far. The likelihood of failure of buildings highly depends both upon where the buildings are located and on what kind of natural hazard is focused, since recent hazard maps reveal significant locality throughout Japan. The current structural design code does not reflect the locality adequately since various design loads such as earthquakes, winds and snows have been determined based on the maximum values observed within Japan in the past. Therefore, not much attention has been paid on imbalance among various design loads in terms of their magnitudes. In this paper, major natural hazards in Japan such as earthquakes, winds and snowfalls are taken and firstly, direct hazard comparison is made on the basis of hazard maps recently developed by Architectural Institute of Japan (AIJ) to see the locality and uniformity among the natural hazards, and then the probability of failure of wooden individual houses due to each hazard is computed using the empirical fragility data. Then, economical consequence due to building failure is combined with the probability of failure to estimate risks from the specified hazard. Finally, the risks among different hazards are compared given a particular site in Japan and the risk due to a single hazard in different sites in Japan are also compared. Observed is less uniformity of risks between different hazards and significant locality within the nation. These risk comparison can provide features of this country and make important suggestion for future improvement of design as well as resource allocation for disaster mitigation. Keywords: Risk Comparison, earthquake, wind, snowfall, individual house risk, regional risk, design load. INTRODUCTION Japan has various natural hazards that have been great threat to building safety so far. The likelihood of structural failure of buildings highly depends both upon where the buildings are located and on what kind of natural hazard is focused, since recent hazard maps reveal significant locality throughout Japan. The current structural design code does not reflect the locality adequately since various design loads such as earthquakes, winds and snows have been determined based on the maximum values observed within Japan in the past. Therefore, not much attention has been paid on imbalance among various design loads in terms of their magnitudes. In this paper, major natural hazards in Japan such as earthquakes, winds and snowfalls are taken and firstly, direct hazard comparison is made on the basis of hazard maps recently developed by AIJ to see the locality and uniformity among the natural hazards, and then the probability of failure of wooden individual houses due to each hazard is computed using the empirical fragility data. Then, economical consequence due to building failure is combined with the probability of failure to estimate risks from the specified hazard. Finally, the risks among different hazards are compared given a particular site in Japan and the risk due to a single hazard in different sites in Japan are also compared. Observed is less uniformity of risks between different hazards and significant locality within the nation. These risk comparison can provide 1 Professor, Graduate School of Engineering, the University of Tokyo, Japan, takada@load.arch.t.u-tokyo.ac.jp 2 Graduate student, Graduate School of Engineering, the University of Tokyo, Japan, horiuchi@load.arch.t.u-tokyo.ac.jp

2 features of this country and make important suggestion for future improvement of design as well as resource allocation for disaster mitigations. Risk definition RISK ASSESSMENT PROCEDURE Risk is defined according to the current ISO document (ISO, 2002) as the combination of the probability of an event and its consequence. The event of interest in this paper is the damage of wooden houses due to the natural hazards since the events from the natural hazards spread throughout the country. The events of occurrence are specified in multiple states, which indicate the different damage states of houses. The damage states or ranks are often used in terms of Rank 1, 2 and 3 which respectively correspond to the state of totally collapsed, half collapsed and partially damaged. The consequence after the event occurs is in this study the monetary loss according to the damage of houses, which will be treated as the ratio of the damage to the initial value of the property. Therefore, the consequence is non-dimensional quantity in this paper. Figure 1 Seismic hazard map Figure 2 Wind hazard map Figure 3 Snowfall hazard map Then, the risk due to a particular hazard is computed as follows, R = P 1 C 1 + ( P 2 " P 1 )C 2 + ( P 3 " P 2 )C 3 = P 1 ( C 1 " C 2 ) + P 2 C 2 " C 3 ( ) + P 3 C 3 (1) where P i is the probability that the damage exceeds the specified rank i, C i is the consequence term, that is, the monetary loss ratio suffered at the specified rank i. For the damages of houses due to all hazards, it is assumed that the monetary loss ratio is commonly 50% for the state of totally collapsed, 20% for half collapsed and 10% for partially damaged, all of which are taken from the damage survey in the 1995 Kobe earthquake. These assumptions should be verified in a future. Hazard estimation (a) Seismic hazard The probability of the specified event can be estimated by combining the hazard and structural fragility curves. For the estimation of different natural hazards, AIJ recently compiled the hazard maps for earthquakes, winds and snowfalls to provide a common statistical basis for determination of design loads (AIJ, 2004; AIJ, 1993). These maps can be made the best use of for the purpose of this study. The intensity of the earthquake ground motion throughout Japan is represented in terms of PGA at the engineering bedrock, as is plotted in Fig. 1, in which the PGA corresponding to the 100-year return period is drawn in a contour map. This map clearly indicates that there is non-uniformity of earthquake ground motions in Japan. The value of PGA at the engineering bedrock can be

3 transformed into that of ground motions at the ground surface where all houses are located, by multiplying a wave amplification factor specified for soil (Midorikawa et al., 1994). (b) Wind hazard As has been done for the case of the seismic hazard estimation, the wind hazard is estimated, based on the AIJ recommendations. AIJ has provides the hazard map for the 10 min-mean wind velocity at the height of 10 meters above the ground, where the ground roughness is categorized into II. The contour map plots the 100-year mean wind velocity, as is shown in Fig. 2, as well as the 500-year value which is not shown in this paper, both of which can be used to determine the Gumbel type distribution of the mean wind velocity. The load effect from the wind hazard map can be computed using the wind direction factor, the factor associated with the vertical distribution and the gust factor. (c) Snowfall hazard The snowfall is based on the annual maximum snow depth data which has been recorded at 136 points and 194 AMeDas observatories in Japan. The statistics of them are modeled as the Gumbel type distribution. AIJ recommendations provide the hazard map for the snowfall as is shown in Fig. 3. Fragility estimation (a) Seismic fragility The seismic fragility is expressed in terms of PGV for each specified damage rank. The fragility curve that is understood as the conditional probability under a given PGV can be written as (Murao et al., 2000) & ( ) = " lnv( T) # $ # lnz Ei ( P Ei T ' % Ei ) + (2) * where P Ei (T) is the conditional probability that the damage of wooden houses exceeds the specified damage rank i under the given PGV v (T) which is a function of the return period T. λ Ei and ζ Ei are, respectively, the logarithmic mean and standard deviation of the fragility of wooden houses. Z denotes the zoning factor that has been used for design in Japan, and unfortunately does not reflect the probabilistic hazard results, and ranges from 0.7 to 1.0 in Japan. Φ is the standard normal cumulative distribution function. The subscript E implies that all subscripted variables are related to earthquakes. Figure 4 shows some fragility curves that have often been modeled as a long-normal distribution with the lognormal standard deviation ζ Ei, say, 0.5. Table 1 lists the statistical parameters determining the distributions. In this table, the damage rank 3 are not written since the Kobe city municipal office did not provide the date for this rank. Table 1 Seismic fragility Damage State Rank λ E ζ E Totally collapsed Half collapsed Partially collapsed Figure 4 Seismic fragility curves (Rank 1)

4 (b) Wind fragility Referring to the literature (Kondo et al, 2002), the fragility curve is modeled as a long-normal distribution where statistical parameters are given in Table 2. & ( ) = " lnw ( T) # $ Wi ( P Wi T ' % Wi ) + (2) * where W(T) is the 10 minutes-mean wind speed corresponding to the return period T, and λ Wi and ζ Wi are parameters determining the distributions. Table 2 Wind fragility Damage State Rank λ Wi ζ Wi Totally collapsed Half collapsed Partially collapsed Table 3 Snowfall fragility Damage State Rank λ Si ζ Si Totally collapsed Half collapsed Partially collapsed It is assumed in this paper that for the above damage rank, the following monetary loss ratio are respectively C W1 =50%, C W2 =20% and C W3 =10%. (c) Snowfall fragility Not much research for estimating house fragility against the snowfall has been done so far. The fragility curve to be used in the following risk estimation is based on the damage data of wooden houses, which have been compiled in the Niigata prefecture during the year of 1980 (NIED, 1982), when the heavy snowfall were experienced. Again, the distribution is assumed to be a lognormal distribution as follows. & ( ) = " ln d( T) # $ Si ( P Si T ' % Si ) + (3) * where d(t) is the snow depth in centimeter, and λ Si and ζ Si are parameters determining the distributions, which are listed on Table 3. Estimation of risk curve The probability of the events thus specified for each hazard is now combined with the consequence, that is, the monetary loss ratio in order to estimate the risk due to the specified hazard. Using Eq. (1), the risk is the function of the selected return period for each hazard T and 500-year return periods are taken as a reference to compare these risks. Once the risk can be expressed in terms of the specified return period, the probability that the risk will occur is equal to the reciprocal of the return period. Then, the relation between the occurrence probability and the intensity of the risk is called a risk curve, which implies how frequently the specified intensity of the consequence occurs. Letting R E (T), R W (T) and R S (T) be earthquake, wind and snow risks to given hazard levels, respectively, the total risk R total can be defined as follows, R total ( T) = R E T ( ) + R W ( T) + R S ( T) (4) Contribution of each risk to the total risk can be a useful index for future judgment about disaster mitigation policy. RISK COMPARISION

5 The risks computed here are twofold. One is called an individual house risk, the other is regional risk. The individual house risk can be translated in such a way that when one wants to see for which risk one has to care the most, he may need an essential basis to make decision for his individual house. In other words, if someone would like to build his house somewhere in Japan, he may need a rational basis to decide where to live. In this particular purpose, the individual house risk may be a useful indicator as well as good information for selecting a place to live in. For the computation of the risk, a single house, in general, is of the utmost concern, and the numbers of houses are not taken into account. On the other hand, the major concern of the central government, for instance, may be the regional risk. Their major concern is on which city or region they must spend their budget most to mitigate the risk. Namely, the regional risk is closely related to more effective regional allocation of the national resource. In the following, the discussion will be made based on the above two types of risks. Comparison of individual house risk Figure 5 Location of selected cities Figure 7 Normalized risk of selected cities The individual house risks of a single house located in some selected major cities in Japan are listed on Table 4. The selected cities are plotted in Fig. 5. The risks are computed for the load intensities that correspond to 100- and 500-year return periods. Figure 6 makes comparison of the individual house risks at major cities. From these results, it can be clearly observed that 1) the individual house risk due to the snowfall is negligible everywhere, 2) the earthquake risk at major cities varies quite significantly, among them, the Shizuoka city shows the highest earthquake risk, and 3) the wind risk to a single house is almost the same disregard to location of cities. The Shizuoka city is the place where a large inter-plate earthquake is considered to occur in a near Table 4 Individual house risks in selected cities 100-year return period 500-year return period Earthquake Wind Snowfall Earthquake Wind Snowfall Sapporo Sendai Tokyo Fukui Shizuoka Osaka Hiroshima Kochi Fukuoka Kagoshima

6 future. The earthquake is named the Tokai earthquake, and the government has reported the occurrence probability of the next event is more than 90% within the next 30 years (Fujiwara et al., 2004). If the above results are normalized by the total risk (Fig. 7), the results can clearly suggest which types of risk must be prioritized. This may lead to optimum resource allocation within each city. Figure 6 Comparison of individual house risk in selected cities Comparison of regional risk It would be very effective for the regional policy-making of disaster mitigation that the risks are compared region by region. Comparison of the total risk of each selected region can give us a useful and rational insight on prioritization for resource allocation. When the regional risk is estimated, the numbers of houses in the region are taken into consideration. The numbers of houses are based on the recent survey data (SBMUAF, 2003). If the property value of a single house is assumed to be 20 million Japanese yen, the regional risk can be expressed in the monetary unit. Table 5 shows the regional risk at each major city for 100- and 500- year return period value. Figure 8 compares the histogram of the regional risks ate the cities. It can be seen from them that, 1) The highest regional risk can be observed at Tokyo, and the second is at Osaka, 2) Kochi and Kagoshima, both of which indicate relatively higher individual risks (see, Fig. 6), reveal less regional risks due to a small number of houses in the two rural cities. Note that figures in Table 5 is the damage of only houses in the cities, and more damage have to be summed up to grasp the overall monetary loss of the big cities. From the above observations, it can be concluded that big cities such as Tokyo and Osaka have highest regional risk due primarily to the numbers of houses there. In Tokyo, a half of the risk comes from earthquakes and another half from winds. Although the obtained results are based on several assumptions, practical treatments and rough statistics, concentration of population and houses on a major city makes the situation worse from the viewpoint of the regional risk in Japan. The snow risk is negligible throughout Japan. The snow prone-regions are limited to the northwestern part of Japan where there are less houses. RISK MAP The hazard maps were shown in chapter 2. Although these maps imply how frequent and severe the intensity of the natural hazards are within Japan, so called risk maps that are the final product

7 through combining the hazard maps with the fragility information of houses can be drawn. In this case, the individual house risks rather than the regional risk are plotted for each natural hazard. Figures 9, 10 and 11 demonstrate the risk map for earthquakes, winds and snowfalls for 500-year return period. Table 5 Regional risks in selected cities (billion yen) 100-year return period 500-year return period Earthquake Wind Snowfall Earthquake Wind Snowfall Sapporo Sendai Tokyo Fukui Shizuoka Osaka Hiroshima Kochi Fukuoka Kagoshima Figure 8 Comparison of regional risks of selected cities CONCLUSIONS Comparison of risks due to natural hazards in Japan is made in this paper. Hazard maps provided by AIJ are fully utilized, and then are transformed into the risk maps associated with the fragility information of wooden houses. Two types of risks; an individual house risk and a regional risk, are proposed. The former risk representation can be used for judging which hazard has the most contribution to the total risk, while the latter representation can be used for making the natural disaster mitigation policy in a national scale. The comparison of risks due to earthquakes, winds and snowfalls in Japan is implemented in this paper. The following important features of this country are found, For the individual house risk, 1) earthquake risk shows significant locality in different locations, 2) wind risk shows almost uniformity throughout the country, 3) snow risk is almost negligible, For the regional risk,

8 4) big cities such as Tokyo and Osaka have highest risk due to densely populated area. Future possible research works will be improvement of each risk assessment procedure, especially estimation of consequence, and consideration of how these risks are combined, and applications of the procedure to total risk management such as insurance. Figure 9 Earthquake risk map Figure 10 Wind risk map Figure 11 Snow risk map ACKNOWLEDGEMENTS The authors acknowledge experts who kindly provided digital data of AIJ hazard maps. This study has been partially supported by the project named Establishment and application of probabilistic assessment method of mutually correlated fragility of urban city against earthquakes sponsored by Ministry of Education, Culture, Sports, Science and Technology (Project No: ). REFERENCES AIJ, 1993, AIJ recommendations for loads on buildings, Maruzen publishing company (in Japanese) AIJ, 2004, AIJ recommendations for loads on buildings, English version available on the web Fujiwara, F., et al., 2004, Outline of strong-motion evaluation in national hazard mapping project of Japan, J. of JAEE, Vol. 4, No.3 ISO/IEC, 2002, Guide 73- Risk management-vocabulary Guidance for use in standards Kondo, H., et al., 2002, Study on wind hazard assessment, Proc. of 17 th Japan symposium on wind engineering, pp (in Japanese) National Research Institute for Earth Science and Disaster Prevention (NIED), 1982, Report on damage of Hokuriku region due to the 56 Heavy snowfall, report of major disaster, No. 17 (in Japanese) Midorikawa, S, et al., 1994, Site effects on strong-motion records observed during the 1987 Chibaken-toho-oki, Japan earthquake, Proc. of 9 th Japan Earthquake Engineering Symposium, Vol. 3, pp Murao, O. and Yamazaki, F., 2000, Fragility functions of houses based on municipal investigation data of the 1995 Kobe earthquake, AIJ, No. 527, pp (in Japanese) Statistics Bureau of Ministry of International Affairs and Communications, 2003, Statistical Survey of Houses and Lands, ( (in Japanese)