S31B-0440
Thickness detection of sedimentary basin by P-to-S-phases of earthquake recordings at Ibaraki-ken, Japan
The Kanto sedimentary basin has very complicated geological formations with various sedimentary layers. The thickness of sedimentary basins varies irregularly from zeros to over several thousand meters down to the pre- Neogene bedrock. Therefore, the understanding the thickness is the fundamental in order to study the seismic ground motions. The vertical seismic profile (VSP) method may give a high-resolution seismic image at a borehole in both of P and S wave information. However, VSP is only limited to a few areas since it is too costly to be widely used. A dozen of deep-wells down to 2000 meters with VSP in the Kanto area have been drilled over past 15 years by NIED. In order to extend the high-quality VSP results farthest, we utilize these VSP data to obtain regression functions for sedimentary depth against the travel time difference of direct P and the direct S waves. On the other hands, the travel time difference between the direct P and the converted PS waves may be detected from P-to-S converted phases by means of all-pass function deconvolved from Receiver Functions (APRF) (Izumi et.al., 1990, Satoh, 2005, Hao, et.al., 2005, 2006). Comparing the classic receiver function for detection of interfaces with abrupt velocity changes in the earth crust, this APRF method can identify P-to-S phase in the near- surface interfaces much clearly, even for a high frequency of 10Hz. We use APRF with the VSP regression functions to estimate the thickness of sedimentary basin beneath each ground motion station. The objective area in this paper is whole of Prefecture Ibaraki-ken, which is located at north-east part of Kanto basin. The northern and northwestern of Ibaraki-ken are mountain areas, and Mt. Tsukuba and Lake kasumigaura are located in the center. A total of 112 strong motion stations, in which 84 stations from Sk-net, 19 K-net, and 14 KiK-net, have been targeted to study the underground structure beneath each station. Over 3000x3 records have been visually identified for the initial P and S phases from events during 1999 to 2005 for Sk-net, and from 1996 to 2007 for K-net and KiK-net. Averagely, about 30 events, whose amplitudes ranged from 2 to 558 gal, are used at each station. The Ts-Tp travel time difference was related both of velocities of Vp and Vs, as well as depth of sedimentary. One of key problems was how to convert the Ts-Tp travel time difference to the depth since a trade-off between factors of the velocity and depth. The regression functions of sedimentary depth against the Ts-Tp gave the statistics solution with a VSP calibration at deep-wells around. Recently, we have obtained the five regression functions at NRT, CHB, EDS, FUT, and YOR in the areas (Hao and Fujiwara, 2007). By combination of the 112 Ts-Tp results and the regression functions, the sedimentary depths in Ibaraki area have been revealed. The results have been confirmed at several areas where there were other deep wells and fault investigations. Acknowledgements The dense strong-motion observations by K-net, Sk-net and KiK-net made this study possible. http://www.j-shis.bosai.go.jp
S31B-0441
Rupture process and strong ground motions of the 2007 Niigataken Chuetsu-Oki earthquake -Directivity pulses striking the Kashiwazaki-Kariwa Nuclear Power Plant-
The Niigataken Chuetsu-Oki earthquake occurred on July 16, 2007, northwest-off Kashiwazaki in Niigata Prefecture, Japan, causing severe damages of ten people dead, about 1300 injured, about 1000 collapsed houses and major lifelines suspended. In particular, strong ground motions from the earthquake struck the Kashiwazaki-Kariwa nuclear power plant (hereafter KKNPP), triggering a fire at an electric transformer and other problems such as leakage of water containing radioactive materials into air and the sea, although the radioactivity levels of the releases are as low as those of the radiation which normal citizens would receive from the natural environment in a year. The source mechanism of this earthquake is a reverse fault, but whether it is the NE-SW strike and NW dip or the SW-NE strike and SE dip are still controversial from the aftershock distribution and geological surveys near the source. Results of the rupture processes inverted by using the GPS and SAR data, tsunami data and teleseismic data so far did not succeed in determining which fault planes moved. Strong ground motions were recorded at about 390 stations by the K-NET of NIED including the stations very close to the source area. There was the KKNPP which is probably one of buildings and facilities closest to the source area. They have their own strong motion network with 22 three-componentsf accelerographs locating at ground-surface, underground, buildings and basements of reactors. The PGA attenuation-distance relationships made setting the fault plane estimated from the GPS data generally follow the empirical relations in Japan, for example, Fukushima and Tanaka (1990) and Si and Midorikawa (1999), even if either fault plane, SE dip or NW dip, is assumed. However, the strong ground motions in the site of the KKNPP had very large accelerations and velocities more than those expected from the empirical relations. The surface motions there had the PGA of more than 1200 gals and even underground motions at the basements of the reactors locating five stories below the ground had the PGA of 680 gals. We simulated ground motions using the characterized source model (Kamae and Irikura, 1998) with three asperities and the empirical Greenfs function method (Irikura, 1986). Then, we found that the source model should be a reverse fault with the NE-SW strike and NW dip to explain the strong motion records obtained near the source area. In particular, strong ground motions in the site of the KKNPP had three significant pulses which are generated as directivity pulses in forward direction of rupture propagation. This is the reason why the strong ground motions in the site of the KKNPP had very large accelerations and velocities. The source model is also verified comparing the observed records at the KKNPP with the numerical simulations by the discrete wavenumber method (Bouchon, 1981).
S31B-0442
Simulation of Long-Period Ground Motions in the Southern Korean Peninsula for Validation of a Three-Dimensional Velocity Model
Recent advances on the understanding of the crustal structure in the southern Korean peninsula have elucidated the importance of path effects on ground motion from explosion sources and natural earthquakes occurred in and around the region. A three-dimensional velocity model of the region is constructed from studies based on the receiver-function analyses of broadband seismograms and the tomographic inversion of surface-wave Green's functions from cross-correlation of short-period ambient noise between pair of accelerograph stations. In order to evaluate the adequacy of the velocity model, we simulate ground motion velocity from the 20 January 2007, Mw 4.5, Odaesan, Korea, earthquake using the three-dimensional finite-difference method. The simulated ground motions are compared with the recorded motions at 23 broadband seismograph stations in the frequency band of 0.1-0.3 Hz. Our crude model for the sourthern Korean peninsula generally estimates well the long-period peak ground motion and dominant waveforms of the recorded motions. In the long period larger than about 3 seconds, the waveform and amplitude of S waves are similar to those of observed seismograms. Developments of P waves and coda of S waves are weak in the simulated seismograms compared to the recorded seismograms. This is because that most of the velocity model volume is based on the three-dimensional interpolation of one- dimensional velocity profiles obtained from receiver function analyses of sparse observation locations, which may have a smoothing effect on the real structure. The simulation also emphasizes the importance of the uppermost crustal structure above about 3 km depth believed to be highly heterogeneous as well as to be trapping most of S-wave energy. Future seismic explorations covering the entire southern Korean peninsula will improve the resolution of our model.
S31B-0443
Effects of Long-Period Ground Motion in Distant Basins: The 1906 San Francisco Earthquake and Comparison with Japanese Cases
Large earthquakes at shallow depths often excite long-period ground motions in distant sedimentary basins and damage large-scale structures. We have reported these effects with simulations of the 2003 Tokachi-oki and 2004 off-Kii peninsula, Japan, earthquakes using finite element method (FEM) of a voxel mesh. Besides of the above, we have performed long-period ground motion simulation for the 1906 San Francisco earthquake using the source model of Wald et al.(1993). Our simulation excited long-period ground motions in the LA basin in at a period of 7 sec. However, the results are not so strong as compared with the velocity response at distant basins during the above Japanese subduction-zone earthquakes. One of the reasons might be that the rupture directivity effect is not so strong for the southern region including the LA basin, because the large asperities of the 1906 San Francisco earthquake are mainly located at the northern region of the hypocenter. To clarify the quantitative response in the LA basin for the 1906 event, we will compare simulation results using other source models of Thatcher et al. (1997) and Song et al. (2007) that consist of larger asperities at the source region than the Wald's source model. Regarding the long-period ground motion simulation, our FEM code has advantage to deal with the effect of sea and topography, and the effect of intrinsic factor with the Rayleigh damping. We showed strong excitation of long- period ground motion with a period 7.4 sec in the Yufutsu basin during the 2003 Tokachi-Oki earthquake. Comparison of the simulated waveforms between with/without sea model has confirmed the significant contribution of long-period ground motion from the sea side of the Yufutsu basin. The intrinsic attenuation (Qp and Qs) is important parameter governing the wave propagation in a long distance. We have introduced it with the Rayleigh damping in the FEM code that can fit the Q value at two different frequency points in order to suit broader frequency range for the constant Q model. Toward realistic ground motion simulation, we will examine the above two effects on the long-period components reproduced by the 1906 San Francisco earthquake.
S31B-0444
Long-Period Ground Motion Characteristics in Osaka Basin, Japan - Examination of 3D Basin Structure Models
Osaka, the most populated area in western Japan, is located inside a sedimentary basin whose size is approximately 60km x 40km and the maximum bedrock is 3km. As there are many long-period structures such as skyscrapers and oil tanks, it is important to understand long-period ground motion characteristics in this area in order for the reliable prediction of long-period ground motions during future mega-thrust earthquakes in the subduction zone of Philippine Sea plate. Iwaki and Iwata (2007) evaluated the applicability of the two existent 3D basin structure models (Kagawa et al., 2004 and Horikawa et al., 2003) by conducting a long-period ground motion simulation of an event that occurred near the hypothetical subduction zone events. Both models reproduced the response spectra of the observed ground motions well, within factor of 2, at most stations. However, at some stations, the predominant periods of the observed ground motions were not well reproduced, which suggests the need for the models to be improved. In this study, we show observed long-period ground motion characteristics in Osaka basin from a number of events, and estimate the effects of 3D basin structure on the characteristics by model simulations. We first examined site-specific characteristics of long-period ground motions using the strong motion records of seven major inland earthquakes (M>6) and four deep earthquakes (M>7). We took the coda part (60s after the S-wave arrival) and calculated horizontal-to-vertical Fourier amplitude spectral ratio (H/V). The H/Vs of 11 events were stable at each station, which suggests they represent the long-period ground motion characteristics controlled by the underground structure rather than events. Then ground motion simulations were carried out by 3D finite-difference method (Pitarka, 1999). The target events were: 1) the largest aftershock of the 2004 Off Kii peninsula earthquake (MJMA6.5) and 2) the 2000 Tottori-ken Seibu earthquake (MJMA7.3), both 2-300km away from Osaka basin but of different backazimuths. Coda H/Vs calculated from the synthetic waveforms of two events are similar to each other, which could be corresponding to the observed characteristics. We compared the observed H/V, 3D simulated H/V, and the theoretical Rayleigh wave H/V obtained from 1D velocity structure beneath each station. We found that 3D simulated H/V explained the observed one better than the 1D H/V at most stations. This reconfirms the importance of the 3D underground structure on long-period ground motion characteristics inside the basin. Time- dependent characteristics of H/V would be discussed in order to improve the basin structure models. We used strong motion data provided by COERKA, NIED, JMA, Denkyo-net, and Osaka City Waterworks Bureau.
S31B-0445
Modeling and Characterization of Long Period Ground Motion and Associated Spectral Response
Recent advances in performance-based structural design and base isolation of structures against earthquake caused damage necessitates the accurate prediction of long period ground motions. In the present study three- dimensional finite difference calculations are performed to assess the spatial distribution of long period spectral response resulting from finite ruptures in heterogeneous earth media. Slip distributions of a number of recent earthquakes in the magnitude range of 6.0 to 7.4 are used to simulate the wave propagation with the staggered- grid finite-difference method developed by Madariaga and Olsen (1998) within the frequency range of 0 to 1 Hz. The resulting ground motion time histories, the associated acceleration and displacement response spectra at various distances from the fault as well as the spatial distribution of spectral response amplitudes at large discrete periods (2 to 10 sec) are studied for the characterization of long period ground motion and structural response in terms of fault normal, fault parallel and mean components. The results are also compared with recently developed NGA and other attenuation relationships. The findings of this study are believed to be particularly useful in the formulation of design elastic displacement spectra for seismic codes, and in zoning studies of seismic hazard for long-period structures.
S31B-0446
Modeling Long-Period Ground Motions for Marmara Region
As a consequence of the change of paradigm in earthquake resistant design stronger design earthquakes now control the seismic design of important structures. These stronger earthquakes include the effects of near-field pulses, fault-normal motions, and near-field deep soil site motions. As a consequence, there is a strong need for the development of robust and reliable techniques for the assessment of long period earthquake ground motions especially for near field conditions. To provide an example to the assessment of long period ground motion a comparative study has been carried out for the Marmara Region, Turkey. The probabilistic earthquake hazard has been investigated using PEER-NGA (2007) and older generation attenuation relationships for PGA and SA (0.2s, 1s, 2s, 4s, 6s, 8s and 10s) corresponding to 50, 10 and 2 percent probabilities of exceedance in 50 years. PGA values were used to obtain the EuroCode whereas, SA(0.2s and 1s) were used to obtain the NEHRP (2003) based response spectra. SA (0.2s, 1s, 2s, 4s, 6s, 8s and 10s) values were used to plot the equi-hazard spectrum. Furthermore associated hazard deaggregation has been conducted for several selected sites to obtain rational estimates of the deterministic long period spectral accelerations and the deterministic spectral shapes. Comparison of the findings indicate significant variation of long period spectral accelerations. The accuracy of seismic design spectra given in current codes is not sufficient at these periods. There is also a need to develop guidelines for the selection of design basis ground motion for long period or highly nonlinear (softening) structure.
S31B-0447
Earthquake Ground Motion in the Valley of Mexico: Basin Effects
We present a study of the ground motion and resulting amplification in the Mexico City Basin due to strong earthquakes in the Mexican Pacific Coast. We propose an approximation of the regional structure and Mexico City's basin and analyze their response to two shallow earthquakes generated near the coast. We compare two sets of three dimensional simulations: the first includes a soft structure similar in shape and properties to the Valley of Mexico, while the second excludes the soft soil deposits. Our 3D computations, with a maximum resolution of 0.75 Hz, reproduce the amplitude and long durations characteristics usually observed in the basin. We confirm that stations inside the Mexican Volcanic Belt experience amplification. In the frequency band 0.2-0.4 Hz additional amplification occurs inside the valley due to the shallow soil deposits in the lake bed region. We compare the normalized durations of the ground motion at several stations against observed data, and speculate on the durations of the soil motion as being a local effect due to the basin's shape and low velocities.
S31B-0448
Variability of kinematic source parameters and its implication on the choice of the design scenario
The near-fault seismic motion recorded during recent earthquakes (Chi Chi earthquake, 1999; Parkfield 2004) showed the high spatial variability of the motion. This variability is controlled by the fault geometry, the rupture process complexity and also by the propagation and site effects. As a result of these observations, the earthquake scenarios can be largely influenced by the vicinity of the seismic source, and the number of available recordings is still not enough to infer a robust parameterization of the ground motion to be used for retrieving multi-parametric predictive equations. In this study we modeled scenarios from different rupture models of a fault similar to the 1980 Irpinia, Italy, earthquake source (Mw 6.9). A discrete wavenumber-finite element technique (COMPSYN; Spudich and Xu, 2002) was used to compute full-wave displacement and velocity time series in the low-frequency band. We investigated the inter and intra event variability as a function of different source rupture parameters (rupture velocity, slip distribution, nucleation point, source time function), whose values depend on the degree of knowledge of the physical model controlling the process. The probability density functions of the simulated ground motion parameters, such as PGD and PGV, can be used to identify a specific scenario matching the engineering requests. http://esse3.mi.ingv.it/index.htm
S31B-0449
Long-period Amplification by Two-Dimensional Resonance in a Deep Alpine Valley
Many numerical and empirical studies have shown that deep sedimentary structures may significantly amplify strong ground motion at long periods, and the impact of this low-frequency amplification on tall buildings was demonstrated during many devastating earthquakes. In the framework of the SHAKE-VAL project we are analysing site effects in the Rhône valley, a deep sedimentary basin in Southern Switzerland. We computed site-to-reference spectral ratios from weak motion recorded on a temporary network of 12 seismometers. At most sites we observe substantial amplification between 0.50 and 0.60 Hz. The frequency of amplification is insensitive to the local sedimentary thickness, and the amplification level reaches a maximum of about 12 in the valley center. These observations are consistent with two-dimensional resonance of the deep basin, and the SH00 and SV0 fundamental modes of resonance can be isolated by rotating the ground motion to the directions perpendicular and parallel to the valley axis. To support our interpretation we performed numerical simulations of the recorded events with a 3-D finite difference method. The synthetic spectral ratios are generally in agreement with the observations at frequencies below 1 Hz. In a last step we estimated how non-linear soil behaviour will affect the long-period part of the signals during strong ground motion. We convolved synthetic time series with the transfer function of the basin and propagated the resulting signals through a shallow sand layer using a fully nonlinear 1-D method. Our results suggest that cyclic mobility will reduce the spectral acceleration at 0.50 Hz by 50% for rock acceleration exceeding 1 ms-1.
S31B-0450
CyberShake 2007: Update on Physics-Based Probabilistic Seismic Hazard Calculations for Southern California
To calculate a physics-based probabilistic hazard curve for a site of interest, we begin with the NSHMP-2002 ERF and identify all ruptures within 200 km of the site of interest. We convert the NSHMP-2002 rupture definition into multiple rupture variations with differing hypocenter location and slip distribution, which results in about 200,000 rupture variations per site. Strain Green Tensors are calculated for the site using the SCEC CVM (v4.0), and then, using reciprocity, we calculate synthetic seismograms for each rupture variation. Peak intensity measures are then extracted from these synthetics and combined with the original rupture probabilities to produce probabilistic seismic hazard curves for the site. Recent improvements include (1) development of a parallel CVM mesh generator, (2) implementation of an efficient 3D visco-elastic finite difference algorithm, (3) efficient storage and retrieval of SGTs, (4) implementation of an efficient earthquake simulation code, and (5) development of an optimized and efficient workflow system. Thus far, we have produced hazard curves for spectral acceleration at a suite of periods ranging from 3 to 10 seconds at about 10 sites in the Los Angeles region. Our preliminary analysis suggests the increased ground motion levels generated in the CyberShake simulations result from rupture directivity and basin response effects. More importantly, we find that the physics-based hazard estimates are much more sensitive to the assumed magnitude-area relation used in the definition of the ruptures than is found in the empirical approach. Our results suggest that a magnitude-area relation based on waveform modeling studies (e.g., Somerville et al, 2006) is the most appropriate parameterization for physics-based hazard calculations. These types of sensitivity analyses will be used to guide the future activities of the CyberShake program.
S31B-0451
3D Modeling of Strong Ground Motion in the Pacific Northwest From Large Earthquakes in the Cascadia Subduction Zone
The Cascadia subduction zone in the Pacific Northwest, USA, generates Great (megathrust) earthquakes with a recurrence period of about 500 years, most recently the M~9 event on January 26, 1700. Since no earthquake of such magnitude has occurred in the Pacific Northwest since the deployment of strong ground motion instruments, a large uncertainty is associated with the ground motions expected from such event. To decrease this uncertainty, we have carried out the first 3D simulations of megathrust earthquakes (Mw8.5 and Mw9.0) rupturing along the Cascadia subduction zone. The simulations were carried out in a recently developed 3D velocity model of the region of dimensions 1050 km by 550 km, discretized into 2 billion 250 m3 cubes with a minimum S-wave velocity of 625 m/s. The model includes the subduction slab, accretionary sediments, local sedimentary basins, and the ocean layer. About 6 minutes of wave propagation for each scenario consumed about 24 Wall-clock hours using a parallel fourth-order finite-difference method with 1600 processors on the San Diego Supercomputer Center Datastar supercomputer. The source descriptions for the Mw9.0 scenarios were designed by mapping the inversion results for the December 26, 2004 M9+ Sumatra-Andaman Islands earthquake (Ji, 2006) onto a 950 km by 150 km large rupture for the Pacific Northwest model. Simulations were carried out for hypocenters located toward the northern and southern ends of the subduction zone. In addition, we simulated two M8.5 events with a source area of 275 km by 150 km located in the northern and central parts of the model area. The sources for the M8.5 events were generated using the pseudo-dynamic model by Guatteri et al. (2004). All sources used spatially-variable slip, rise time and rupture velocity. Three major metropolitan areas are located in the model region, namely Seattle (3 million+ people), Vancouver (2 million+ people), and Portland (2 million+ people), all located above sedimentary basins amplifying the waves incident from the subduction zone. The estimated peak ground velocities (PGVs) for frequencies less than 0.5 Hz vary significantly with the assumed rise time. Using a mean rise of 32 s, as estimated from source inversion of the 2004 M9+ Sumatra-Andeman event (Ji, 2006), PGVs reached 40 cm/s in Seattle and 10 cm/s in Vancouver and Portland. However, if the mean rise time is decreased to about 14 s, as suggested by the empirical regression by Somerville et al. (1999), PGVs are increased by 2-3 times at these locations. For the Mw8.5 events, PGVs would reach about 10 cm/s in Seattle, and about 5 cm/s in Vancouver and Portland. Combined with extended duration of the shaking exceeding 1 minute for the Mw8.5 events and 2 minutes for the Mw9 events, these long-period ground motions may inflict significant damage on the built environment, in particular on the highrises in downtown Seattle. However, the strongest shaking arrives 1-2 minutes after the earthquake nucleates, indicating that an early warning system in place may help mitigate loss of life in case of a megathrust earthquake in the Pacific Northwest. Additional efforts should analyse the simulated displacements on the ocean bottom for tsunami generation potential.