Simulation of Long-Period Ground Motions for the 1923 Kanto Earthquake (M≈8)

We performed numerical simulations of long-period ground motions in Tokyo for the 1923 Kanto, Japan, earthquake (M_s=7.9 to 8.1) which devastated Tokyo, Yokohama, and their environs, and caused more than 130,000 fatalities. We used reflection-transmission matrices and the discrete wavenumber integration method to compute ground motions for fault models placed in layered structures. The objective of this study is twofold: 1) to estimate the long-period response spectrum for the 1923 Kanto earthquake; and 2) to investigate the effects of various source parameters on the ground motion so that we can assess the variability of estimated ground motions, and apply the results to a broader class of earthquakes than just a single design earthquake specifically fior the 1923 event. The Kanto earthquake was recorded in Tokyo with a Ewing seismograph and an Imamura seismograph. The ground motions estimated by earlier investigators from these two seismograms differed significantly. This conflict can be reconciled if we assume that the solid friction of the Ewing seismograph was very high during shaking. Solid friction can reduce the resonance of the instrument, and long-period ground motions with large amplitudes can be recorded correctly. We conclude that the ground motion of the Kanto earthquake had a very large long-period component with a velocity response spectrum of 120 cm/sec (5% damping) at a period of 13 sec. The velocity response spectrum at a period of 7.5 sec is estimated to be about 50 cm/sec. Numerical simulations produced a wide range of ground motions and response spectra, even with a given fault geometry and seismic moment. For a rupture model initiating from the southwestern end of the fault plane the most probable epicenter of the 1923 earthquake the computed response spectra have a range of 10 to 100 cm/sec at a period of 7.5 sec, which brackets the observed level. All the response spectra computed for this model have a peak in a period range of 10 to 13 sec. The slip distribution and the rupture direction significantly influence simulated ground motions. Large subevents in a shallow structure enhance the ground motion significantly, especially if the rupture propagation is toward the site. One of our extreme models, which has large slip of about 8 min the shallow crust at the western end of the fault plane, can produce a large ground motion comparable to that estimated from the Ewing seismogram. Reducing the rise time or increasing the rupture propagation increases the spectral amplitude at periods shorter than 5 sec. Also, if the site is located on a very soft sediment, significant (a factor of 1.4) amplification occurs. The basin structure beneath Tokyo would increase the duration of ground motion significantly. Although increased duration does not significantly affect the response spectrum, it will play an important role in the nonlinear response of structures