RESEARCH ACTIVITIES IN 1999

The followings are the summaries of researches for the themes related to prediction of strong ground motions in this year.


Source Process of the 1999 Kocaeli, Turkey, Earthquake

Sekiguchi, Haruko Tomotaka Iwata, and Kojiro Irikura
The multi-time window linear waveform inversion was performed for the 1999 Kocaeli, Turkey earthquake. We focused on the possibilities of rupture propagation with super-shear velocity or rupture triggered by P-wave in the east of the hypocenter. Such extraordinary ways of rupture propagation are indicated by the short duration of observed waveforms and the short P-S time at a station SKR which is located along the surface breaks in the east of the hypocenter. The appropriateness of the inverted source models was judged using ABIC and a source model with P-wave triggering assumption was selected as a best model. This source model suggests the possibility of rupture extending onto the Duzce fault, although the slips on this segment were not well constrained. The final slip distribution is strongly heterogeneous. Several large asperities are found at deeper part on the fault plane (more than 10km depth). A small asperity is triggered by P-wave near the SKR station.

Source Process of the 1999 Chi-Chi, Taiwan, Earthquake

Sekiguchi, Haruko Tomotaka Iwata, and Kojiro Irikura
We also performed waveform inversion for the 1999 Chi-Chi, Taiwan, earthquake using near-source strong motion data. We assumed a fault plane model from information of surface rupture, hypocenter, moment tensor solution of tele-seismic data and aftershock distribution. Different dip angles of aftershock distributions observed are shown between northern and southern parts and we assumed a two-plane fault model constructed by one plane with 0degree dipping for the northeastern deeper part and another plane with 29 degrees for the other part of the fault plane. We used 20 station data (velocity 0.1-1.0Hz) whose epicentral distance is less than 150km. GreenIUJs functions are calculated using two different 1D underground structures for hanging wall and foot wall sides. A final slip distribution from this inversion is shown as below. Relatively small asperities are observed in the southern part. Source effects on near-fault ground motions will be discussed.

High-Frequency Radiation Process of the 1994 Northridge Earthquake

Yasumaro Kakehi
From the accumulation of source process analyses of the actual earthquakes, we can abstract useful information for the source modeling in the quantitative strong ground motion prediction. As a part of such an approach, high-frequency radiation process of the 1994 Northridge earthquake is investigated. From the envelope inversion of the acceleration seismograms observed at 5 stations close to the fault plane, distribution of radiation intensities of high-frequency (5 - 10 Hz) waves on the fault plane is estimated. The empirical Green's functions are used for the calculation of synthetic waveforms in the process of the envelope inversion. The inversion result shows that high-frequency waves are radiated at the edges of the large-slip areas. This indicates that the source images are different between high- and low-frequencies. Therefore, the source model should be constructed taking the target frequency range into consideration in the strong motion prediction process. From a physical viewpoint, high frequency radiation at the edges of large-slip areas can be interpreted as stopping phases.

Strong Motion Evaluation in the Near-Fault Region Considering the Slip-Velocity Function of the Source

Hiroshi Kawase and Shin'ichi Matsushima
We have established a four-asperity source model for the Hyogo-ken Nanbu earthquake of 1995 by using a forward modeling technique (Matsushima and Kawase, 1998) and a deconvolved bedrock motion at the JMA Kobe station (Kawase, 1996). In the course of the forward modeling we found that near-fault ground motions are primarily characterized by the several directivity pulses each of which is generated by the so-called forward rupture directivity within a separated asperity. We also found that the waveform of the directivity pulse is controlled by the size of an asperity and the slip-velocity function within it. The pulse-width of the directivity pulse is proportional to the size of an asperity, while the peak velocity amplitude of the directivity pulse is proportional to the slip velocity. Under the assumption that our four-asperity model is a realistic source of the Hyogo-ken Nanbu earthquake, we calculate velocity synthetics for a 1-D structure to understand the characteristics of the strong motions in the near- and far-filed from the source. We found that forward rupture directivity is quite strong not only in the near-fault region but also in the far-field region, which we must take into account for quantitative strong motion prediction.

Quantitative Estimation of the Basin-Edge Effect in Kobe

ShinIUJichi Matsushima and Hiroshi Kawase
We derived a relatively simple multiple asperity source model of the Hyogo-ken Nanbu (Kobe) earthquake of 1995 and reproduced strong ground motions in Kobe considering a 3-D basin structure using 3-D FDM with 4th ordered staggered-grid scheme (Graves, 1996). The calculated synthetic waveforms match the data fairly well at stations located over a wide range in Kobe. The distribution of PGV forms a belt-like shape that resembles the damage concentrated area during the earthquake. Finally we separate the direct S-waves and the edge-induced waves in the basin to see the contribution of the IRJedge effectISJ (Kawase, 1996) in the distribution of PGV. With only the direct S-wave PGV in most part of the basin does not vary so much. On the other hand, PGV of the edge-induced wave result in a distribution similar to the damage belt and this distribution is strongly influenced by the complex structure of the basin edge. From these results we conclude that in order to simulate strong ground motions accurately at sites like Kobe that lies near the basin edge and where the source rests underneath, it is important to take into account of the edge effect caused by the deep basin structure and the slip velocity function of the source.

Simulations of Long-Period Ground Motions of Large-Scale Sedimentary Basins

Yoshi Hisada, Jacobo Bielak, Omar Ghattas,.and R. O'Hallaron
The purpose of this project is to develop a methodology for simulating long-period strong motions of sedimentary basins (the Osaka and Kanto basins), using FEM (Finite Element Method) and a simplified analytical method. First of all, we simulated the long-period strong motions (> 0.9 second) in the Kobe area for the mainshock (1/17/95) and the aftershock (2/2/95) of the Hyogoken-Nanbu (Kobe) earthquake (Hisada et. al., 1998). In order to model ground motions in the realistic and highly heterogeneous basin structure, we used an unstructured 3D FEM, which runs efficiently on parallel computers. The simulation results agree well with the observations, and show the strong 3D basin effects, especially within the disaster belt. Second, we developed an analytical method using a simplified 2-D basin model to quantitatively understand the basin edge effects. The basin model consists of a single flat layer with a rectangular edge, which is surrounded by rigid bedrock. We can analytically express the basin response due to the bedrock excitation by the superposition of the vertical 1D waves and the horizontal edge waves; the latter consists of the evanescent and surface waves. As compared with the 1D waves, the maximum amplitude of the 2D waves is about half at the edge, because of the destructive interference between the 1D and the edge waves. On the contrary, the amplitude of the 2D waves becomes about two times larger at the distance of half wavelength from the edge, because of the constructive interference. Eventually, the amplitudes of the 2D waves become same as the 1D at distances more than the one wavelength from the edge. We will examine these phenomena for more generalized cases using 3D FEM.

Macro-Micro Analysis for Strong Motion Prediction

Muneo Hori and Tsuyoshi Ichimura
Higher spatial and time resolution is essential for the estimation of strong motion distribution in a metropolis. Modeling ground motions, however, is not an easy task as the ground and geological structures are complicated and available data of the structure are limited. While the length-scale is different, regarding a metropolis as a heterogeneous body consisting of wildly changing structures, we can apply techniques of micromechanics established for composite materials. A macro-micro analysis is one example of such application. This analysis method takes the following procedures: 1) model a metropolis including its underground structure as a statistically heterogeneous body, making use of available measured data; 2) determine two deterministic bodies that bound the expectation of the statistical body; and 3) compute the wave propagation process in these bodies applying the singular perturbation expansion. Here, the two techniques of micromechanics are used, the generalized Hashin-Shtrikman variational principle to determine the two bounding bodies, and the singular perturbation expansion that is commonly used for the homogenization. In numerical computation, the singular perturbation leads to coupled problems for first- and second-order terms in the expansion, which correspond to responses viewed in larger and smaller length scales, respectively. Indeed, the first-order solution of low resolution is obtained for the whole body, and the second-order solution of high resolution is obtained for each part of the body. The structure used in computing the first-order solution is determined by analyzing the second-order solution, and the second-order solution is obtained by using the first-order solution as input data. As an example, the strong motion distribution of Yokohama City of 10km length scale is computed in a spatial resolution of 1m. The comparison of the numerical solution with actual data measured at several sites is satisfactory, and the validity of the proposed analysis method is verified.

Simulation of Broad-Band Strong Ground Motion during the Kocaeli, Turkey, Earthquake

Katsuhiro Kamae
Strong ground motions in near-source from the Kocaeli earthquake have been simulated based on the heterogeneous source model and the hybrid simulation technique. The heterogeneous source model composedB!!Jof four asperities has been constructed based on the slip distribution derived from the source inversion analysis. Simulations have been carried out by the hybrid scheme combined the 1-D theoretical method for long-period motions (>1 sec) with the stochastic simulation technique for short-period motions (<1 sec). We assumed that ground motions are generated from only four asperities. The four asperities are roughly characterized by the comparison of the synthetics with the observed motions in three stations (SKR, YPT, GYN). As an example, the comparison between the synthetic and the observed velocity and acceleration motions at SKR station is shown. The estimated the largest as well as the combined asperity areas are consistent with those derived from the empirical self-similar scaling relationships by Somerville et al. (1999).

Semi-Empirical Method based on Variable-Slip Rupture Model
Kazuo Dan and Toshiaki Sato
A variable-slip rupture model for large earthquakes, obtained by the source inversion of long-period seismic waves, was taken into account in a semi-empirical method for simulating strong-ground motions in the wide-period range. The source spectrum of each of the sub-faults was extrapolated in the short-period range by the omega-square model with two corner circular frequencies. The one is due to the temporal integration of the slip-velocity time function. The other is due to the spatial integration of the slip-velocity time function on the sub-fault. The method was applied to the variable-slip rupture model for the 1923 Kanto, Japan, earthquake of MS 8.2 obtained by Wald and Somerville (1995). The pseudo velocity response spectrum of the simulated strong-ground acceleration at TOK (Tokyo JMA) was about 60 cm/s on an average in the wide-period range of 0.2 to 10 seconds, and was consistent with that of the record, whose clipped parts had been restored by Yokota et al. (1989), observed at HNG (Hongo, Tokyo) during the Kanto earthquake. The instrumental seismic intensities (JMA, 1996) of the simulated accelerations at TOK and YOK (Yokohama JMA) were also consistent with the seismic intensities of 6, compiled by the JMA (1983). The instrumental seismic intensity of the simulated accelerations at KNS (soil site in Odawara) was consistent with the seismic intensity of 7, estimated from the overturned tomb stones and the collapsed houses (Monobe, 1925).

A Theoretical Omega-Squared Model for Predicting Near-Source Strong Ground Motions at Broad-Band Frequencies

Yoshi Hisada
The recent earthquakes have been drastically demonstrated that the necessity of a methodology for predicting near-source strong ground motions at broad-band frequencies, i.e., from coherent characteristics at low frequencies (e.g., long-period directivity pulses and permanent displacement offsets) to random characteristics at high frequencies (e.g., acceleration). For this purpose, we proposed the omega-inverse-squared model (Hisada, 2000), which is a theoretical omega-square model, and is modified from the k-square model of Bernard et al. (1996) by considering the spatial variation in slip distribution, slip velocity, and rupture velocity. Using a hybrid method, which is the combination of the omega-inverse-squared model at low frequencies (0 to 1 Hz) and the Kamae-Irikura's method (Kamae, et al, 1998) at high frequencies (0.5 to 10 Hz), we successfully simulated the near-field strong motions during the 1985 Michoacan earthquake at broadband frequencies (Hisada, 2000). We are planning apply this method to the 1995 Kobe, 1999 Kocaeli (Turkey) and Chi-Chi (Taiwan) earthquakes.




RETURN TO UJ11 Home