tsunami fragility – a new measure to identify tsunami damage · 2012-01-25 · tsunami fragility...

Tsunami Fragility

Paper:

Tsunami Fragility– A New Measure to Identify Tsunami Damage –Shunichi Koshimura∗, Yuichi Namegaya∗∗, and Hideaki Yanagisawa∗∗∗

∗Graduate School of Engineering, Tohoku UniversityAoba 6-6-11-1104, Aramaki, Aoba-ku, Sendai 980-8579, Japan

E-mail: [email protected]∗∗Geological Survey of Japan, National Institute of Advanced Industrial Science and Technology

C7, 1-1-1 Higashi, Tsukuba 305-8567, JapanE-mail: [email protected]

∗∗∗Tokyo Electric Power Services Co., Ltd.3-3, Higashiueno 3-Chome, Taito-ku, Tokyo 110-0015, Japan

E-mail: [email protected][Received June 29, 2009; accepted November 5, 2009]

Abstract. Tsunami fragility (fragility curve, or fragilityfunction) is a new measure, we propose, for estimat-ing structural damage and fatalities due to tsunami at-tack, by integrating satellite remote sensing, field sur-vey, numerical modeling, and historical data analysiswith geographic information system (GIS). Tsunamifragility is expressed as the structural damage prob-ability or fatality ratio related to hydrodynamic fea-tures of tsunami inundation flow, such as inundationdepth, current velocity and hydrodynamic force. Itexpands the capability of estimating potential tsunamidamage in a quantitative manner.

Keywords: fragility curve, tsunami damage estimation,remote sensing, numerical modeling, historical tsunamis

1. Introduction

In tsunami damage estimation efforts, several empiri-cal relationships between tsunami hazard and vulnerabil-ity have been used. Shuto (1993) proposed the tsunamiintensity scale to discuss the structural damage based onthe empirical data from historical tsunamis in Japan, interms of the damage and local tsunami height [1], and thishas been widely used in tsunami disaster assessment bythe Japanese government as a measure of tsunami dam-age. When the local tsunami inundation depth exceeds2 m, for example, Shuto’s tsunami intensity scale sug-gests, complete destruction of wooden houses as shownin Fig. 1. Izuka and Matsutomi (2000) suggested a struc-tural destruction threshold related to the hydrodynamicfeatures of tsunami inundation flow throughout the fieldsurveys and laboratory experiments [2]. And some engi-neering studies have proposed tsunami design forces onstructures based on the laboratory experiments [3, 4], buthave not suggested the procedures for estimating struc-tural damage.

Since the 2004 Sumatra-Andaman earthquake tsunamidisaster, numerous efforts have been made to identify thetsunami damage mechanisms by widely deployed post-tsunami survey teams reporting tsunami height, inunda-tion zone extent and damage [5–9]. These efforts have ledto new understandings of local aspects of tsunami inunda-tion flow and damage mechanisms.

However, their findings based on the inspection of lo-cal aspects of tsunami damage make it difficult to identifythe “vulnerability” in a quantitative manner. The natureof vulnerability is associated with multitude of uncertainsources, such as hydrodynamic features of tsunami in-undation flow, structural characteristics, population, landuse, and any other site conditions. To view tsunami vul-nerability comprehensively requires humongous amountsof damage data, whereas post-tsunami survey rarely pro-vides sufficient data because of limited survey time andhuman resources. As Shuto (1993) concluded, degree ofdamage may change with these uncertainties and a statis-tical approach to these uncertainties should be taken.

Herein, we propose Tsunami fragility (fragility curveor fragility function) as a new measure for estimatingtsunami damage. Tsunami fragility is defined as the struc-tural damage probability or fatality ratio with particularregard to the hydrodynamic features of tsunami inunda-tion flow, such as inundation depth, current velocity andhydrodynamic force. In principle, the development oftsunami fragility requires that tsunami hazard informa-tion and damage data should be used synergistically. Wethus incorporate several approaches to constructing thetsunami fragility.

In order to obtain tsunami hazard information such asinundation depth and current velocity, we performed a nu-merical modeling of tsunami inundation with some modelvalidations, especially focusing on the 2004 Sumatra-Andaman earthquake tsunami. In terms of the damagedata including structural damage and fatalities, we use therecent advances of remote sensing technologies expand-ing capabilities of detecting spatial extent of tsunami af-

Journal of Disaster Research Vol.4 No.6, 2009 479

Koshimura, S., Namegaya, Y., and Yanagisawa, H.

Tsunami intensity 0 1 2 3 4 5

Tsunami height (m) 1 2 4 8 16 32

Damage : wooden house partly dam-aged

Damage : masonry house no data

Damage : reinforced concretebuilding

Damage : fishing boats damage

withstand completelydestroyedno data

completely damaged50 damage

completely destroyed

completely destroyedwithstand

Fig. 1. Tsunami intensity scale and damage [1], modified from the original figure.

fected areas and damage on structures. In addition, somefield data and historical documents are used to obtain thetsunami fragility from historical events.

Throughout the data integration and statistical analysis,we propose a concept and framework developing tsunamifragility, and apply it in several approaches according toavailable data types from recent and historical events.

2. Developing Tsunami Fragility – Methods

Fragility curves (or fragility functions) have conven-tionally been developed in performing seismic risk analy-sis of structural systems to identify structural vulnerabilityagainst strong ground motion using damage data associ-ated with historical earthquakes and spatial distribution ofobserved or simulated seismic responses [10]. And theyhave been implemented in estimating structural damageagainst potential seismic risks in which various uncertainsources such as seismic hazard, structural charachteris-tics, soil-structure interaction are involved [11, 12].

In earthquake engineering studies, the fragility curvesare defined by the following Eq. (1) or (2) ;

PD(x) = Φ[

lnx−λξ

]

=∫ x

−∞

1√2πξ t

exp

(− (ln t −λ )2

2ξ 2

)dt (1)

PD(x) = Φ[

x−µσ

]

=∫ x

−∞

1√2πσ

exp

(− (t −µ)2

2σ2

)dt . (2)

where PD(x) is the damage probability of structures, as afunction of x or lnx such as the maximum ground accel-eration, velocity, and seismic intensity. Here, PD(x) is ex-pressed by the standard lognormal or normal cumulativedistribution function Φ[(lnx − λ )/ξ ] or Φ[(x − µ)/σ ],with two statistical parameters (λ ,ξ ) or (µ,σ), as themean of lnx (or x), and the standard deviation respec-tively.

To develop tsunami fragility, we take a statistical ap-proach synergistically using remote sensing, numericalmodel results, field surveys and historical documents, infive steps.

1 Damage data acquisition : obtaining damage datafrom satellite images, field surveys or historical doc-uments, e.g. numbers of destroyed or survived struc-tures with its spatial information.

2 Tsunami hazard estimation : estimating the hydro-dynamic features of tsunami by numerical modeling,field surveys and from historical documents.

3 Data assimilation between the damage data andtsunami hazard information : correlating the dam-age data and the hydrodynamic features of tsunamithrough the GIS analysis.

4 Calculating damage probability : determining thedamage probabilities by counting the number ofdamaged or survived structures, for each range of hy-drodynamic features above.

5 Regression analysis : developing the fragility curvesby regression analysis of discrete sets of damageprobability and hydrodynamic features of tsunami.

In the sections that follow, we apply the above proce-dure in several approaches to construct tsunami fragility,according to the data types such as satellite data, numeri-cal models, field surveys and historical documents.

3. Tsunami Fragility Determined from Satel-lite Remote Sensing and Numerical Model-ing

Taking an advantage in satellite remote sensing, weidentify the spatial distribution of structural damage bytsunami. The highest spatial resolution of commercial op-tical satellite imaging is up to 60-70 centimeters (Quick-Bird owned by DigitalGlobe) or 1 meter (IKONOS op-erated by GeoEye). Fig. 2 shows the result of visualinterpretation [13] on structural damage by using a setof pre and post-tsunami (the 2004 tsunami) satellite im-ages (IKONOS) from Banda Aceh, Indonesia. Throughinspecting a set of pre and post-tsunami satellite imagesvisually or manually, presence of building roofs can beinterpreted. The advantage of using high-resolution opti-cal satellite images for damage interpretation is the capa-bility of understanding structural damage visually and en-ables us to comprehend its spatial extent in regional scalewhere post-tsunami survey hardly get through because of

480 Journal of Disaster Research Vol.4 No.6, 2009

Tsunami Fragility

� � � � � ��

��

�� !�"��

#�$�$� ��!�

��

� �

� ��

� �

� �

%��

�&��'��&�

�&��'��&�

�&��'��&�

�&��'��&�

�&��'��&�

�&��'�!&�!&��'�"&�

"&��'��&��&��'��&�

��

��

�� #�$�$�

%�� (��

Fig. 2. (a) Spatial distribution of structural damage interpreted from the pre and post-tsunami satellite images (IKONOS) [13].Black dots indicate the interpreted structures as destroyed, and the gray dots as survived. The arrow points the expanded regionshown in the right panels (b) pre-tsunami and (c) post-tsunami satellite images and (d) interpreted damage. (e) The maximumtsunami inundation depth obtained from the numerical model [14, 15].

the limitation of survey time and resources. However,note that no structural types were identified by the inter-pretation of satellite images. Also, the damage featurewhich can be identified from satellite images is only struc-tural destruction or major structural failure which revealschange of roof’s shape, namely “collapsed” and “majoror severe damage.” Accordingly, the interpretation “De-stroyed” in Fig. 2 means “collapsed” or “major or severedamage,” and “Survived” is either of “moderate,” “minor,”“slight” and “no” damage.

To obtain the tsunami hazard information such as in-undation depth and current velocity, we performed a nu-merical modeling of the 2004 Sumatra-Andaman earth-quake tsunami [14, 15], using high-resolution bathymetryand topography data. The model results were validatedby the field measurements of post-tsunami survey teams[6, 7], using Aida’s formula [16] in terms of the reliabilityof tsunami numerical model (see ref.[15] for detail).

Damage interpretation shown in Fig. 2(a) is combinedwith the numerical model results, e.g. Fig. 2(e), to obtainthe tsunami damage statistics as shown in Fig. 3. UsingGIS, we sampled all of the structures in the tsunami inun-dation zone and made a table (spread sheet) of structureID, damage interpretation (Destroyed or Survived) andtsunami hydrodynamic features (inundation depth, cur-rent velocity and hydrodynamic force) spatially equiva-

0

200

400

600

800

1000

1200

0.2

0.4

0.6

0.8

1.0

1.2

1.3

1.5

1.7

1.9

2.1

2.2

2.4

2.6

2.8

2.9

3.1

3.3

3.5

3.7

3.9

4.0

4.2

4.3

4.5

4.7

4.8

5.0

5.1

5.3

5.6

6.7

8.2

Inundation depth (< x m)

DestroyedSurvived

Number of

Fig. 3. Histogram of the numbers of destroyed and survivedstructures in terms of inundation depth range. Each inunda-tion depth range is determined by exploring a range whichincludes approximately 1,000 structures.

lent to the position of each structure. After sorting thetable by each level of hydrodynamic features, we deter-mined the groups of structures to calculate damage proba-bility so that roughly 1000 structures are involved in eachgroup. Then we determined the damage probability ineach group according to the range of inundation depth,current velocity and hydrodynamic force obtained by the



��

��

��

��

��

��

��

��

Fig. 4. A discrete set of damage probabilities and the me-dian values of inundation depths that were compiled fromsample data.

�

�

�

�

�

�

�

��

�

��

��

��

Fig. 5. An example of the plot on normal probability paper.

numerical model. And as a result of counting the numberof destroyed and survived structures within each inunda-tion depth range (group), we obtain a relationship betweenthe damage probability and inundation depth, as a discreteset of structural damage probabilities and tsunami inunda-tion depths shown in Fig. 4. Then, we explore this rela-tionship with the form of fragility curve by performing theregression analysis.

Taking an analogy of earthquake engineering studies[10, 11, 17], we assume that the cumulative probability PDof damage occurrence is given as either Eq. (1) or (2)with two statistical parameters, (λ ,ξ ) or (µ,σ ). Here,the statistical parameters λ (or µ) and ξ (or σ ) are ob-tained by plotting x (or lnx) and the inverse of Φ (Φ−1)on normal or lognormal probability paper, and conduct-ing the least-squares fitting of this plot, as shown in Fig. 5.Hence, these parameters are obtained by taking the inter-cept (= λ or µ) and the angular coefficient (= ξ or σ) inEq. (1) or (2).

Through regression analysis, the parameters are de-termined as shown in Table 1, to obtain the best fit offragility curve with respect to the maximum inundationdepth (measured above the local ground level) dmax (m),the maximum current velocity vmax (m/s) and the maxi-

Table 1. Statistical parameters of tsunami fragility curves(Fig. 6) for structural damage. R2 is the coefficient of deter-mination obtained through the least-squares fitting.

Fragility curve µ σ λ ξ R2

(a) dmax (m) 2.99 1.12 N/A N/A 0.99(b) vmax (m/s) N/A N/A 0.80 0.28 0.97(c) F (kN/m) N/A N/A 1.47 0.75 0.99

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

� � ��

��

��

Fig. 6. Tsunami fragility curves for structural destruction,in terms of (a) the maximum inundation depth, (b) the max-imum current velocity and (c) the maximum hydrodynamicforce obtained from the numerical model. The solid linesare the best-fitted curves of the plot (◦ : the distribution ofdamage probabilities) with the parameters in Table 1.

mum hydrodynamic force F acting on a structure per unitwidth (kN/m). Here F is defined as the maximum dragforce per unit width of structures;

F =12

CDρ max{v2d}×10−3 . . . . . . . (3)

where CD is the drag coefficient (CD = 1.0 for simplic-ity), ρ water density (= 1,000 kg/m3), v current velocity(m/s) and d inundation depth (m), and both of v and d areobtained at each time step of the tsunami inundation mod-eling. Note that the tsunami fragility with respect to theinundation depth is given by the standardized normal dis-tribution function with µ and σ , while those to the currentvelocity and hydrodynamic force are by the standardizedlognormal distribution functions with λ and ξ . The selec-tion of which curve is applied should be made by checkingits fit to the datasets.

Fragility curves shown in Fig. 6 indicate the damageprobabilities of structural destruction equivalent to the hy-drodynamic features of tsunami inundation flow. Housesin Banda Aceh, for example, were especially vulnera-


Tsunami Fragility

ble when the local inundation depth exceeded 2 or 3 m,the current velocity exceeded 2.5 m/s or hydrodynamicload exceeded 5 kN/m. Note that the observed struc-tural damage at a site might consist of both damage bytsunami and strong ground motion. Major structure typesin the tsunami-affected area were low-rise wooden house,timber construction, and non-engineered RC constructionlightly reinforced, and it was reported that the large num-ber of the wooden houses survived the earthquake withminor damage and non-engineered RC structures weredegraded by strong ground motion, then they were de-stroyed by the tsunami [18]. We supposed that the struc-tural destruction was likely to be induced by the tsunamiinundation, but many of the structures were degraded bystrong ground motion before the tsunami attack. In thissense, the proposed tsunami fragility may involve thestructures with minor damage or degraded seismic per-formance. Note also that the tsunami damage on struc-tures were caused by both hydrodynamic force/impactand the impact of floating debris, i.e. these facts are re-flected on the damage probabilities but not on the nu-merical model results (the estimated hydrodynamic fea-tures). Thus, the present tsunami fragility may indicateoverestimation in damage probabilities to the hydrody-namic features of tsunami inundation flow. Further to bementioned is that tsunami fragility proposed herein is forassessing the number of damaged (destroyed) structuresby applying tsunami fragility curves to the number of ex-posed structures against a given hydrodynamic conditionof tsunami. This is not for a prediction whether a struc-ture is destroyed or survives under a given probability ofoccurrence.

The proposed tsunami fragility curves here was basedon the regression analysis of the relations between themodeled tsunami hazards and the damage probabilitiesthat were sampled in each group of approximately 1000data. How many data should be included in each group (socalled data bin) still needs some discussions in statisticalpoint of view. Different selection of bin size (the rangeto determine the damage probability) may cause differ-ent result of regression. In addition, we assumed that thenumerical model results represent the features of tsunamiinundation flow in the study area without errors. This isthe issue that should be discussed as statistical analysis oftsunami fragility curves considering the proper selectionof bin size and uncertainties in the numerical model.

4. Tsunami Fragility for Fatality Estimation

Tsunami fragility for fatality is determined by using thepost-tsunami data in terms of the number of dead, missingand survivors. Fig. 7 shows the spatial distribution of theratio of dead, missing and survivors in each desa (village)in Banda Aceh city (as of 12 April 2005), normalized bythe pre-tsunami desa population [13]. GIS analysis ofthe fatality information and the numerical model resultsin Fig. 2(e) yields a fragility curve for tsunami fatalitiesas the relationship between the fatality ratio (both dead

� � � � � ��

� ��

��

��

��

��

��

Fig. 7. Proportion of fatalities in Banda Aceh city [13], cal-culated by using reported number of fatality and pre-tsunamipopulation.

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Fig. 8. Tsunami fragility curves for fatality in terms of (a)the inundation depth and (b) inundation height. The solidline is the best-fitted curve of the plot (◦ : the distribution offatality ratio) with Eq. (2).

and missing) and the hydrodynamic features of tsunami.Based on tsunami fatality data in Fig. 7, the representa-tive value of local hydrodynamic feature of tsunami inun-dation or the inundation height is calculated by taking themedian value of modeled inundation depths and inunda-tion heights within each desa.

Figure 8 shows the tsunami fragility expressed as thefatality ratio with regard to the representative values ofthe maximum inundation depth dmax (measured above thelocal ground level) and inundation height hmax (measuredabove the pre-tsunami tide level) calculated by taking themedian value of the numerical model results within eachdesa as shown in Fig. 7. The fragility curve is determinedby assuming the standardized normal distribution functionof Eq. (2) with the parameters of Table 2 obtained throughthe least-squares fitting, where x is the median value of theinundation depth or inundation height (m) in each desa,calculated by using of the numerical model results.

Note that the fatality ratio distribution is the result ofthe post-tsunami investigation based on the pre-tsunamiregistration data [13]. It is highly unknown where theresidents were affected by the tsunami inundation flow,because it is easily guessed that the residents who were



��!�"��

#$

�

%

&

'

*

��

��

��

��

�

�

� ��

+

��

�

��

�

�

,��!� -�/�� :��!��!��;��

=

>

?

� +

@

Fig. 9. The result of visual interpretation of structural damage in Banda Aceh city andthe points of field measurements. See their original paper [19] for all the points andareas investigated.


Tsunami Fragility

Table 2. Statistical parameters of tsunami fragility curvesfor fatality (Fig. 8).

Fragility curve µ σ R2

(a) dmax (m) 3.92 1.15 0.62(b) hmax (m) 5.37 0.76 0.72

aware of tsunami arrival have evacuated and tried to sur-vive. In other words, the fragility curve of Fig. 8 does notindicate the human’s survival possibility according to thelocal hydrodynamic features of tsunami inundation flow.Also, taking median to obtain the representative values oftsunami inundation depth according to each desa reflectshigher variance of the plot compared with that of Fig. 6.For the above reasons, this fragility function should be in-terpreted as a macroscopic measure of tsunami impact, i.e.the occurrence of tsunami fatality significantly increasewhen the local inundation depth exceeds approximately2 m and the inundation height 4.5 (m), and almost impos-sible to survive when the local inundation depth exceeds6 m.

5. Tsunami Fragility from Satellite RemoteSensing and Field Survey

Developing tsunami fragility may also be viewed usingsatellite images and post-tsunami surveys. Namegaya andTsuji (2006) investigated the structural damage in BandaAceh city by the 2004 tsunami, using the visual inspec-tion of QuickBird pre and post-tsunami satellite imagesacquired on June 23, 2004 and December 28, 2004, infour areas of Banda Aceh city together with the mea-surements of tsunami inundation depth and height [19].Fig. 4 is showing their result of visual interpretation ofstructural damage in Banda Aceh city and the points oftsunami measurements. The markers • and ◦ in the figuredenote survived and destroyed (washed-away) structuresinterpreted from the satellite images focusing on the pres-ence of their roofs.

Using field survey results presents difficulties in cor-relating the tsunami heights at all the points where thestructural damage was inspected. In this case, the dam-age probability is calculated by counting the number ofsurvived and destroyed structures within the dashed-linecircles of 100 m diameter (see Fig. 4) and the tsunami in-undation depth or height is represented by the measuredvalue at the center of each solid-line circle. Consequently,the relationships between the damage probabilities andtsunami heights are obtained at 13 to 15 points in BandaAceh city (see Namegaya and Tsuji (2006) for details).Table 3 is the result obtained at each area equivalent toFig. 4. The tsunami fragility curves are determined asFig. 10, with statistical parameters of Table 4. Note thattwo fragility curves in Fig. 10(a), although very similar,show differences in approaches of damage data compila-tion (including number of samples and the inspected area)and methods to obtain tsunami hazard information.

Table 3. The damage probabilities and measured tsunamiheights obtained by the visual inspection of satellite imagesand field measurements. The area and points A to O areequivalent to Fig. 4.

Area hmax dmax Destroyed Pre-tsunami Damage(m) (m) structures structures probability

A 7.0 5.4 40 41 0.98B 7.1 5.0 51 57 0.89C 7.6 4.7 32 45 0.71D 7.1 5.0 72 83 0.87E 8.2 4.9 46 57 0.81F 6.6 3.7 51 60 0.85G 4.6 2.5 2 31 0.06H 4.7 2.1 12 30 0.40I 12.0 – 72 75 0.96J 9.7 7.9 60 61 0.98K 6.1 3.3 51 56 0.91L 6.6 3.8 36 50 0.72M 6.8 4.5 38 50 0.76N 6.5 2.6 14 30 0.47O 4.8 – 6 31 0.19

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

Fig. 10. Tsunami fragility curves for structural destruction,in terms of (a) the maximum inundation depth and (b) themaximum inundation height, by using the visual inspectionof satellite images and field measurements. The dashed linein (a) indicates the fragility curve from Fig. 6(a).

6. Tsunami Fragility from Historical Data

In constructing the tsunami fragility from historicalevents, we incorporated the historical tsunami data onlocal tsunami damage and height. In Japan, the post-tsunami surveys would be conducted by many differ-ent organizations and individuals. After the 1896 MeijiSanriku earthquake tsunami, which caused approximately22,000 casualties, the damage survey efforts were con-ducted by the central government [20] and an engineerSoshin Yamana delegated by Iwate prefectural govern-ment (his report was published by Yamashita (1982) withcomments and interpretations) [21]. Seismological re-searchers and Japan Meteorological Agency also con-ducted the survey after the 1933 Showa Sanriku earth-quake tsunami which caused approximately 3,000 casu-alties [22, 23].

Hatori (1984) compiled the house damage data fromthe historical documents of the 1896 Meiji Sanriku, 1933Showa Sanriku and the 1960 Chile tsunami events aslisted in Table 5. He defined the structural damage prob-



Table 4. Statistical parameters for fragility curves (Fig. 10).

x µ σ R2

dmax (m) 3.33 1.25 0.62hmax (m) 6.07 1.36 0.53

Table 5. Historical tsunami data in Japan used for develop-ing tsunami fragility.

Data compilation Events Original data[References] [References]Hatori (1984) [25] 1896 Meiji-Sanriku [21]

1933 Showa-Sanriku [22]1960 Chile [23]

Shuto (1987a, 1993) 1896 Meiji-Sanriku [20, 21, 24, 27–30][1, 26]

ability as Eq. (4) by counting the number of houses inthree damage categories ; destroyed/washed-away, mod-erate and only flooded, in each reported area or settlementwith tsunami height ;

PD =a+b/2

a+b+ c. . . . . . . . . . . . . (4)

where a, b and c is the number of the houses catego-rized as destroyed/washed-away, moderate damage andonly flooded, respectively.

Shuto (1987a, 1993) also compiled the documents andreports from the 1896 Meiij Sanriku tsunami (Table 5),and determined the damage probability with four damagecategories ;

PD =a+b+ c/2

a+b+ c+d. . . . . . . . . . . (5)

where a, b, c and d is the number of the houses cate-gorized as washed-away, completely destroyed, moderatedamage and only flooded, respectively. To increase thereliability of data, he conducted the numerical modelingand the additional field survey to determine the reliabilityof the documents.

Figure 11 plots historical data (damage probability ver-sus tsunami height and inundation depth) compiled byHatori (1984) and Shuto (1987a, 1993), and the fragilitycurves (solid lines) obtained by the least-squares fittingof Eq. (2). Since the historical data is highly dispersed,the dashed lines are also added to indicate the maximumand minimum limits by the authors’ interpretation (prob-ably with less statistical meaning). The statistical param-eters of fragility curves with the solid and dashed linesare summarized in Table 6. The high dispersion is proba-bly caused by numerous uncertain factors in terms of thereliability of historical data. It is quite unknown, for ex-ample, how the tsunami heights were measured (datum)and represented in their reports, e.g. single or multiplemeasurements. Accordingly, these fragility curves shouldbe interpreted as a coarse measure with uncertainty, e.g.2 m tsunami is equivalent to cause 0-30 % of probabilitythat a house would be destroyed (Fig. 11(d)). In anotherway, by using fragility curves, the magnitude of tsunami

1.0

0.8

0.6

0.4

0.2

0.0

PD

20151050

Tsunami height (m)

1933 Showa (Hatori) Fragility Historical data

(b)1.0

0.8

0.6

0.4

0.2

0.0

PD

20151050

Tsunami height (m)

1896 Meiji (Hatori)FragilityHistorical data

(a)

1.0

0.8

0.6

0.4

0.2

0.0

PD

20151050

Tsunami height (m)

1960 Chile (Hatori) FragilityHistorical data

(c) 1.0

0.8

0.6

0.4

0.2

0.0

PD

20151050

Tsunami height (m)

All (Hatori)FragilityHistorical data

(d)

1.0

0.8

0.6

0.4

0.2

0.0

PD

20151050

Tsunami height (m)

1896 Meiji (Shuto)FragilityHistorical data

(e) 1.0

0.8

0.6

0.4

0.2

0.0

PD

1086420

Inundation depth (m)

1896 Meiji (Shuto)

FragilityHistorical

data

(f)

Fig. 11. Historical tsunami data of Hatori (1984) and Shuto(1987a, 1993), and tsunami fragility curves. (a) 1896 MeijiSanriku tsunami by Hatori, (b) 1933 Showa Sanriku tsunamiby Hatori, (c) 1960 Chile tsunami by Hatori, (d) Three eventsby Hatori, (e) and (f) 1896 Meiji Sanriku tsunami by Shuto.

Table 6. Statistical parameters of tsunami fragility curves(Fig. 11). µ,σ for the regression of the historical data (solidline) and µ ′,σ ′,µ ′′,σ ′′ for upper and lower limits.

Fragility curve µ σ R2 µ ′ σ ′ µ ′′ σ ′′

(a) 5.84 3.28 0.30 2.80 1.35 9.00 5.00(b) 5.09 2.50 0.70 2.90 1.35 7.00 2.50(c) 4.66 1.09 0.80 N/A N/A N/A N/A(d) 5.97 2.66 0.55 2.80 1.30 10.0 3.60(e) 6.05 2.49 0.24 2.00 0.90 10.5 5.00(f) 5.49 1.26 0.36 N/A N/A N/A N/A

can be speculated from the documented damage, e.g. 30% of structural damage would be potentially caused bythe tsunami of 2.1-7.4 m height, as shown in Fig. 11(d),as an empirical relationships between tsunami hazardsand local vulnerability learned from the historical Sanrikutsunami disasters.

Historical tsunami fragility aims to identify the rela-tionships among the tsunami hazards, the damage and un-certain historical documents. For instance, large numbersof descriptions can be found in historical documents, say-ing “An abnormal tide reached to the entrance of a shrine


Tsunami Fragility

which was X m above the sea level” or “An abnormaltide penetrated in the village to cause Y houses washed-away.” To identify the origin or cause of the descriptionsabove, the former requires the interpretation of potentialdamage equivalent to the reported tsunami height of X m.Also the latter requires the potential tsunami height equiv-alent to the reported amount of damage Y .

7. Concluding Remarks

We proposed a new measure called Tsunami fragilitythroughout the statistical analysis of tsunami damage datainterpreted from the high-resolution satellite images orfield survey, numerical modeling and historical docu-ments, to identify the relationship between tsunami haz-ard and vulnerability. Tsunami fragility is expressedby the structural damage probability or fatality ratio asthe functions of hydrodynamic features of tsunami, suchas inundation depth, current velocity and hydrodynamicforce. Especially, the integration of satellite remote sens-ing and numerical modeling leads to a significant knowl-edge on structural vulnerability against the 2004 Sumatra-Andaman earthquake tsunami, in Banda Aceh, Indonesia.

We suggest that tsunami fragility is implemented foran assessment of structural damage and fatalities withinthe exposed area against potential tsunami hazard scenar-ios. Multiplying the number of exposed structures andpopulations by the damage probability from the fragilitycurves equivalent to the estimated tsunami hazards pro-vides the quantitative estimation of tsunami damage. Itis still highly speculative, however, to say that the pro-posed tsunami fragility can become an universal measureof tsunami impact or damage. As stated in the introduc-tion, the tsunami damage should be characterized by nu-merous uncertain factors. In this sense, tsunami fragilityproposed here includes some of uncertainties, but not all.In other words, they may not be applicable in consideringtsunami vulnerability when changing the areas of inter-est or considering other tsunami scenarios. Thus, we alsosuggest that careful use and interpretations are requiredin using proposed tsunami fragility when applying. Webelieve that further more precise investigations from the2004 and other historical events can expand the applica-bility of tsunami fragility.

AcknowledgementsThis research was financially supported, in part, by the Indus-trial Technology Research Grant Program in 2008 (Project ID :08E52010a) from New Energy and Industrial Technology Devel-opment Organization (NEDO), and the Grant-in-Aid for Scien-tific Research (Project Number : 19681019) from the Ministry ofEducation, Culture, Sports, Science and Technology (MEXT) ofJapan.

References:[1] N. Shuto, “Tsunami intensity and disasters, Tsunamis in the World,”

Kluwer Academic Publishers, pp. 197-216, 1993.[2] H. Izuka and H. Matsutomi, “Damage estimation due to tsunami in-

undation flow,” Proceedings of Coastal Engineering, JSCE, Vol.47,pp. 381-385, 2000 (in Japanese).

[3] R. Asakura, K. Iwase, T. Ikeya, M. Takao, N. Fujii, and M. Omori,“An experimental study on wave force acting on on-shore structuresdue to over flowing tsunamis,” Proceedings of Coastal Engineering,JSCE, Vol.47, pp. 911-915, 2000 (in Japanese).

[4] H. Yeh, “Design Tsunami Forces for Onshore Structures,” Journalof Disaster Research, Vol.2, No.6, pp. 531-536, 2007.

[5] J. Borrero, “Field survey of northern Sumatra and Banda Aceh, In-donesia after the tsunami and earthquake of 26 December 2004,”Seismological Research Letters, Vol.76, No.3, pp. 309-318, 2005.

[6] H. Matsutomi, T. Sakakiyama, S. Nugroho, and M. Matsuyama,“Aspects of inundated flow due to the 2004 Indian Ocean tsunami,”Coastal Engineering Journal, Vol.48, No.2, pp. 167-195, 2006.

[7] Y. Tsuji, Y. Tanioka, H. Matsutomi, Y. Nishimura, T. Kamataki,Y. Murakami, T. Sakakiyama, A. Moore, G. Gelfenbaum, S.Nuguroho, B. Waluyo, I. Sukanta, R. Triyono, and Y. Namegaya,“Damage and height distribution of Sumatra earthquake – Tsunamiof December 26, 2004, in Banda Aceh city and its environs,” Jour-nal of Disaster Research, Vol.1, No.1, pp. 103-115, 2006.

[8] K. Fujima, Y. Shigihara, T. Tomita, K. Honda, H. Nobuoka, M.Hanzawa, H. Fujii, H. Otani, S. Orishimo, M. Tatsumi, and S.Koshimura, “Survey results of the Indian Ocean tsunami in the Mal-dives,” Coastal Engineering Journal, Vol.48, No.2, pp. 91-97, 2006.

[9] K. Satake, T. T. Aung, Y. Sawai, Y. Okamura, K. S. Win, W. Swe,C. Swe, T. L. Swe, S. T. Tun, M. M. Soe, T. Z. Oo, and S. H. Zaw,“Tsunami heights and damage along the Myanmar coast from theDecember 2004 Sumatra-Andaman earthquake,” Earth, Planets andSpace, Vol.58, pp. 243-252, 2006.

[10] O. Murao and F. Yamazaki, “Development of fragility curves forbuildings based on damage survey data of a local government af-ter the 1995 Hyogoken-Nanbu earthquake,” Journal of Structuraland Construction Engineering, Vol.527, pp. 189-196, 2000 (inJapanese).

[11] M. Shinozuka, M. Q. Feng, J. Lee, and T. Naganuma, “Statisti-cal analysis of fragility curves,” Journal of Engineering Mechanics,Vol.126, pp. 1224-1231, 2000.

[12] Federal Emergency Management Agency, “Multi-hazard Loss Es-timation Methodology, HAZUS-MH MR3 (Earthquake Model),”Technical Manual, pp. 5-35, 2003.

[13] Japan International cooperation Agency (JICA), “The study on theurgent rehabilitation and reconstruction support program for Acehprovince and affected areas in north Sumatra, Final Report (1),”Vol.IV : Data book, 2005.

[14] T. Oie, S. Koshimura, H. Yanagisawa, and F. Imamura, “Numericalmodeling of the 2004 Indian Ocean tsunami and damage assessmentin Banda Aceh,” Indonesia, Annual Journal of Coastal Engineering,JSCE, Vol.53, pp. 221-225, 2006 (in Japanese).

[15] S. Koshimura, T. Oie, H. Yanagisawa, and F. Imamura, “Developingfragility functions for tsunami damage estimation using numericalmodel and post-tsunami data from Banda Aceh, Indonesia,” CoastalEngineering Journal, No.3, pp. 243-273, 2009.

[16] I. Aida, “Reliability of a tsunami source model derived from faultparameters,” Journal of Physics of the Earth, Vol.26, pp. 57-73,1978.

[17] K. R. Karim, and F. Yamazaki, “A simplified method of constructingfragility curves for highway bridges, Earthquake Engineering andStructural Dynamics,” Vol.32, pp. 1603-1626, 2003.

[18] M. Saatcioglu, A. Ghobarah, and I. Nistor, “Performance of struc-tures in Indonesia during the December 2004 great Sumatra earth-quake and Indian Ocean tsunami,” Earthquake Spectra, Vol.22,No.3, pp. S295-S319, 2006.

[19] Y. Namegaya and Y. Tsuji, “Distributions of the Swept AwayHouses in Banda Aceh City, Indonesia, due to the 2004 IndianOcean Tsunami Estimated by Satellite Images,” Annual Journalof Coastal Engineering, JSCE, Vol.53, pp. 286-290, 2006 (inJapanese).

[20] T. Iki, “Field survey of Sanriku tsunami disasters, Bulletin of theImperial Earthquake Investigation Committee,” Vol.11, pp. 5-34,1896 (in Japanese).

[21] F. Yamashita, “Tragic histories of great Sanriku tsunami disasters,”Seijisha, 413p. 1982 (in Japanese).

[22] Earthquake Research Institute, “Reports on the 3 March, Showa8 Sanriku tsunami,” Bulletin of the Earthquake Research Institute,Tokyo Imperial University, Supplementary Vol.1, 248p., 1934 (inJapanese).

[23] Japan Meteorological Agency, Sendai Branch, “Report of the 24May, Showa 35, Chilean earthquake tsunami,” 1961 (in Japanese).



[24] Iwate Prefectural Government, “Report of the recovery processfrom the 1960 Chilean earthquake tsunami disaster,” 251p., 1969(in Japanese).

[25] T. Hatori, “Damage probability of houses by tsunamis,” Bulletin ofthe Earthquake Research Institute, Vol.59, pp. 433-439, 1984 (inJapanese).

[26] N. Shuto, “Evolution of tsunami disasters,” Tsunami EngineeringTechnical Report, Tohoku University, Vol.4, pp. 1-41, 1987a (inJapanese).

[27] N. Shuto and C. Goto, “Field survey of great Sanriku tsunami– from Raga, Hiraiga, Shimanokoshi, Omoto and Shimokonari,”Tsunami Engineering Technical Report, Tohoku University, Vol.2,pp. 39-45, 1985a (in Japanese).

[28] N. Shuto and C. Goto, “Field survey of great Sanriku tsunami– from Okkirai,” Tsunami Engineering Technical Report, TohokuUniversity, Vol.2, pp. 46-53, 1985b (in Japanese).

[29] F. Imamura, C. Goto, and N. Shuto, “Study on Numerical TsunamiForecasting System – Accuracy of numerical models,” Tsunami En-gineering Technical Report, Tohoku University, Vol.3, pp. 23-87,1986 (in Japanese).

[30] N. Shuto, J. Sayama, and K. Fujima, “Field survey of great Sanrikutsunami – from Ofunato,” Tsunami Engineering Technical Report,Tohoku University, Vol.4, pp. 101-113, 1987b (in Japanese).

Name:Shunichi Koshimura

Affiliation:Associate Professor, Graduate School of Engi-neering, Tohoku University

Address:Aoba 6-6-11-1104, Aramaki, Aoba-Ku, Sendai 980-8579, JapanBrief Career:2000-2002 JSPS Research Fellow, National Oceanic and AtmosphericAdministration2002-2005 Research Scientist, Disaster Reduction and Human RenovationInstitute2005- Associate Professor, Tohoku UniversitySelected Publications:• S. Koshimura, T. Oie, H. Yanagisawa, and F. Imamura, “Developingfragility functions for tsunami damage estimation using numerical modeland post-tsunami data from Banda Aceh, Indonesia,” Coastal EngineeringJournal, No.3, pp. 243-273, 2009.• S. Koshimura, Y. Hayashi, K. Munemoto, and F. Imamura, “Effect of theEmperor seamounts on trans-oceanic propagation of the 2006 Kuril Islandearthquake tsunami,” Geophysical Research letters, Vol.35, L02611,doi:10.1029/2007GL032129, 24, 2008.• S. Koshimura, T. Katada, H. O. Mofjeld, and Y. Kawata, “A method forestimating casualties due to the tsunami inundation flow,” Natural Hazards,Vol.39, pp. 265-274, 2006.Academic Societies & Scientific Organizations:• Japan Society of Civil Engineers (JSCE)• Institute of Social Safety Science• Japan Associaiton for Earthquake Engineering (JAEE)• Japan Society for Computational Engineering and Science (JSCES)• Americal Geophysical Union (AGU)

Name:Yuichi Namegaya

Affiliation:Postdoctoral Research Fellow, Geological Sur-vey of Japan, National Institute of Advanced In-dustrial Science and Technology

Address:Site 7, Higashi 1-1-1, Tsukuba 305-8567, JapanBrief Career:2005-2007 JSPS Research Fellow, Earthquake Research Institute, theUniversity of Tokyo2007- Postdoctoral Research Fellow, Geological Survey of Japan, NationalInstitute of Advanced Industrial Science and TechnologySelected Publications:• Y. Namegaya, Y. Tanioka, K. Abe, K. Satake, K. Hirata, M. Okada, andA. R. Gusman, “In situ measurements of tide gauge response andcorrections of tsunami waveforms from the Niigataken Chuetsu-okiearthquake in 2007,” Pure and Applied Geophysics, Vol.166, pp. 97-116,2009.• Y. Namegaya and K. Satake, “Tsunami generated by the 2007 NotoHanto earthquake,” Earth, Planets and Space, Vol.60, pp. 127-132, 2008.Academic Societies & Scientific Organizations:• Japan Society of Civil Engineers (JSCE)• Seismological Society of Japan (SSJ)• Americal Geophysical Union (AGU)

Name:Hideki Yanagisawa

Affiliation:Company Member, Tokyo Electric Power Ser-vices Company Limited

Address:Higashi-Ueno 3-3-3, Taito-ku, Tokyo 110-0015, JapanBrief Career:2008-2009 Post-Doctoral Research Fellow, Graduate School ofEngineering, Tohoku University2009- Company Member, Tokyo Electric Power Services CompanyLimitedSelected Publications:• H. Yanagisawa, S. Koshimura, K. Goto, T. Miyagi, F. Imamura, A.Ruangrassamee, and C. Tanavud, “Damage of mangrove forest by the2004 Indian Ocean tsunami at Pakarang Cape and Namkem, Thailand,”Polish Journal of Environmental Studies, Vol.18, No.1, pp. 35-42, 2009.• H. Yanagisawa, S. Koshimura, K. Goto, T. Miyagi, F. Imamura, A.Ruangrassamee, and C. Tanavud, “The reduction effects of mangroveforest on a tsunami based on field surveys at Pakarang Cape, Thailand andnumerical analysis, Estuarine,” Coastal and Shelf Science, Vol.81, pp.27-37, 2009.Academic Societies & Scientific Organizations:• Japan Society of Civil Engineers (JSCE)• Japan Society for Mangroves


tsunami fragility – a new measure to identify tsunami damage · 2012-01-25 · tsunami fragility...

Documents