Asia Conference on Earthquake Engineering Technical Proceedings

SEISMIC VULNERABILITY EVALUATION OF URBAN STRUCTURES IN METRO

PART 1: GENERATION OF STRONG GROUND MOTION FROM A SCENARIO EARTHQUAKE OF THE WEST VALLEY FAULT

Nelson Pulido, Bartolome Bautista, Leonila Bautista, Hisakazu Sakai, Hiroshi Arai and Tetsuo Kubo

ABSTRACT : The West Valley Fault System, which crosses from the south to the north, poses a very important seismic hazard to the city. We performed a broadband frequency strong ground motion simulation in the Metro Manila region, for an outcropping engineering bedrock site condition, based on several fault rupture scenarios and a multi- asperity model. We considered two possible scenarios of fault rupture within the West Valley fault system. The first scenario by assuming a total fault rupture length of 63km (Mw=6.76) and the second scenario we assumed a shorter fault rupture length of 34km (Mw=6.47). We calculated the ground motion at two sites located in the southern part of the fault, in the Muntinlupa city, by constraining the structure velocity model for our calculations from results of microtremor array measurements at both sites. The bedrock ground motion obtained in this study is used as input motion to calculate the nonlinear soil response at the Muntinlupa sites, and subsequently to study the seismic performance of existing school buildings in Metro Manila.

KEYWORDS: Seismic Vulnerability, West Valley Fault, Earthquake Scenario, Strong Ground Motion, Muntinlupa city.

1 INTRODUCTION

The Valley is a pull-apart basin bounded by escarpments of the East and West Valley fault systems (Figure 1). Right-lateral movement of the West and East Valley faults suggests clockwise rotation of the fault block underlying the valley (Nelson et. al. 2000). Landforms indicative of latest Pleistocene and Holocene strike-slip faulting, such as offset stream terraces and shutter ridges, are widespread along the traces of the East and West Valley fault systems north of the River. Landforms suggesting repeated rupture of the West Valley fault have not been identified South of the (Punongbayan et. al., 1996). Based on these observations Nelson et al. (2000) concluded that the chance of an earthquake larger than M 7 on faults of the Marikina Valley seems small.

227 Asia Conference on Earthquake Engineering Technical Proceedings

Table 1. Fault parameters for Scenarios 1 and 2

In the present study we considered two possible scenarios of fault rupture within the West Valley fault. The first one by assuming a total fault rupture length of 63km (Mw=6.76) from the northern splay of the West Valley fault up to slightly South of Muntinlupa city (scenario 1). The second scenario we assumed a shorter fault rupture length of 34km (Mw=6.47), not going southward of the Pasig River (Table 1). The above observations suggest that scenario 2, although less critical than scenario 1 from the point of view of the radiated strong ground motion, could have a larger probability to occur. We calculated an outcrop engineering bedrock ground motion at two sites located in the southern part of the fault, at the Muntinlupa city by using a broadband frequency simulation methodology (Pulido at al. 2003). The structure velocity model for our calculations was constrained by the results of microtremor array measurements performed at both sites.

2 METHODOLOGY

We estimate the broadband frequency (0.1Hz to 20 Hz) near fault ground motion from a hybrid simulation technique that combines deterministic wave propagation modelling for the low frequencies with a stochastic technique for the high frequencies. The basic idea of the simulation methodology is to evaluate the strong ground motion radiated from a finite fault source model composed of asperities embedded in a flat layered velocity structure. The ground motion at a particular site is obtained from the contributions of the seismic radiation from all the asperities in the fault plane that are assumed to have a finite area.

2.1 Low Frequency Ground motion To calculate low frequency ground motion (0.1 to 1.0Hz) we subdivide the asperity into several subfaults or point sources and simply add the time delayed ground motion from them by applying a constant rupture velocity. The seismogram from each point source is obtained numerically by the Discrete Wave Number method of Bouchon (1981), which computes the wave propagation in a flat-layered crustal velocity structure, for a particular focal mechanism and source moment function. The point source moment function is defined as a smoothed ramp as follows:

⎛ ⎛ 4*(t − τ ) ⎞⎞ M 0 ⎜ ⎜ 2 ⎟⎟ (1) M (t) = *⎜1+ tanh ⎟ 2 ⎜ ⎜ τ ⎟⎟ ⎝ ⎝ ⎠⎠

228 Asia Conference on Earthquake Engineering Technical Proceedings

Table 2. Asperity parameters for Scenarios 1 and 2

Note: parameters of scenario 2 in parenthesis when different to those of scenario 1.

where Mo is the point source seismic moment, t is the rupture time, and τ is the asperity rise time.

2.2 High Frequency Ground Motion High frequency ground motion (1 to 20 Hz) is calculated from a finite asperity as previously but the ground motion radiated from the point sources is obtained by using the stochastic approach of Boore (1983) modified by introducing a frequency dependent radiation pattern model (Pulido et. al. 2003). The procedure of summation of the point source contributions differ from the one applied for the low frequencies; for high frequencies the summation is obtained by applying the empirical Green’s function method proposed by Irikura (1986), which is very efficient for the radiation of high frequency ground motion from finite faults.

The frequency dependent radiation pattern Rpi(θ,φ, f) is introduced in order to account for the effect of the pattern on intermediate frequency ground motions (1 to 3 Hz). The i component of acceleration Fourier spectra for a point source is obtained as follows:

R (θ ,φ, f )M S( f , f )F e −πfR Q( f )β P( f , f ) A ( f ) = pi 0 c s max (2) i 4πρβ 3 R

1 3 ⎛∆σ a ⎞ (3) fc = 49000β ⎜ ⎟ ⎝ M o ⎠

229 Asia Conference on Earthquake Engineering Technical Proceedings

Figure 1. Landsat image of the Metro Manila region showning the West and East Valley fault systems.

1 (4) P( f , fmax ) = ⎛ f ⎞ ⎜1+ ⎟ ⎝ fmax ⎠ where Mo is the point source seismic moment (in Nm in eq. 3), S( f, fc ) is the omega square source model (Brune 1970) with corner frequency fc (eq. 3), ∆σ is the point source stress drop (in Mpa), Fs is the amplification factor due to the free surface, R is the station-point source distance and ρ and β are the average density and S-wave velocity of the media. The exponential term accounts for the regional attenuation of Q which increases with the frequency as a power law of the form af b, where a and b determine the strength of attenuation. P is the high-frequency cut-off of the point-source acceleration spectra for frequencies above fmax (eq. 4). More details of the simulation technique are explained in Pulido et al. (2003).

3 FAULT AND ASPERITY PARAMETERS

230 Asia Conference on Earthquake Engineering Technical Proceedings

Figure 2. Velocity model at sites A and B. We assume the same model at both sites for depths below 4km. For shallower depths we use results from microtremors array measurements at every site.

To obtain the fault seismic moment and asperity area we used the empirical scalings of Somerville (1999). The asperities seismic moment, stress drops and rise time were calculated using the asperity model of Das and Kostrov (1986) and Boatwright (1987). The rupture velocity was assumed as 90% of the average S-wave velocity, and allowed to vary randomly in a prescribed interval (as for the rise time), to improve the high frequency generation (Table 2). The velocity model was obtained by overlapping the crustal velocity model used for routine epicenter locations at PHIVOLCS, with the shallow (<2 km) velocity model obtained from microtremor array measurements at sites A and B in Muntilupa (Arai et al. 2003) (Figure 2). The S-wave velocity for the shallower layer of our model, at sites A and B corresponds to an engineering bedrock (Vs=700m/s)(Figure 2). Therefore our simulation of the ground motion will be performed for an engineering bedrock outcropping site. The simulation of the ground motion by considering the non-linear response of soil layers with smaller S-wave velocity is performed in a companion paper (Sakai et. al. 2003). We use a high frequency attenuation Q value obtained from aftershocks of the 1990 earthquake (Kanao et al. 1992). The total asperity area at every fault segment was calculated from an empirical ratio of the total asperity area (Sa) to the fault rupture area (S) (Somerville et al. 1999):

S a S = 0.22 (5) Results from a dynamic model of the rupture of a circular fault (with radius R ) with an asperity (with radius r ) at its center (Das and Kostrov, 1986, Boatwright 1987), combined with equation (5), suggest that the ratio between the asperity stress drop to the fault average stress drop is equal to 0.47. If we assume an asperity model the total seismic moment can be calculated as follows (Boatwright 1987): 16 M a = ∆σr 2R (6) o 7

231 Asia Conference on Earthquake Engineering Technical Proceedings

Figure 3. PGV distribution from two scenario earthquakes of the West Valley fault (scenarios 1 and 2). We assume two fault segments and equal number of asperities depicted by thin and thick black lines. Hypocenter is shown by a star. Simulation is performed every 5 km by assuming an S-wave velocity of 800 m/s.

Replacing (5) into (6) we obtain the stress drop of asperities: M a ∆σ = 11.07 o (7) a S 3 2

Subsequently the seismic moment from the asperities is calculated by using the Brune’s crack model (1970):

3 2 16 ⎛ S ⎞ a M asp = ∆σ a ⎜ ⎟ (8) 7 ⎝ π ⎠

4 RESULTS

Scenario 1 produced the most critical ground motion at Muntinlupa among the two scenarios

232 Asia Conference on Earthquake Engineering Technical Proceedings

Figure 4. EW component of Acceleration and Velocity waveforms and response spectra at sites A and B, for scenario 1.

Figure 5. NS component of Acceleration and Velocity waveforms and response spectra at sites A and B, for scenario 1.

233 Asia Conference on Earthquake Engineering Technical Proceedings

Figure 6. Comparison of empirical PGA and PGV attenuation laws, with the simulated values from scenarios 1 and 2. Simulation is performed for Vs=800 m/s. considered. As shown in Figure 3, a very important directivity effect is observed towards Muntinlupa, for the case of a unilateral southward rupture propagation. The location of the asperities plays a very important role for the estimation of peak ground motions. In the present simulations the asperity locations were choosen in order to generate the ground motion with the largest possible PGV at Muntinlupa (sites A and B). The ground motion from scenario earthquake 2, although less critical than scenario earthquake 1 for the Muntinlupa city (sites A and B, Figure 3), has a larger probability to occur, as might be inferred from observations by Punonbayan et. al. (1996) of low activity of the West Valley fault south of the Pasig river. In Figure 4 we can observe that large directivity pulses of approximately two seconds are generated in the EW velocity component of ground motion. On the other hand the NS ground motion is much smaller than the EW ground motion at sites A and B. This particular feature is explained by the fact that the EW component of ground motion at sites A and B almost corresponds to a maximum in the nodal plane of SH-waves (maximum radiation) in the fault- perpendicular component. On the other hand the small PGV values in the NS component are

234 Asia Conference on Earthquake Engineering Technical Proceedings associated with a minimum in the nodal plane of SH-waves for the fault-parallel component for sites close to the fault (Figure 5). The velocity and acceleration response spectra also show clear peaks around 2 sec for the EW component (Figure 4). Site B has in general larger peak ground motions compared with site A, because of the difference in the velocity model. The shallower layer of Vs=700m/s and 400m thickness at site B is responsible of producing larger ground motions around 2 sec compared with site A. The largest PGA and PGV reach values of around 700 cm/s2 and 100 cm/s at site B and 400 cm/s2 and 66 cm/s at site A. We obtained a good agreement between our simulations for scenarios 1 and 2 and some PGA and PGV empirical attenuation laws for a similar fault type (strike-slip), and site conditions (Vs = 800 m/s) (Figure 6).

5 CONCLUSIONS

We have calculated the strong ground motion at an outcrop engineering bedrock soil condition at two sites located in the Muntinlupa city, south of Metro Manila. The ground motion was obtained for two possible scenario earthquakes within the West Valley fault system. We obtained that a hypothetical M6.8 from the West Valley fault, with an unilateral southward rupture propagation, generates the most critical bedrock ground motion at Muntinlupa city. However this scenario might have a smaller probability to occur compared with other scenarios with smaller fault length with a fault rupture not going southward of the Pasig River. Large directivity pulses around two seconds are observed in the EW component of ground motion for a site located very close to the fault line (within 2 km). For sites located at larger distances from the fault this might not be the case. Empirical attenuation laws for the PGA and PGV for similar fault type and site conditions provide a good constraint on the fault and asperities parameters of our model.

REFERENCES

Arai, H., T. Kubo, 2003. Strong Motion Estimation in Metro Manila () from a Scenario Earthquake of the West Valley Fault (Part 2), Proceedings of the 2nd Annual Meeting of Japan Association for Earthquake Engineering, 152-153.

Boatwright, J., 1988. The Seismic Radiation from Composite Models of Faulting, Bull. Seism. Soc. Am., 78, 2, 489-508.

Bouchon, M., 1981. A simple method to calculate Green’s functions for elastic layered media, Bull. Seism. Soc. Am., 71, 4, 959-971.

235 Asia Conference on Earthquake Engineering Technical Proceedings

Brune, J.N., 1970. Tectonic stress and the spectra of seismic shear waves from earthquakes, J. Geophys. Res., 75, 4997-5009.

Das, S. and B. V. Kostrov, 1986. Fracture of a single asperity on a finite fault: A model for weak earthquakes?, In Das, S. J. Boatwright, and C. Scholz, eds. : Earthquake Source Mechanism, AGU., Washington D. C., 91-96.

Irikura, K., 1986. Prediction of strong acceleration motion using empirical Green’s function, 7th Japan. Earthq. Eng. Symp. 151-156.

Kanao, M., K. Ito, 1992. Attenuation of Coda Waves in the Source Area of the 1990 July 16 Luzon Earthquake, Philippines, Bull. Disas. Prev. Res. Inst., Kyoto University, 42 (2), No. 365.

Nelson, A.R., S. Personius, R. Rimando, R. Punongbayan, N. Tungol, H. Mirabueno, A. Rasdas, 2000. Multiple Large Earthquakes in the Past 1500 Years on a Fault in Metropolitan Manila, the Philippines, Bull. Seism. Soc. Am., 90, 1, 73-85.

Pulido, N. and T. Kubo, 2003. Near-Fault Strong Motion Complexity of the 2000 Tottori Earthquake (Japan) from a Broadband Source Asperity Model, Tectonophysics (in press).

Punongbayan, R. S., J. A. Daligdig, G. M. Besana, N. M. Tuñgol, and R. E. Rimando, 1996, The Marikina Valley fault system: active faulting in the eastern Metro Manila area, report of the Philippine Institute of Volcanology and Seismology, submitted as a journal paper to the Philippine Journal of Volcanology and Seismology, 36pp.

Sakai, H., H. Arai, T. Kubo, B. Bautista, M. L. Bautista, 2004. Seismic Vulnerability Evaluation of Urban Structures in The Metro-Manila, Part 3: Response of Shallow Soil Layers, Proceedings of the 2nd Annual Meeting of Japan Association for Earthquake Engineering, 154-155.

Somerville, P., K. Irikura, R. Graves, S. Sawada, D. Wald, N. Abrahamson, Y. Iwasaki, T. Kagawa, N. Smith and A. Kowada, 1999. Characterizing Crustal Earthquake Slip Models for the Prediction of Strong Ground Motion, Seismological Research Letters., 70, No.1, 59-80.

236 Asia Conference on Earthquake Engineering Technical Proceedings

ABOUT THE AUTHORS

Nelson Pulido, Dr. Sc, is Researcher at the Structural Performance Team, Earthquake Disaster Mitigation Research Center, National Research Institute for Earth Science and Disaster Prevention (EDM), 4F Human Renovation Museum, 1-5-2 Kaigan–dori, Wakihama, Chuo-ku Kobe 651-0073, Japan. (E-mail: [email protected])

Bartolome C. Bautista, Dr. Sc., is Chief Science Research Specialist, Philippine Institute of Volcanology and Seismology, Philippines (E-mail: [email protected])

Maria Leonila P. Bautista, Dr. Sc., is Science Research Specialist, Philippine Institute of Volcanology and Seismology, Philippines (E-mail: [email protected])

Hisakazu Sakai, Dr. Eng., is Researcher at the Structural Performance Team, Earthquake Disaster Mitigation Research Center, National Research Institute for Earth Science and Disaster Prevention, Japan. (E-mail: [email protected])

Hiroshi Arai, Dr. Eng., is Deputy Team Leader, Structural Performance Team, Earthquake Disaster Mitigation Research Center, National Research Institute for Earth Science and Disaster Prevention, Japan (E-mail: [email protected])

Tetsuo Kubo, Dr. Eng., is the Team Leader of the Structural Performance Team at EDM, as well as Professor of the Department of Architecture, Graduate School of Engineering, University of Tokyo, Japan (E-mail: [email protected])

ACKNOWLEDGEMENTS We would like to thank Mr. Ishmael C. Narag and Ms. Esmeralda C. Banganan from the Philippine Institute of Volcanology and Seismology (PHIVOLCS) for their invaluable help during the array microtremors observations at Muntinlupa. Also we would like to express our gratitude to Engineer Oscar Oquendo, officer of the Disaster Action of Muntinlupa city for his help in local coordination of the drilling sites.

237