→ Japanese

Ground Motion and Rupture Process of the 2003 Sanriku-Minami Earthquake Obtained from Strong Motion Data of K-NET and KiK-net

*Aoi, S., H. **Sekiguchi *T. Kunugi, *R. Honda and *H. Fujiwara
(*National Research Institute for Earth Science and Disaster Prevention
**National Institute of Advanced Industrial Science and Technology)

  The Sanriku-Minami earthquake (2003/5/26, 18:24 JST, Mw 7.0), an intraslab earthquake of the Pacific Plate, occurred at about 70 km depth, several kilometers offshore, near the boundary of Miyagi and Iwate Prefectures (38.806N, 141.685E; NIED, Hi-net) (Figs. 1 and 2). Its ground motion was widely felt in the northern half of Japan and recorded at about 800 stations of nationwide strong motion networks (Fig. 3), K-NET (Kinoshita, 1998) and KiK-net (Aoi et al., 2000). At five stations, including two JMA (the Japan Meteorological Agency) stations, PGA (peak ground acceleration) exceeded 1g (980 gal) and 0.5 g within the epicental distance of about 100 km.
  We estimated the source process of this earthquake by the multi-time window linear waveform inversion method (Hartzell and Heaton, 1983) using the strong motion data at 17 stations (Fig. 4) in 0.05-0.5 Hz range. Set up for the inversion analysis is shown in Table. We assumed the rupture starting point at (38.820N, 141.657E, 72.0km), the hypocenter determined from P-wave arrivals at 30 nearby Hi-net stations (Shiomi, personal commu.) (Fig. 1).
  Total slip distribution, the comparison of observed and synthetic waveforms and the time progression of rupture are shown in Figs. 5, 6 and 7, respectively. Rupture mainly progressed toward north. There are two areas with large slips, one at the hypocenter on the southern segment and the other on the northern segment a little deeper than the first one. The slips in these areas cause the two pulses which well explain the two remarkable pulses in the observed waveforms. The first pulse in the observed waveforms could not be modeled when a single segment fault model with mechanism of the CMT solution was adopted. The total moment of the source model is 7.6x10**19Nm (Mw=7.2). The first time window triggering front propagation velocity that gives the smallest misfit of the waveform fitting was 3.0 km/s.

Table: Data
Acceleration data at 17 stations of K-NET and KiK-net (NIED)
Integrated into velocity
Band-pass filter: 0.05-0.5 Hz
Table: Set up for the waveform inversion
Fault plane model:
   North segment: area: 18 km x 28 km, strike/dip: 2.9/87.2
   South segment: area: 10 km x 28 km, strike/dip: 190./69.
   Subfault size 2.0 km x 2.0 km
   # of time windows: 6 (Rise time 1.0 s at 0.5 s interval)
   First time window triggering front propagation velocity: 3.0 km/s
Green's function:
   Methodology for computation of synthetics
      : Discrete wavenumber method (Bouchon, 1981)
      : RT matrix method (Kennette and Kerry, 1983)
      : Convolution of moving dislocation
                                  (Sekiguchi et al., 2002)
   Velocity structure model: Ukawa et al. (1984)
Constraints:
   Variation of rake angle
      :within 45 degrees from the rake of CMT solution
   Smoothing of slip in space and in time