Miura Earthquake Damage Estimation In Metro Manila

ARTICLE IN PRESS

Soil Dynamics and Earthquake Engineering 28 (2008) 764–777 www.elsevier.com/locate/soildyn

Earthquake damage estimation in Metro Manila, Philippines based on seismic performance of buildings evaluated by local experts’ judgments Hiroyuki Miuraa,!, Saburoh Midorikawab, Kazuo Fujimotoc, Benito M. Pachecod, Hiroaki Yamanakae a

Center for Urban Earthquake Engineering, Tokyo Institute of Technology, G3-3, 4259 Nagatsuta, Midori-ku, Yokohama 226-8502, Japan b Department of Built Environment, Tokyo Institute of Technology, Yokohama, Japan c Department of Risk and Crisis Management System, Chiba Institute of Science, Choshi, Japan d Vibrametrics Inc., Quezon, Philippines e Department of Environmental Science and Technology, Tokyo Institute of Technology, Yokohama, Japan Received 31 May 2006; received in revised form 7 September 2007; accepted 11 October 2007

Abstract Building damage due to a scenario earthquake in Metro Manila, Philippines is estimated based on seismic performance of the buildings evaluated by local experts’ judgments. For the damage estimation, building capacity curves and fragility curve are developed from questionnaire to the local experts of structural engineering. The Delphi method is used to integrate the experts’ opinions. The derived capacity curves are validated by comparing with the result of pushover analysis for typical buildings. Building responses due to simulated ground motions are estimated by the capacity spectrum method. Damage ratios are calculated from the fragility curves and the building responses. Distributions of the damaged buildings are computed by multiplying the damage ratios and the building inventory. The distribution and the amount of the damaged buildings in this study show significant difference from the estimation with the capacity curves of HAZUS, suggesting the importance of evaluation of the region-specific building performance. r 2007 Elsevier Ltd. All rights reserved. Keywords: Building damage estimation; Seismic performance; Capacity spectrum method; Local experts; Delphi method; Metro Manila

1. Introduction Population growth and urban expansion in mega-cities increase vulnerability to disasters in developing countries. In order to establish efficient earthquake disaster mitigation planning, earthquake loss estimation is indispensable. In particular, building damage estimation is important for loss estimation since the damaged buildings result in great economic loss and casualties. In order to carry out building damage estimation, it is necessary to evaluate following three points: (1) estimating the ground motions due to a scenario earthquake by modeling the source and the underground structure in the area of interest, (2) evaluating the damage ratio based on !Corresponding author. Tel.: +81 45 924 5602; fax: +81 45 924 5574.

E-mail address: [email protected] (H. Miura). 0267-7261/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.soildyn.2007.10.011

the seismic performance of the local buildings, and (3) computing the damage distribution and the number of damaged buildings by multiplying the damage ratio by the building inventory. Therefore, it is important to gather the data for underground structure, vulnerability of buildings and building inventory. This study is mainly focused on the evaluation of the building performance for the damage estimation in a developing country. One of the standardized tools for earthquake loss estimation is HAZUS [1] developed in the US. In HAZUS, the seismic performance of typical buildings in the US is given. The seismic performance of buildings, however, should be region-specific because of the different design level and construction quality in each region. Therefore, it is not appropriate to apply the building performance in HAZUS to other regions. For developing countries, simple tools for loss estimation have been proposed in RADIUS

ARTICLE IN PRESS 765

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

[2] and GESI [3]. The reliability of the estimation by RADIUS or GESI, however, would not be high because the tools were developed for highly simplified loss estimation. The Philippines is one of developing countries located in a zone of high seismicity. Metro Manila, the capital of the Philippines, is a mega-city that is highly populated in the urban areas. Building damage estimation due to scenario earthquakes in Metro Manila has been conducted based on HAZUS [4], GESI [5], and vulnerability functions constructed from observed damage data of the 1990 Luzon earthquake [6]. The vulnerability functions, however, were developed only for low-rise buildings in the Philippines. It is necessary to examine the seismic performance of mid-rise and high-rise buildings for more reliable damage estimation in urban areas. The capacity spectrum method (e.g., [7,8]) is a simplified procedure estimate non-linear building response from the capacity of a building and the demand of ground motion on the building. In the method, the seismic performance of buildings can be incorporated rationally. For obtaining the capacity of the buildings, it is a valid way to integrate experts’ judgments when the available experimental and actual damage data to evaluate the building performance is limited. In this study, questionnaire for local experts of structural engineering in Metro Manila is applied to develop seismic capacity curves of the buildings for more reliable building damage estimation. The derived capacity curves are validated by comparing with result of pushover analysis for typical buildings. Building damage due to a scenario earthquake is computed by multiplying damage ratio estimated from the capacity curve and simulated ground motion by building inventory. The estimated damage distribution is compared with that by the capacity curves of HAZUS to examine the effects of the region-specific building performance on the damage estimation. 2. Earthquake environment in Metro Manila, Philippines Metro Manila consists of seventeen cities and municipalities including Manila, Makati, Quezon and Marikina. Fig. 1 shows the location of Metro Manila and the urban sprawl [9]. In around 1950 the urbanized area was less than 100 km2 with a population of 1.6 million, but now is expanded to more than 600 km2 with a population of 10 million. In the old areas in Metro Manila such as Manila city, densely built-up area with low-rise and mid-rise buildings has been developed. In the newly developed commercial zones such as Makati and Marikina, many high-rise buildings have been constructed. According with the sprawl of the urbanized area, new commercial zones have been expanded. Fig. 2 shows the geomorphological classification map of Metro Manila [10]. The area is divided broadly into three parts: Central plateau, Coastal lowland and Marikina valley. The central plateau is on stiff soils with an elevation

Developing Period : - 1948 : - 1966 : - 1975 : - 1996

Fig. 1. Location of Metro Manila with urban sprawl after Doi and Kim [9].

Manila

Sierra Madre Range

Manila Bay Makati

Laguna de Bay

0

10 km

Geomorphological Unit : Coastal Lowland : Marikina Valley : Central Plateau : Mountain

Fig. 2. Geomorphological classification map and active faults in Metro Manila after Matsuda et al. [10].

of 15–30 m. The coastal lowland extending along the Manila bay is on soft sand and clay deposits with a thickness of several to 40 m. The Marikina valley is bounded by the central plateau and the Sierra Madre range, and consists of a delta and a muddy flood plain. The thickness of the surface soft deposits reaches 50 m at a maximum. Since the Luzon Island including Metro Manila is located between the Eurasian Plate to the west and the Philippine Sea Plate to the east, the seismic and volcanic activities are high. After the Spanish Empire colonized the Philippines in the 15th century, description or accounts of earthquakes have been maintained in various letters and chronicles. The historical earthquake data in Metro Manila, as well as the instrumentally derived earthquake data gathered in the 20th century, have been compiled in

ARTICLE IN PRESS 766

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

the previous study [11]. According to the earthquake data, seismic intensity more than VII in Modified Rossi-Forel intensity scale have been recorded for 28 times during recent 400 years. As an example of the recent earthquakes, in the Luzon earthquake of July 16, 1990 (M7.8), Intensity VII was recorded and minor building damage was caused in Metro Manila. The average return period for a destructive earthquake (Intensity VIII) was roughly estimated at about 80 years [11]. In the Metro Manila area, there are two major active faults. One is the West valley fault located between the central plateau and the Marikina valley, and another is the East valley fault situated between the Marikina valley and the Sierra Madre range. Trench-excavation survey at the northern end of the West valley fault suggests the recurrence of hundreds rather than thousands of years [12]. Besides, these faults have high potential to produce a damaging earthquake with magnitude of 6–7 [12]. Disaster mitigation planning to the earthquakes triggered by these faults seems as urgent issue for Metro Manila.

3. Flow of building damage estimation 3.1. Overview Fig. 3 shows the flowchart of the building damage estimation adopted in this study. After setting parameters for a fault model of a scenario earthquake in Metro Manila, ground motions at surface are computed using hybrid simulation technique [13,14] and soil response analysis [15] based on underground structure model. Building response due to the ground motion is evaluated by the capacity spectrum method. First, the buildings existing in Metro Manila are classified into several Scenario Earthquake

Hybrid Simulation and Soil Response Analysis

Computation of Surface Ground Motion Capacity Curves Capacity Spectrum Method Fragility Curves Estimation of Damage Ratio Building Inventory Distribution and Amount of Building Damage Fig. 3. Flow of building damage estimation.

categories. Capacity curve for each category is developed by integrating the experts’ opinions. The non-linear response of the building is estimated from the capacity curve and demand curve converted from the ground motion spectrum. Damage state for each building category is determined by the building response and fragility curves. Finally, combining the damage state of each building category and building inventory data, the distribution of the building damage is computed. 3.2. Ground motion estimation The West valley fault is selected as the source of a scenario earthquake because the fault is closer to the central part of Metro Manila. The ground motions due to the West valley fault are simulated using the fault model and the underground structure model. Fig. 4(a) and (b) shows the fault model and major fault parameters used in the simulation. After determining the fault length from the geomorphology in and around the fault, the other fault parameters such as the fault width, the seismic moment, the area of asperities and the average slip are estimated based on the recipe for predicting strong ground motions [16]. The fault length and the moment magnitude of the earthquake are set as 40 km and Mw 6.7, respectively. Two asperities are located in the fault and the rupture starts from northern bottom of the fault. The underground structure model with a 500 m mesh system is constructed from the about 400 boring data, the geomorphological classification map [10] and the geophysical explorations [17]. The ground motions on the engineering bedrock with the shear-wave velocity of about 400 m/s are computed by the hybrid simulation technique [13,14]. The simulation technique consists of the stochastic green function for ground motion with short period (less than 1 s) and the 3-D finite difference method for ground motion with long period (more than 1 s). The ground motions at the surface are computed by the soil response analysis with the SHAKE program [15]. The surface soils in Metro Manila are broadly classified into three types: clay, sand and gravel. The dynamic soil properties proposed in the previous study [18] are applied in the computation. Fig. 5 illustrates the relationships between the shear modulus ratio, damping factor and shear strain for each soil type used in the analysis. Fig. 4(c) shows the computed peak ground velocity (vectorial summation of two horizontal motions) on the surface. Fig. 6 indicates 5%-damped velocity response spectrum and demand curves at Ermita and Quezon computed from the simulated ground motions. The demand curve is defined by the relationship between the spectral response displacement and the response acceleration. The maximum velocity response at Ermita reaches almost 5 m/s, while the response at Quezon is less than 1 m/ s. This is because that the ground motion at Ermita is strongly amplified due to the thick soft soil deposits in the coastal lowland area.

ARTICLE IN PRESS H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

South

767

North

4km

Surface

16.6km

0.5km

13.6km

As1

As2

3.1km Bg2

Bg1

41km Starting point of rupture

Strike (deg.)

North

N200E

South

N190E

Dip Angle (deg.) Fault area (km × km) Asperity Area As1 (km × km)

As2

Moment Magnitude (Mw) Total M0(dyne ·cm) As1 As2 As1 As2 Average Slip (m) Bg1 Bg2

90 41 × 16.6 7×6 10 × 10 6.7 1.68 × 1026 3.01 × 1025 9.49 × 1025 1.6 2.3 0.3 0.4

Fig. 4. (a) Fault model of scenario earthquake. As1, 2 and Bg1, 2 show areas of asperities and backgrounds, respectively. (b) Fault parameters. (c) Distribution of peak ground velocity due to the scenario earthquake.

0.4 G/G0 Clay Sand Gravel

0.2

0.5 h Clay Sand Gravel

0

10−6

10−5

10−4 Shear Strain, γ

10−3

Damping factor, h

Shear modulus ratio, G/G0

1

0 10−2

Fig. 5. Relationships between shear modulus ratio, damping factor and shear strain proposed by Imazu and Fukutake [18].

3.3. Building inventory In Metro Manila, there had been the building inventory data digitized from 1/10,000 scale topographic maps edited in 1989 [19]. The authors have updated the inventory data using the satellite remote sensing data [20]. In the inventory data, the attribute for number of stories, which mainly controls the vibration period during ground shaking, is included for each building. The inventory for the mid-rise and high-rise buildings was updated using the high-

resolution satellite IKONOS images. The dotted squares in Fig. 7 indicate the coverage of the images that cover about 75% of Metro Manila including the major commercial areas such as Manila, Makati, Quezon and Marikina. The locations of the newly constructed buildings were extracted from the difference between the IKONOS images and the existing inventory data. The number of stories was estimated for each building using the shadow lengths of the buildings obliquely observed from the satellite. The inventory for the low-rise buildings was updated from the land cover classification map derived from the multitemporal Landsat images [21]. The detail of the analysis for the updating is described in the authors’ previous study [20]. Fig. 7 shows the updated building distribution with a 500 m mesh system. The total number of buildings in the updated inventory was estimated at about 1.29 million. According to the recent national census in 2000 [6], the total number of buildings in Metro Manila is approximately 1.32 million. The updated number of the buildings shows good agreement with the census data. Due to the updating, the number of buildings is increased by about 40% over the 15-year period. As shown in Fig. 7, the buildings are densely concentrated in the western coastal area such as Manila. A lot of the buildings are distributed also in the northern, southern and eastern areas with the expansion of the urbanized areas as shown in Fig. 1.

ARTICLE IN PRESS 768

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

10

10 Vel. Response Spectrum (m/s)

Vel. Response Spectrum (m/s)

Ermita

1

0.1 NS EW

Quezon

1

0.1 NS EW

0.01

0.01 0.1

1 Period (s)

0.1

10

30

1 Period (s)

10

30 Ermita

Quezon

20

SA (m/s/s)

SA (m/s/s)

20

10

10 NS EW

NS EW

0

0 0.2

0.4 SD (m)

0.6

0.8

0.2

0.4 SD (m)

0.6

0.8

Fig. 6. Five percent damped velocity response spectra and demand curves at Ermita and Quezon.

In order for rational building damage estimation, not only the number of stories but also the structural type and the design vintage for each building are indispensable. Only the footprints and the number stories, however, are assigned in the inventory data. The estimation of structural type and design vintage for each building is discussed in Section 5.1. 4. Evaluation of seismic performance of buildings 4.1. Classification of buildings

0

10 km

No. of buildings 1,000 – 500 – 999 300 – 499 100 – 299 1 – 99

: Coverage of IKONOS images Fig. 7. Building distribution of inventory data updated by Miura and Midorikawa [20].

In the capacity spectrum method, building response during ground shaking is estimated from an intersection of building capacity curve and demand curve. The capacity curve needs to be obtained for each building type. First, buildings in Metro Manila are classified considering the structural type, the number of stories and the design vintage as shown in Table 1. The structural types are classified into three major categories: CHB (Concrete hollow block building), C1 (Reinforced-concrete moment frame building) and C2 (Reinforced-concrete shear wall building). CHB buildings

ARTICLE IN PRESS H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

4.2. Building capacity curves derived from experts’ judgments

Table 1 Classification of buildings in Metro Manila Structural types

769

Stories

Design vintage

CHB

Concrete hollow block

1–3

Sub-type 1, 2, 3

C1L C1M C1H

Concrete moment frame

1–3 4–7 8–15

Sub-type 1, 2, 3 Sub-type 1, 2, 3 Sub-type 1, 2, 3

C2H C2V C2E C2S

Concrete shear wall

8–15 16–25 26–35 36

Sub-type Sub-type Sub-type Sub-type

1, 2, 3 1, 2 1, 2 1

Sub-type 1: Constructed after 1992. Sub-type 2: Constructed between 1972 and 1991. Sub-type 3: Constructed before 1971.

are typically single-family or small, multiple-family dwellings that are usually not engineered. Seismic resistance of these buildings depends on mostly on CHB walls, which are usually provided with lintel beams and vertical stiffeners at an average spacing of a few meters. C1 buildings have a frame of reinforced-concrete columns and beams. Lateral loads of these buildings are resisted by beamcolumn frame action. C2 buildings are mostly tall buildings having concrete shear walls that are usually bearing walls as vertical components of the lateral-force-resisting system. Other structural types, such as wooden buildings, bamboo buildings and steel buildings, are existed in Metro Manila. According to the questionnaire to the building officials and the local government engineers in Metro Manila [22], the percentages of other structural types in the city/municipality were estimated at approximately 20%. Since the number of other structural types is limited, all the buildings are classified into the three major building types (CHB, C1 and C2) in this study. The range of stories is classified into six categories: lowrise buildings (1–3 story), mid-rise building (4–7 story) and high-rise buildings (8–15, 16–25, 26–35, 36+ story). The national structural code of the Philippines (NSCP) was firstly established in 1972 [23]. The code has been revised in 1981, 1986, 1992 and 2001 [24–27]. Generally, design base shear coefficients increased in NSCP1981 from NSCP1972, then decreased in NSCP1986. As a result of the lessons learned from the 1990 Luzon earthquake, significant changes in special requirements for earthquake resistant design of RC buildings were formally incorporated in NSCP1992. The design base shear coefficients increased in NSCP2001 for all the building types, especially for buildings located near the fault. The increase of design base shear in NSCP 2001 was mainly motivated by observations in the 1994 Northridge earthquake and the 1995 Kobe earthquake. Considering the period for the revision of the building code, the design vintages of the buildings are classified into three categories: Sub-type 1 (built after 1992), Sub-type 2 (built between 1972 and 1991) and Sub-type 3 (built before 1971).

To construct the capacity curve of each building category, the two-round questionnaire is applied to the experts of structural engineering comprised of the professors and the local engineers in Metro Manila [28]. The responses of the experts are integrated by the Delphi method (e.g., [29]). The Delphi method is based on a structured process for collecting and distilling knowledge from a group of experts by means of a series of questionnaires interspersed with opinion feedback. The method has been also utilized to obtain estimates of the damage due to earthquakes in ATC-13 [30]. In the first round of the questionnaire, 22 experts participated. In the second round, 21 experts joined the survey. Five engineers who participated in the first round were not able to join the second round because of their urgent obligations, and four engineers are added in the second round survey. A total of 26 experts participate the questionnaire. The questionnaire documents with instruction and explanatory notes for seismic capacity of buildings are distributed to the experts by mail in both round surveys. The responses of the experts are gathered also by mail. To make parameters queried in the surveys more relevant, a follow-up workshop among the experts is organized after the second round questionnaire. The capacity curve consists of spectral displacement and acceleration at yield- and ultimate-capacity points. The questionnaires are mainly composed of the questions for six parameters: anticipated natural vibration period of each building type, seismic mass of building, design strength, strength at yield and ultimate point, and ductility. Selfrated experience/knowledge level (Ei) and certainty level (Ci) of each respondent (i) are also asked in the questionnaires. As with the ATC-13 study, the responses are processed by computing a weighting factor, ECi factor, defined as the following equation: EC i ¼

E 4i C i . n P E 4i C i

(1)

i¼1

Here, n in the equation indicates the number of the respondents. Higher EC indicates higher self-evaluation of the response. Fig. 8 illustrates an example of the results in the first round and the second round surveys. The horizontal axes represent EC and the vertical axes represent l the ratio of the ultimate strength to the yield strength for C1L Sub-type 1 building. Solid point indicates the response of each respondent. Solid line and dotted lines show the average of the responses and its standard deviation, respectively. As shown in the first round survey in Fig. 8(a), difference of experience and certainty level between the respondents is not significant since all the EC factors show smaller than 0.15. In the second round survey shown in Fig. 8(b),

ARTICLE IN PRESS 770

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

4

4 C1L (Sub-Type1)

3

λ

C1L (Sub-Type1)

: Ave. : Ave. ± σ

: Ave. : Ave. ± σ

3

2

2

1

1

0

0 0.1

0

0.2

0.3

0.1

0

0.2

EC

0.3

EC

Fig. 8. Comparison of EC factors between first and second round. (a) First round. (b) Second round.

Table 2 Data for capacity curves and fragility curves derived by the Delphi method Type

Sub-type

Capacity curve DY (m)

AY (m/s/s)

Fragility curve DU (m)

AU (m/s/s)

bc

Displacement at damage state (m) Slight

Moderate

Extensive

Complete

CHB

1 2 3

0.002 0.002 0.002

3.82 4.02 4.12

0.010 0.007 0.007

5.00 5.98 5.98

0.005 0.005 0.005

0.007 0.007 0.007

0.018 0.018 0.018

0.045 0.045 0.045

0.7 0.7 0.7

C1L

1 2 3

0.008 0.005 0.005

2.94 2.84 3.04

0.058 0.018 0.014

4.10 4.31 4.21

0.021 0.019 0.019

0.037 0.032 0.030

0.10 0.088 0.075

0.26 0.23 0.19

0.5 0.5 0.5

C1M

1 2 3

0.020 0.021 0.019

1.96 2.74 2.74

0.150 0.083 0.067

2.74 3.92 4.21

0.035 0.035 0.035

0.061 0.061 0.057

0.17 0.17 0.14

0.42 0.42 0.35

0.5 0.6 0.6

C1H

1 2 3

0.064 0.10 0.10

1.57 2.84 3.14

0.54 0.44 0.36

2.01 4.21 4.70

0.054 0.054 0.054

0.11 0.11 0.093

0.32 0.32 0.25

0.86 0.86 0.64

0.4 0.6 0.7

C2H

1 2 3

0.060 0.093 0.08

1.37 2.45 2.25

0.40 0.34 0.24

1.86 3.72 3.43

0.038 0.038 0.038

0.094 0.094 0.079

0.28 0.28 0.22

0.75 0.75 0.56

0.4 0.6 0.5

C2V

1 2

0.13 0.23

0.98 2.45

0.75 0.83

1.57 3.72

0.075 0.075

0.19 0.19

0.56 0.56

1.5 1.5

0.4 0.6

C2E

1 2

0.21 0.38

0.98 2.45

1.40 1.30

1.47 3.33

0.11 0.11

0.28 0.28

0.54 0.84

2.2 2.25

0.4 0.6

C2S

1

0.39

1.18

2.80

1.57

0.17

0.43

1.29

3.4

0.6

DY: displacement at yield point (m), DU: displacement at ultimate point (m), AY: acceleration at yield point (m/s/s), AU: acceleration at ultimate point (m/s/s).

however, EC factors of some respondents show higher than 0.15. It indicates that the number of respondents who evaluate their certainty level in the second round higher than in the first round is increased. Besides, the standard deviation in the second round is declined to about 0.3 while that in the first round is about 0.5. It means that the responses for the parameter are converged with approximately 1.5 by the opinion feedback. Similar convergence is also observed in other parameters. The spectral values at

the yield and ultimate points for each building category are determined from the median values of the responses in the second round survey. Fig. 8 illustrates the derived capacity curves of the building types highlighted in bold face type in Table 1. Table 2 shows the spectral displacements and accelerations at yield and ultimate points for all the capacity curves derived in this study. Fragility curve is a probability function of being in, or exceeding, a damage state for a given spectral displacement.

ARTICLE IN PRESS 771

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

The fragility curve is developed from the result of the second round survey based on following equation:

4.3. Comparison with pushover analysis

# ! "$ 1 Sd P½dsjS d # ¼ F ln , bc S¯ d;ds

In order to validate the capacities derived from the Delphi method, they are compared with result of the pushover analysis [32] for typical buildings in Metro Manila. Analytical values for the capacity of a structure can be obtained from the pushover analysis. The pushover analysis is applied to two-story RC building (C1L Subtype 1) and 10-story RC building (C1H Sub-type 3). Fig. 10 illustrates the frame geometry of the two-story and 10-story building. The two-story and the 10-story buildings represent a typical school and residential building, respectively. In the pushover analysis, the sizes and the reinforcements of the members of the frame are determined based on the drawings of the buildings. The material models such as shear-strain relationships for concrete and reinforcement steel are defined basically based on the design practice in the Philippines [33,34]. The distributions of lateral loading assumed in the analysis are based on fundamental mode shape of the frames. Fig. 11 shows the building capacity curves derived from the pushover analysis with the capacity curves derived from the Delphi method. For the C1L building, the capacity displacement by the Delphi method is smaller than that by

(2)

where P[ds|Sd] is the probability of a particular damage state, ds (slight, moderate, extensive and complete), at the given spectral displacement, Sd. S¯ d;ds and bc are the median value and its standard deviation of spectral displacement at which the building reaches the damage state. F is the standard normal cumulative distribution function. Here, S¯ d;ds is expressed as multiplication of drift ratio and building height. Since the drift ratio at a damage state of buildings in Metro Manila is poorly examined, the drift ratio of the nearest building type in HAZUS is used in this study considering the structural type, building height and design level [31]. bc is expressed as square root of sum of squares of the standard deviations derived from all the answers in the second round survey. Fig. 9 illustrates the constructed fragility curves of the low-rise building (CHB and C1L), the mid-rise building (C1M) and the high-rise building (C1H). Table 2 also shows the displacements at the damage states and bc for all the building categories.

10

10

10

CHB (Sub-Type3)

C1M (Sub-Type3)

C2V (Sub-Type1)

C1L (Sub-Type3)

C1H (Sub-Type2)

C2E (Sub-Type1)

0

0.01

0.02 SD (m)

0.03

0.04

SA (m/s/s)

SA (m/s/s)

5

5

0

0.1

0.2 0.3 SD (m)

0.4

0.5

5

0

1

2 SD (m)

Fig. 9. Building capacity curves derived from the Delphi method. Low-rise, Mid- and high-rise and high-rise.

1 Damage Ratio

SA (m/s/s)

C2S (Sub-Type1)

0.8 0.6 0.4 0.2 0

0.1

0.2

0.3 0

SD (m) Damage state :

0.1

0.2

0.3 0

SD (m) : Slight

0.1

0.2

SD (m) : Moderate

: Extensive

0.3 0

0.1

0.2

0.3

SD (m) : Complete

Fig. 10. Fragility curves derived from the Delphi method. CHB: Sub-Type3, C1L: Sub-Type3, C1M: Sub-Type3 and C1H: Sub-Type2.

3

ARTICLE IN PRESS 772

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

the pushover analysis, and the capacity accelerations (strengths) are comparable or smaller. Here, the capacity of C1L building by the Delphi method would represent standard residential/commercial buildings because the number of residential/commercial low-rise buildings is predominant in the urban area. As described before, the capacity by the pushover analysis represents a typical school building. The difference between the seismic capacities is caused because public buildings such as school generally would have higher potential to resist for seismic loading than residential/commercial buildings. For the C1H building, on the contrary, the capacity accelerations by the Delphi method are little higher than those by the pushover analysis. Only the lateral load bearing elements such as columns and beams are modeled in the pushover analysis. The actual high-rise building,

however, is likely stronger than the result of this analysis because non-structural elements such as partition walls provide additional strength in actual high-rise building. It indicates that the capacity curves of actual building would correspond better with the curves by the Delphi method. Although the number of the examined cases is limited, the capacity curves derived by the Delphi method are consistent with those by the pushover analysis. The capacity curves derived by the Delphi method are compared also with result of static lateral loading experiment for existing buildings in Metro Manila [35]. According to the force–displacement curve obtained from the experiment for an existing two-story CHB building, the displacements at the yield and ultimate points were approximately 0.004 and 0.006 m, respectively. As shown in Fig. 12 and Table 2, the displacements at the yield and ultimate points of the CHB building are estimated at about 0.002 and 0.007–0.01 m, respectively. The capacity of the CHB building derived from the Delphi method shows good agreement with that of the actual building. 5. Building damage estimation 5.1. Selection of structural type and design vintage

31.0 m

3.2 m

3.0 m

4.45 m

4.0 m 2.5 m

7.0 m

3.8 m 3.8 m

7.6 m 22.8 m

8

8

6

6 SA (m/s/s)

SA (m/s/s)

Fig. 11. Frame geometry of 2-story and 10-story buildings for pushover analysis. (a) 2-story building. (b) 10-story building.

The building damage due to the scenario earthquake is estimated by using the derived capacity curves, the fragility curves, the simulated ground motion and the building inventory data. The number of damaged buildings of the low-rise (CHB and C1L), the mid-rise (C1M), and the highrise (C1H) buildings are computed by multiplying the distribution of the damage ratio in each damage state by the building inventory data. The structural types of both CHB and C1L are contained in the low-rise buildings in Metro Manila. Besides, the strengths of the buildings vary in each design vintage. As described before, the building population of the categories needs to be approximately estimated although

4

2

0

4

2

0.05

0.1 SD (m)

0.15

0

0.1

0.2 0.3 SD (m)

0.4

Push-Over Analysis (Sub-Type1)

Push-Over Analysis (Sub-Type3)

Delphi Method (Sub-Type1)

Delphi Method (Sub-Type3)

0.5

Fig. 12. Comparison of capacity curves derived from the push-over analysis and the Delphi method. (a) C1L Building. (b) C1H Building.

ARTICLE IN PRESS 773

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

the ratios between the number of CHB and that of C1L in each region are approximately estimated at 2:1, 1:2 or 0:3 as shown in the table. As shown in Fig. 1, the urbanized areas had covered approximately 60% of Metro Manila by 1975 including major residential and commercial zones. It indicates that Sub-type 3 design vintage is predominant for the low-rise and mid-rise buildings. Therefore, Sub-type 3 is applied for CHB, C1L and C1M buildings in the damage estimation. Since most of the high-rise buildings would be rather newer than the low-rise and mid-rise buildings, the Sub-type 2 is adopted for C1H buildings in the estimation.

the structural type and the design vintage of each building are not included in the inventory data. The building population in Metro Manila has been investigated by the questionnaires for the building officials and the local government engineers [22]. In the survey, the number of buildings for each building type in each city/ municipality was approximately estimated by the building officials and engineers. Based on the survey, the relationship between the percentages of CHB and C1L buildings in each city/municipality is broadly classified into three categories as shown in Table 3. In the region A such as Caloocan, Valenzuela and so on, the number of CHB is dominant compared with that of C1L. On the contrary, the number of C1L is dominant in the region B such as Marikina, Makati. In the region C, almost all the low-rise buildings consist of C1L. Based on the result of the survey,

5.2. Results of building damage estimation In order to examine effects of the region-specific building performance, the damage estimation of this study is compared against that with capacity curves of nearest building types in HAZUS. To compare with the damage of CHB buildings, URML buildings in HAZUS is used because the structural type almost corresponds with CHB. Low-code is adopted for URML, C1L and C1M buildings in HAZUS because Sub-type 3 buildings in Metro Manila would not be fully engineered. Moderate-code is applied to C1H buildings in HAZUS since Sub-type 2 C1H buildings would have a certain level of resistance for seismic loading. Fig. 13 shows the comparison of the capacity curves derived from the Delphi method and the curves of HAZUS used in the damage estimation. The yield and ultimate

Table 3 Approximately estimated ratio of number of CHB buildings and that of C1L buildings in each city/municipality Region

A

B C

City/municipality

Ratio

Caloocan, Valenzuela, Quezon, Navotas, San Juan, Mandaluyong, Manila, Pasig, Pasay, Pateros, Paranaque, Muntinlupa Marikina, Makati, Taguig, Las Pinas Malabon

CHB

C1L

2

1

1 0

2 3

SA (m/s/s)

10 CHB (Sub-Type3)

C1L (Sub-Type3)

URML (Low-code)

C1L (Low-code)

5

0

0.04 SD (m)

0.08

0

0.04 SD (m)

0.08

SA (m/s/s)

10 C1M (Sub-Type3)

C1H (Sub-Type2)

C1M (Low-code)

C1H (Moderate-code)

5

0

0.05 SD (m)

0.1

0

0.3

0.6

SD (m)

Fig. 13. Comparison of capacity curves by the Delphi method and HAZUS used in the damage estimation. Low-rise, low-rise, mid-rise and high-rise.

ARTICLE IN PRESS 774

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

Fig. 14. Building distribution and distribution of damaged buildings. (a) Building distribution. (b) Extensive or complete damage (with capacity curves of this study). (c) Extensive or complete damage (with capacity curves of HAZUS). (d) Building distribution. (e) Extensive or complete damage (with capacity curves of this study). (f) Extensive or complete damage (with capacity curves of HAZUS). (g) Building distribution. (h) Moderate damage (with capacity curves of this study). (i) Moderate damage (with capacity curves of HAZUS).

ARTICLE IN PRESS 775

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777 Table 4 Comparison of number of damaged buildings Building type

Low-rise (1–3 story)

Mid-rise (4–7 story)

High-rise (8–15 story)

Damage state

(a) Estimation with capacity curves of this study

(b) Estimation with capacity curves of HAZUS

No. of damaged buildings

Ratio (%)

No. of damaged buildings

Ratio (%)

Complete Extensive Moderate Slight Total

114,900 66,700 123,300 86,900 391,800

9.0 5.2 9.6 6.8 30.6

295,800 245,700 235,000 161,500 938,000

23.1 19.2 18.3 12.6 73.2

Total no. of buildings

1,281,400

Complete Extensive Moderate Slight Total

240 407 413 311 1371

8.4 14.2 14.4 10.8 47.8

634 918 927 219 2698

22.1 32.0 32.3 7.6 94.0

Total no. of buildings

2869

–

2869

–

Complete Extensive Moderate Slight Total

5 91 452 160 708

0.6 11.2 55.7 19.7 87.2

14 153 363 147 677

1.7 18.8 44.7 18.1 83.4

Total no. of buildings

812

–

812

–

strengths (accelerations) by the Delphi method are higher than those of HAZUS in all the types. On the other hand, the displacement of the ultimate point by the Delphi method show smaller than those of HAZUS except for the C1H building, indicating the ductility of the buildings in Metro Manila is lower than that in the US. Fig. 14 shows the distribution of the damaged buildings due to the scenario earthquake based on the capacity curves developed in this study and those of HAZUS. Fig. 14(a), (d) and (g) shows the distributions of the lowrise, mid-rise and high-rise buildings in the inventory data, respectively. Fig. 14(b) and (c) shows the distribution of the completely or extensively damaged low-rise buildings. Fig. 14(e) and (f) shows the distribution of the completely or extensively damaged mid-rise buildings. Fig. 14(h) and (i) shows the distribution of the moderately damaged high-rise buildings. Table 4 shows the number of the damaged buildings and the damage ratio at each damage state. The damage distribution of the low-rise buildings in this study is significantly different from that with the capacity curves of HAZUS. In this study, the damage is concentrated only in the soft soil areas such as the coastal lowland and the Marikina valley. In the estimation with the capacity curves of HAZUS, on the contrary, the severe damage is distributed to the whole area of Metro Manila. According to the number of damaged buildings shown in Table 4, the number of damaged buildings in this study is larger than that in the other estimation. As shown in Fig. 13, the low strength of the capacity in HAZUS causes

–

1,281,400

–

the severe damage not only in the soft soil area but also in the stiff soil area such as the central plateau. This trend is also observed in the damage distribution of the mid-rise buildings as illustrated in Fig. 14(e) and (f). Most of the high-rise buildings would suffer moderate damage but not severe damage. One of the reasons is the spectral characteristic of the ground motion. The magnitude of the scenario earthquake Mw 6.7 is not large enough to generate a strong ground motion with long period more than several seconds, which contributes to the response of higher buildings. As shown in Fig. 14(h) and (i), the significant difference between the distributions of the moderately damaged high-rise buildings is not observed in the estimations. As shown in Table 4, the number of the completely or extensively damaged buildings in the estimation with the capacity curves of HAZUS is larger than that in this study since the capacity strength in HAZUS is rather small. 6. Conclusions The seismic performance of the buildings in Metro Manila, Philippines is evaluated by integrating the local experts’ judgments for the building damage estimation. First, the buildings are classified into 20 categories according to the structural type, the number of stories and the design vintage. The questionnaire is applied to the local experts in Metro Manila to integrate the opinions of the experts by the Delphi method. The building capacity curve and the fragility curve for each building category are

ARTICLE IN PRESS 776

H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777

developed from the result of the questionnaires. The derived capacity curves are consistent with the result of the pushover analysis. It indicates that the integration of the experts’ opinions provide the reliable seismic performance for the local building. The ground motions due to a scenario earthquake are computed using the simulation technique based on the underground structure model. The capacity spectrum method is applied to estimate the building response due to the simulated ground motion. The damage ratios are calculated from the fragility curves and the building responses. The distributions of the damaged buildings are estimated by multiplying the damage ratios and the building inventory data. In the estimation of this study for the low-rise and midrise buildings, the severely damaged buildings are mainly concentrated in the soft soil areas such as the coastal lowlands and the Marikina valley. In the estimation with the capacity curves of HAZUS, on the contrary, the severe damage is obtained not only in the soft soil areas but also in the stiff soil areas such as the central plateau. The differences of the damage distributions are caused by the capacity curves used in the estimations. These results indicate the importance of the evaluation of the regionspecific building performance for the reliable building damage estimation. Acknowledgments This study was done as a part of Development Earthquake and Tsunami Disaster Mitigation Technologies and Their Integration for the Asia-Pacific Region (EqTAP) Project sponsored by MEXT (Ministry of Education, Culture, Sports, Science and Technology) of Japan. References [1] Federal Emergency Management Agency (FEMA). HAZUS99 SR2 technical manual. Washington, DC, 1999. [2] United Nations Initiative Towards Earthquake Safe Cities. Risk assessment tools for diagnosis of urban areas against seismic disasters (RADIUS). International Decade for Natural Disaster Reduction, 2000. [3] GeoHazards International and United Nations Center for Regional Development (UNCRD) Disaster Management Planning Hyogo Office. Global earthquake safety initiative (GESI). Pilot project. Final report, 2001. [4] Midorikawa S, Fujimoto K, Arai J. Preliminary assessment of building damage due to a scenario earthquake in Metro Manila, Philippines. Paper no. LE-1f. In: Proceedings of the 7th US national conference on earthquake engineering; 2002. [5] Hasegawa K, Hayashi H, Topping K, Maki N, Tatsuki S, Banba M, et al. Development of participatory seismic risk assessment procedure that reflects community needs: a case report from Marikina City, Metro Manila, Philippines. Paper no. 2256. In: Proceedings of the 13th world conference on earthquake engineering; 2004. [6] Japan International Cooperation Agency (JICA), Metropolitan Manila Development Authority (MMDA), Philippines Institute of Volcanology and Seismology (PHIVOLCS). Earthquake impact reduction study for Metropolitan Manila, Republic of the Philippines. Final report, 2004.

[7] Applied Technical Council (ATC). Seismic evaluation and retrofit of concrete buildings. Redwood City: ATC-40; 1996. [8] Chopra AK, Goel RK. Capacity-demand-diagram methods based on inelastic design spectrum. Earthquake Spectra 1999;15(4):637–56. [9] Doi K, Kim K. Role of strategic modeling approach in the formulation of the framework of sustainable metropolitan policies. In: Proceedings of 1st workshop on environment conservation of Metro Manila; 1998. p. 13–25. [10] Matsuda I, Enomoto T, Banganan EL, Narag IC. Regional division of Metro Manila on the basis of geological and geomorphological conditions. Bull Inst Sci Technol, Kanto Gakuin Univ 1998;25: 101–12. [11] Daligdig JA, Besana GM. Seismological hazards in Metro Manila. In: Proceedings of the natural disaster prevention and mitigation in metropolitan Manila area. UNCHS (Habitat) Project INT/90/70; 1993. p. 9–41. [12] Nelson A, Personius SF, Rimando RE, Punongbayan RS, Tungol N, Mirabueno H, et al. Multiple large earthquakes in the past 1500 years on a fault in Metropolitan Manila, The Philippines. Bull Seism Soc Am 2000;90:73–85. [13] Kamae K, Irikura K, Pitarka A. A technique for simulating strong ground motion using hybrid Green’s function. Bull Seism Soc Am 1998;88:357–67. [14] Irikura K, Kamae K. Strong ground motions during the 1948 Fukui earthquake—estimation of broad-band ground motion using a hybrid simulation technique. J Seisimol Soc Jpn 1999;52:129–50 (in Japanese with English abstract). [15] Schnabel PB, Lysmer J, Seed HB. SHAKE: a computer program for earthquake response analysis of horizontally layered sites. Report no. EERC 72-12. Earthquake Engineering Research Center, University of California, Berkeley; 1972. [16] Irikura K, Miyake H. Prediction of strong ground motions for scenario earthquakes. J Geol 2001;110(6):849–75 (in Japanese with English abstract). [17] Yamanaka H, Takezono M, Ogata Y, Eto K, Banganan EL, Narag IC, et al. Estimation of S-wave velocity profiles in Metro Manila from long-period microtremor array measurement. In: Proceedings of the international workshop on the integration of data for seismic disaster mitigation in Metro Manila; 2000. p. 13–29. [18] Imazu M, Fukutake K. Dynamic shear modulus and damping of gravel material. In: Proceedings of 21th Japan national conference on soil mechanics and foundations engineering; 1986. p. 509–12 (in Japanese). [19] Japan International Cooperation Agency (JICA). 1/10,000 scale urban contoured map. Report on establishment of graphic information base project of the national capital region, Republic of the Philippines, 1989. [20] Miura H, Midorikawa S. Updating GIS building inventory data using high-resolution satellite images for earthquake damage assessment. Earthquake Spectra 2006;22(4):151–68. [21] Yamazaki F, Mitomi H, Yusuf Y, Matsuoka M. Urban classification of Metro Manila for seismic risk assessment using satellite images. In: Proceedings of 5th multi-lateral workshop on development of earthquake and Tsunami disaster mitigation technologies and their integration for the Asia-Pacific Region; 2003. p. 1–8 (CD-ROM). [22] Vibrametrics Inc. Survey among building officials and local government engineers on population distribution in Metro Manila. EqTAP Metro Manila case study, Final report, 2003. [23] Association of Structural Engineers of the Philippines. National structural code of the Philippines (NSCP). First edition, 1972. [24] Association of Structural Engineers of the Philippines. National structural code of the Philippines (NSCP). Second edition, 1981. [25] Association of Structural Engineers of the Philippines. National structural code of the Philippines (NSCP). Third edition, 1986. [26] Association of Structural Engineers of the Philippines. National structural code of the Philippines (NSCP). Fourth edition, 1992. [27] Association of Structural Engineers of the Philippines. National structural code of the Philippines (NSCP). Fifth edition, 2001.

ARTICLE IN PRESS H. Miura et al. / Soil Dynamics and Earthquake Engineering 28 (2008) 764–777 [28] Vibrametrics Inc. Survey of experts’ judgment on earthquake capacity of selected building types in Metro Manila. EqTAP Metro Manila case study, Final report, 2003. [29] Linstone HA, Turoff M. The Delphi method: techniques and application. Addison Wesley Longman Publishing Co.; 2002. [30] Applied Technical Council (ATC). Earthquake damage evaluation data for California. Report ATC-13. Redwood City, 1985. [31] Vibrametrics Inc. Development of fragility curves for selected building types. EqTAP Metro Manila case study, Final report, 2003. [32] Seismo Struct. Computer program for static and dynamic nonlinear analysis of framed structure. /http://www.seismosoft.comS, 2002.

777

[33] Vibrametrics Inc. Pushover analysis of an internal frame of a standard 2-story school building. Final report of EqTAP Metro Manila case study, 2003. [34] Vibrametrics Inc. Pushover analysis of an internal frame of a 10-story sub-type3 (oldest) C1H building. Final report of EqTAP Metro Manila case study, 2003. [35] Suzuki S, Tanaka S, Horie K, Maki N, Mizukoshi K, Ohmori T, et al. A study on dynamic parameter properties and earthquake response characteristics of non-engineered housings in Marikina city, Philippines. J Soc Safety Sci 2005;7:1–8 (in Japanese with English abstract).