Third International Symposium on the Effects of Surface Geology on Seismic Motion Grenoble, France, 30 August - 1 September 2006 Paper Number: 105 BROADBAND STRONG MOTION SIMULATION OF THE 2004 NIIGATA- KEN CHUETSU EARTHQUAKE: SOURCE AND SITE EFFECTS Nelson PULIDO 1, Masashi MATSUOKA 2 1 National Research Institute for Earth Science and Disaster Prevention (NIED), Tsukuba, Japan 2 Earthquake Disaster Mitigation Research Center (EDM), Kobe, Japan ABSTRACT - In this study we analyzed the contribution of fault rupture process and site effects to the extremely large recorded near-fault ground motions and widespread road damage during the 2004/10/23, M6.8 Niigataken Chuetsu earthquake. For that purpose we calculate seismic-bedrock broadband ground motions every 250m within the near-fault region, and incorporate site amplifications to simulated PGV values, from a 7.5-arc second (250m) Vs30 map of the Niigata region. We derive the expressions relating average S-wave velocity of the shallow soil to PGV amplification values with respect to a seismic bedrock condition. These relationships are useful to predict PGV when a detailed 3D velocity structure is not available. Our simulated ground motions are in good agreement with observed strong motion data and damage distribution to roads during the Chuetsu earthquake. Predicted PGV values at a dense grid cell provide an useful information at localities with no strong motion records. 1. Introduction The 2004/10/23, M6.8 Niigata-ken Chuetsu earthquake is the largest damaging earthquake in Japan since the 1995 Kobe earthquake. The earthquake recorded accelerations above 1.7g and velocities of 130 kines. The Chuetsu earthquake and its strong aftershocks sequence produced an extensive damage to roads around the epicentral region. In order to allow a detailed comparison between damage to roads and ground motion indices it is necessary to calculate ground motion at a finer spacing than the one provided by observed ground motions (K-NET and KiK-net) and JMA observed instrumental intensities. For that purpose we calculate seismic-bedrock broadband ground motions every 250m within the near-fault region, and incorporate site amplifications to simulated PGV values, from a 7.5-arc second (250m) Vs30 map in the Niigata region. Forward seismic-bedrock ground motions (by a hybrid method that combines wave propagation within a flat-layered crustal model at low frequencies, with a semi-stochastic approach at high frequencies), are calculated from a broadband multi-asperity source model obtained by optimizing the fitting to observed Fourier spectra and acceleration envelopes of near-fault ground motions (Pulido and Kubo, 2004). 657
Figure 1. KiK-net and K-NET stations used to estimate site amplifications. Color scale corresponds to a 7.5 arc-second (250m) map of average S-wave velocity for the upper 30m, within the Niigata region (Wakamatsu et. al., 2005). 2. Site Amplifications within Niigata Prefecture 2.1 Inversion of spectral ratios We derive the expressions relating the average S-wave velocity of the upper 5, 10, 20 and 30m to PGV amplification factors with respect to a seismic bedrock site (Vs=2600 km/s), by using site amplification functions obtained from a two-step spectral inversion technique; we estimated site amplifications at 31 strong motion stations in Figure 1 (23 K-NET and 8 KiK-net), using records from 50 aftershocks (M3 to 5) of the Niigata-ken Chuetsu earthquake. We first estimated relative spectral ratios at every site, including borehole bottom sites, with respect to the surface recordings of a reference borehole site (FKSH07 KiK-net site), and perform the spectral inversion to obtain relative amplification functions. Then we use the theoretical 1-D response of the reference borehole site to obtain corrected amplification functions with respect to an S-wave velocity of 2600 m/s, which is the velocity value at the bottom of the reference borehole site (Figure 2). Our inversion assumes an spherical attenuation (1/R geometrical spreading) and a frequency dependent Q. We obtained a Q value of 63 f 1.14 from inversion (Figure 2). This value is in agreement with other Q values obtained for the region (Jin and Aki, 2006). To check the inverted site effects values we calculate the ratios of surface to bottom site effects at KiK-net sites and compare them with the observed borehole response using all the aftershocks database (Figure 3). We can observe a very good agreement between the ratios from inverted site effects (black lines) and observed boreholes response (gray lines) (Figure 3). 658
Figure 2. Site amplifications at 31 KiK-net and K-NET stations within the Niigata region, estimated from inversion of spectral ratios with respect to a reference borehole site (FKSH07). The site effects has been corrected by the outcrop 1-D response of the reference site (plotted without free surface factor within the Figure) to yield amplification with respect to a S-wave velocity of 2600 m/s (S-wave value at the bottom of the borehole). 659
Figure 3. Ratios of surface to bottom inverted site effects for KiK-Net borehole stations in Niigata prefecture region (black lines). Observed surface to bottom ratios for all aftershocks database are shown in gray lines. Yellow lines are observed averages ratios. Red line depicts the theoretical outcrop response for the reference station (for spectral inversion). Borehole depths are shown within each box (in meters). 2.2 PGV amplification to seismic bedrock and Vs30 Using the reference station-corrected site effects we de-convolved the site amplification functions from all the aftershocks records and calculated the ratio of observed to sitecorrected PGV values. In this way we obtain PGV amplifications with respect to a seismic bedrock (Vs=2600 m/s). These PGV amplifications include the contribution from all frequencies as we used frequency dependent site response functions to obtain them. Finally we calculate the time averaged S-wave velocities for different soil column thickness (5 to 30m) from log information at each site, and derive the best fitting log-linear expressions relating PGV amplification and average S-wave velocities. We obtained that the PGV amplification with respect to the seismic bedrock is better correlated with the average S-wave velocity of the upper 10m of soil (Vs5 and Vs10) than with Vs30 (Figure 4), which is a widely used index to estimate PGV amplifications with respect to an engineering bedrock site (Vs=600m/s). This result implies that most of the amplification due to site effects is experienced in the very shallow layers, as can be clearly observed in sites like K-NET Ojiya. The PGV amplification with respect to a seismic bedrock (Vs=2600m/s) is obtained as follows: log PGV amp = 1.83 0.53 log Vs30 (1) In Figure 4 (Vs30) we can observe that the PGV amplification with respect to a seismic bedrock obtained in this study, is twice the PGV amplification with respect to an engineering bedrock (Vs30=600m/s) (Midorikawa et al., 1994). 660
Figure 4. PGV amplification with respect to a seismic bedrock condition (Vs= 2600m/s) for the average S-wave velocity for different soil column thickness. Correlation values (R) and best-fitting equations are shown inside each panel. Table I. Source Model and attenuation parameters from inversions Parameter Value Stress drop asperity 1 Stress drop asperity 2 Stress drop asperity 3 Stress drop asperity 4 100 bar 271 bar 187 bar 61 bar Figure 5. Multi-asperity model of the 2004 Niigata-ken Chuetsu earthquake. Slip model was obtained from inversion of near-fault strong motions and teleseismic data (Yagi, 2004). f max 5 Hz Q(f ) for Niigata region 63 f 1.14 This factor of 2 represents the contribution of the deep velocity structure to PGV values. In a later section we will use information of a 250m Vs30 map of the Niigata region together with equation 1 to predict the ground motion at a finer grid than observed data (KiK-net, K-NET and JMA stations including municipalities). Despite Vs10 gives a better correlation to PGV amplification than Vs30, there is no information available on a Vs10 map for Japan, and therefore we will use the Vs30 map available for the Niigata region (Wakamatsu et al., 2005). 3. Source Model Estimation We estimated an optimum multi-asperity model of the Niigata-ken Chuetsu earthquake by maximizing the fitting to observed acceleration Fourier spectra and velocity envelopes of near-fault ground motions. For this purpose we use a genetic algorithm scheme combined with a broadband strong motion simulation procedure (Pulido and Kubo, 2004). For the inversion, we selected 10 stations (3 KiK-net, 7 K-NET) within 50 km from the fault, as shown by filled triangles and squares in Figure 6. Before performing the inversion, we deconvolved all observed waveforms to a seismic-bedrock ground condition (Vs=2600), by 661
ESG2006, Grenoble, 30/08-01/09/2006 Figure 6. Simulated PGV values of the 2004 Niigata-ken Chuetsu earthquake (color scale). Black contour lines represent the observed JMA instrumental intensities during the Chuetsu earthquake. The figure inset shows the comparison between simulated and observed PGV values at KiK-net and K-NET sites within the Niigata region. Filled symbols represent stations used for asperity parameters estimation. using the site effects functions obtained in the previous section. We used a slip model obtained from inversion of near-fault strong motion and teleseismic waveforms (Yagi, 2004), as an initial source model (Figure 5). Our optimum source model consists of four asperities the first one centered at the hypocenter (Figure 5). In Table 1 we show the asperity parameters obtained from the inversion. We obtained that asperity 2 has a very large stress drop (271 bar), which makes a significant contribution to the extremely large PGA value of 1.7g, recorded at the Tokamachi K-NET station (NIG021). 4. Strong Motion Simulation within Niigata Prefecture In order to compare damage to civil infrastructures during the 2004 Niigata-ken Chuetsu earthquake, to ground motion indexes, we estimated the ground motion every 250m. Forward strong motions are first calculated at a seismic-bedrock for a 1km grid mesh within the Niigata region, from an optimum multi-asperity model (Figure 5) combined with 662
ESG2006, Grenoble, 30/08-01/09/2006 Figure 7. Simulated PGV values of the 2004 Niigata-ken Chuetsu earthquake (color scale) for a region around the ruptured fault (black box). Open circles show damage points to roads during the Chuetsu earthquake. a broadband strong motion simulation procedure (Pulido and Kubo, 2004). Then we linearly interpolate the seismic bedrock ground motions to a 250m mesh corresponding to the grid cells of a Vs30 map available for the Niigata region (Wakamatsu and Matsuoka 2005). Finally we apply the PGV amplification factors with respect to the seismic bedrock obtained in this study (equation 1), to every interpolated seismic bedrock PGV, and get the ground motion at the surface (Figures 6 and 7). In Figure 6 we show the simulated PGV values for the Niigata region. For comparison we included the observed JMA instrumental intensity values (black contour lines in Figure 6) obtained from JMA intensity recorders located at every municipality (approximately one instrument per municipality). Municipalities boundaries within the Niigata prefecture are depicted by white lines in Figure 6. From the overall picture of the Niigata region we can see that our simulated PGV s follow a similar trend compared to JMA recorded intensities. However, looking at a fine scale we can observe that our simulation is able to provide a detailed picture of peak ground motions at a municipality level, were there are only one or two JMA records available (Figure 6 and Figure 7). In order to evaluate the accuracy of our simulations forward ground motions were calculated at all the K-NET and KiK-net sites within the Niigata prefecture, incorporating PGV amplifications from equation 1, and using actual Vs30 values from log information data at every site. We obtained a good agreement to observed strong ground motions as displayed at the inset in Figure 6. The simulated PGV values for the 10 sites used for estimation of asperity parameters are within a 40% deviation from observed values, as shown by filled circles in Figure 6 inset. We obtained large values of PGV at the western and southern regions of Ojiya city as well as the entire Kawaguchi and Yamakoshimura towns, where simulated PGV s are above 80 kines for large areas (Figure 7). We obtained that asperity 1 has a large contribution to the ground motion at western Ojiya-city and northern Kawaguchi town. 663
ESG2006, Grenoble, 30/08-01/09/2006 Figure 8. Theoretical radiation pattern for SH-waves of the 2004 Niigata-ken Chuetsu earthquake. A star depicts the hypocenter. Large ground motions at Yamakoshimura town are generated by asperity 3, and ground motions at southern Ojiya and southern Kawaguchi are mainly generated by asperity 2 (Figure 7). 4.1 Effect of Radiation Pattern on ground motion Our simulated PGV values show a complex pattern around the hypocenter. We can observe that ground motion is relatively small just above the hypocenter, compared with the adjacent areas around the hypocenter (Figures 6 and 7). This feature may be explained by the radiation pattern of a reverse type earthquake like the Chuetsu earthquake, because for this type of mechanism a minimum nodal plane of SH-waves corresponds approximately to a region above the hypocenter (Figure 8). Radiation pattern of SH-waves within the Niigata region was calculated for a point source located at the hypocenter of the Chuetsu earthquake, by assuming a flat layered velocity structure and tracing the rays from this point source to a set of receivers spaced at 1 km, across the entire Niigata prefecture region (Figure 8). For the simulation of the high frequency component of ground motion the radiation pattern coefficients are gradually smoothed from these theoretical radiation pattern values to an isotropic radiation pattern at frequencies around 5Hz, following Pulido and Kubo (2004). 5. Strong Motion and Damage to Roads In Figure 7 we show the simulated PGV for the epicentral region of the 2004 Chuetsu 664
earthquake. Damage points to major local and prefecture roads (Sakai et al., 2006) are shown by open circles and black lines in Figure 7. We can observe that PGV is above 40 kines for the most of the region (Figure 7). In southern Ojiya, northern Kawaguchi and Yamakoshimura regions, where simulated PGV is above 100 kines, a large concentration of road damage was observed (Sakai et al. 2006). On the other hand simulated PGV values at Kashiwazaki-city, 30 km North-West of the mainshock epicenter, are relatively larger than the surrounding areas. This is also in good agreement with the large road damage concentration at Kashiwazaki-city (Sakai et al., 2006). 6. Discussions and Conclusions We derived expressions relating average S-wave velocity of soil to PGV amplification with respect to a seismic bedrock condition for the Niigata region, by using results from a spectral ratios inversion of aftershocks recordings from the Chuetsu earthquake. These equations were derived for the Niigata region and therefore they are not necessarily applicable for other regions. However they represent a good tool for approximately estimating PGV values when information of a detailed 3D velocity structure (above the seismic bedrock and below the engineering bedrock) is not available. Future studies may address the calculation of similar equations for other regions with different geological conditions. We obtained that the PGV amplification with respect to the seismic bedrock is better correlated with the average S-wave velocity of the upper 10m of soil (Vs5 and Vs10) than with Vs30, which is a widely used index to estimate PGV amplifications with respect to an engineering bedrock site. The reason for this is probably the large impedance between the upper 10m of soil and the seismic bedrock, compared with the upper 30m. We simulated ground motions at a 250m grid in the Niigata region by incorporating site effects and using a multi-asperity model derived from near-fault ground motions. Our agreement to observed KiK-net, K-NET and JMA intensity records within the Niigata region is good. We also obtained a good correlation between observed damage to roads and simulated PGV values. We are able to reproduce large ground motions in regions with a high concentration of road damage, and with few available ground motion recordings, highlighting the importance of ground motion prediction. We obtained a low Q value (Q=63f 1.14 ) for the Niigata prefecture region, in agreement with others values obtained from coda Q analysis for the region. This low Q value for the Niigata region, has been found to correspond to comparatively large GPS strain rates within Japan. Acknowledgments K-NET and KiK-net strong motion data used in this study was provided by NIED. 7. References Jin, A. and K. Aki, (2005). High-resolution maps of Coda Q in Japan and their interpretation by the brittleductile hypothesis, Earth Planets Space, 57, 403-409. Midorikawa, S., M. Matsuoka, and K. Sakugawa, (1994). Site effects on strong-motion records observed 665
during the 1987 Chiba-ken Toho-oki Japan earthquake, The 9 th Japan Earthquake Engineering Symposium, 3, 85-90, 1994. Pulido N., and T. Kubo, (2004). Near-Fault Strong Motion Complexity of the 2000 Tottori Earthquake (Japan) from a Broadband Source Asperity Model, Tectonophysics, 390, 177-192. Sakai, H., K., Hasegawa, N. Pulido and T. Sato, (2006). Relationship between strong motion and road damage during the 2004 Mid-Niigata earthquake, Journal of Structural Engineering (in Japanese), 52a, 1, 301-308 Wakamatsu, K., M., Matsuoka, and H. Sakakura, (2005). Development of an engineering geomorphologic classification map of 250 meters grid cell for the Niigata Region and its applications. Abstract Volume of the 28 th JSCE Earthquake Engineering Symposium (in Japanese), 4. Yagi, Y., (2004). Source process of the 2004/10/23 Mid-Niigata prefecture earthquake (in Japanese), http://iisee.kenken.go.jp/staff/yagi/eq/20041023/source.pdf. 666