Evaluating a kinematic method for generating broadband ground motions for great subduction zone earthquakes: Application to the 2003 Mw 8.3 Tokachi‐Oki earthquake

We compare broadband synthetic seismograms with recordings of the 2003 Mw">MwMw 8.3 Tokachi‐Oki earthquake to evaluate a compound rupture model, in which slip on the fault consists of multiple high‐stress‐drop asperities superimposed on a background slip distribution with longer rise times. Low‐frequency synthetics (<1  Hz"><1  Hz<1  Hz) are calculated using deterministic, 3D finite‐difference simulations and are combined with high‐frequency (>1  Hz">>1  Hz>1  Hz) stochastic synthetics using a matched filter at 1 Hz. We show that this compound rupture model and overall approach accurately reproduces waveform envelopes and observed response spectral accelerations (SAs) from the Tokachi‐Oki event. We find that sufficiently short subfault rise times (i.e., <∼1–2  s"><∼1–2  s<∼1–2  s) are necessary to reproduce energy ∼1  Hz">∼1  Hz∼1  Hz. This is achieved by either (1) including distinct subevents with short rise times, as may be suggested by the Tokachi‐Oki data, or (2) imposing a fast‐slip velocity over the entire rupture area. We also include a systematic study on the effects of varying several kinematic rupture parameters. We find that simulated strong ground motions are sensitive to the average rupture velocity and coherence of the rupture front, with more coherent ruptures yielding higher response SAs. We also assess the effects of varying the average slip velocity and the character (i.e., area, magnitude, and location) of high‐stress‐drop subevents. Even in the absence of precise constraints on these kinematic rupture parameters, our simulations still reproduce major features in the Tokachi‐Oki earthquake data, supporting its accuracy in modeling future large earthquakes.