This article is the first of a pair of articles that estimate regional and local strong motions from the 1923 Kanto, Japan, earthquake. This Ms 8.2 earthquake caused the most devastating damage in the metropolitan area in Tokyo history. In this article, we first calibrate wave propagation path effects with a moderate-sized modern event. This event, the Odawara earthquake of 5 August 1990 (M 5.1), is the first earthquake larger than M 5 in the last 60 years near the hypocenter of the 1923 Kanto earthquake. We estimate the source parameters based on a grid-search technique using body-waveform data bandpass filtered from 1 to 10 sec at four local stations, because accurate source parameters are critical for calibrating the propagation effects. We find that the Odawara earthquake had a depth of 15.3 km, a dip of 35°, a rake of 40°, a strike of 215°, a seismic moment of 3.3 × 1023 dyne-cm, a source duration of 0.65 sec, and a stress drop of 170 bars.
Next, we investigate the effects of the propagation paths to the local and regional stations where seismograms of the 1923 Kanto earthquake were recorded, by comparing recorded waveforms with synthetic seismograms built with the calibration event. Path-specific flat-layered velocity models are estimated along travel paths from the event to stations Hongo (epicentral distance R = 82 km) in Tokyo, Gifu (R = 213 km), and Sendai (R = 374 km) using forward modeling. In constructing the velocity model for the Gifu station, we use STS-1 broadband seismograms recorded at the nearby Inuyama station. Consequently, at periods greater than 3 sec, the velocity models for stations Hongo and Gifu can successfully reproduce both body waves and direct surface waves, and the velocity model for Sendai station can explain the predominant direct surface waves. In the companion article (Sato et al., 1998), these velocity models are used to examine the adequecy of the variable-slip rupture models of the 1923 Kanto earthquake (Wald and Somerville, 1995; Takeo and Kanamori, 1992) to explain recorded seismograms and also to simulate strong motions from that event.