Acceleration fields in 3-D elastic media due to a propagating rupture at constant super-shear velocities are evaluated. The elastodynamic Green's tensor and its gradient are obtained in closed-form for arbitrary compressional and shear Mach-numbers and arbitrary source time-function. It is found that the acceleration field consists of three main phases: a Mach-wave, a starting phase, and a stopping phase. The stationary phase approximation is used to obtain the explicit dependence of the Mach-wave peak amplitude on the source, propagation, and media elements. It is shown that the Mach-wave plays a dominant role in the acceleration signal of the near-field zone.

View full abstract

Predictions of seismic intensities are attempted based on a three-dimensional attenuation structure and amplifying factors at stations obtained in the previous study for the Tohoku district, Japan. An empirical relation between `earthquake size' related to the intensity and JMA magnitude M_J is also used. The first prediction is to map spatial distribution of the minimum magnitude M_J of an earthquake which causes intensity 5 at a specific JMA station. The result shows that the effects of regional variation of attenuation and local site condition at the station are important in assessing the seismic hazard. Another attempt is made to predict an annual number of felt earthquakes at each JMA station based on the JMA earthquake catalogue for a period from 1963 to 1984. A comparison of predicted number with the observed one shows a good coincidence between them for each year. This may suggest that the proposed method can successfully predict intensities. However, a comparison of the predicted cumulative number for each intensity grade with the observed one shows that the cumulative numbers for intensities 2, 3, and 4 are underestimated by the proposed method. This result may suggest a nonlinearity between the JMA intensity scale and the logarithm of acceleration, which has been suggested by other evidences.

View full abstract

We determined fault plane solutions and focal depths for seven events which occurred in the Japan trench outer slope, using comparison between synthetic and observed long- and short-period teleseismic seismograms. The five normal fault solutions obtained in this study have nearly vertical nodal planes parallel to the trench and their depth range, 2-16 km below the seafloor, is in accord with bending of the oceanic lithosphere prior to subduction. The occurrence of both types of normal faulting, downthrow of both the continental and oceanic sides, is likely to be associated with the formation of horst-graben structures in the outer trench slope near the trench axis. One normal fault solution (31 km below the seafloor) that is significantly deeper than the other normal fault solutions has a nearly vertical nodal plane striking parallel to the ENE magnetic anomaly lineations in the region. This event is also characterized by its lower stress drop or slower rupture than the other events. This event would represent a reactivation of the initial structural fabric inherited from the accretion at the mid-oceanic ridge and indicates that the bending stress is small at this depth. One reverse fault solution, 41 km below the seafloor, is in accord with the compressional bending stress in the deeper portion of the lithosphere. These events indicate that a neutral surface of bending stress is around 30 km depth in this region at least for the past few decades, consistent with the elastic thickness of lithosphere in adjacent regions. This implies that the lithosphere is not entirely in deviatoric tension in this region, suggesting that the 1933 Sanriku earthquake may be a large bending event.

View full abstract

Computation of tsunami waveforms for actual topography is usually made by a time-stepping finite-difference method which needs a long computation time and must be repeated for different initial conditions. We propose here an alternative computation method by a superposition of normal modes. Although the eigenvalue problem to obtain the normal mode solutions needs more memories, once we get the solution, the waveforms can be computed by merely summing up the normal modes with weights determined by the initial condition. We present the waveforms for a simple water basin computed by both methods. The results show that the new method is useful for a completely or almost closed basin including the tsunami source.

View full abstract