HR: 1330h
AN: S42D-0197    [PDF]
TI: Grid Approach for Kinematic Source Inversion
AU: Fujii, Y
EM: fujii@geo.kyushu-u.ac.jp
AF: Dept. of Earth & Planet. Sci., Kyushu Univ., Hakozaki 6-10-1, Fukuoka, 812-8581 Japan
AU: * Takenaka, H
EM: takenaka@geo.kyushu-u.ac.jp
AF: Dept. of Earth & Planet. Sci., Kyushu Univ., Hakozaki 6-10-1, Fukuoka, 812-8581 Japan
AB: In seismic tomography for local parameterization of the spatial functions to be reconstructed (e.g., seismic velocities) there are two basic approaches: one is the block (cell) approach and the other is the grid approach. In the block approach the medium under study is divided into many blocks (cells) and the seismic velocity in every block is taken as unknown parameter. In the grid approach grid nodes are adopted to model the earth structure and velocities at the nodes are taken as unknown parameters; the velocity at any point in the model is calculated by interpolation from those at the nodes surrounding the point. In the grid method the velocity is continuous everywhere, while in the block method artificial velocity discontinuities are introduced into the model between the blocks. In kinematic source inversion the above two approaches, the block approach and the grid approach, can also be considered. The multi-time window linear inversion (e.g., Hartzell and Heaton, 1983) belongs to the block approach, which is now a most popular method and has been widely applied to broadband seismic records including teleseismic and strong-ground motion records to estimate spatio-temporal slip distribution on seismic fault planes of middle to large earthquakes. In this method the fault plane is conventionally divided into many subfaults (cells) to express spatial variation, and each subfault is represented by a single or many point sources placed at constant intervals on the planar surface. Such conventional subfault discretization corresponds to the block approach of seismic tomography, and it is not continuous at the subfault boundaries. Since this conventional subfault discretization is used by most existing methods for kinematic source inversion applied to waveform data, they may all be categorized into the block approach. One exception is Ide and Takeo's (1997, JGR) method, which belongs to the grid method. In their method the spatio-temporal distribution of slip velocity is expanded with the linear b-spline basis functions in 2D space as well as in time. This discretization can give a slip distribution continuous everywhere. In the present study we propose an extension of this grid method. Main extension is that the effect of the rupture propagation is included even under the sub-grid level by using the presentation of slip velocity with an explicit rupture propagation term, which has recently been exploited by Takenaka et al. (2002) for extracting spatially continuous slip distribution from results of conventional source inversion. Furthermore our method can also treat slip direction change on the fault plane unlike Ide and Takeo's original method. In the presentation we demonstrate the feasibility of our method by using simulation data.
DE: 3260 Inverse theory
DE: 7200 SEISMOLOGY
DE: 7209 Earthquake dynamics and mechanics
DE: 7212 Earthquake ground motions and engineering
SC: Seismology [S]
MN: 2003 Fall Meeting