We perform kinematic and dynamic inversion of the 24 July 2008 (Mw 6.8) Northern Iwate intermediate depth earthquake in Japan using strong‐motion records from the K‐NET and KiK‐net networks. The rupture of this moderate magnitude earthquake is modeled as a simple elliptical patch. The optimal solutions are found comparing observed and synthetic records using an L2 norm and the neighborhood algorithm to search for the best solution, followed by an exploration of solution space with a Monte Carlo technique. The geometry of the rupture, rupture velocity, and slip distribution are estimated by kinematic inversion. The rupture geometry, stress, and friction parameters are obtained by dynamic inversion. Both approaches converge to very similar source models with semiminor axes of 4 km, maximum slip of about 4 m, and large stress drops in the 30–45 MPa range. Rupture duration was less than 3 s because of very high sub‐Rayleigh rupture propagation speeds. Energy release rate for the best models was in the range 23–36 MJ/m2, a rather large value for events of this size. For both kinematic and dynamic inversion we found families of solutions that fit the strong‐motion data within a certain error, confirming the strong trade‐off among inverted parameters. Finally, we demonstrate that dynamic inversion solutions are controlled by the dynamic similarity parameter κ and by seismic moment M0. These two parameters define a region of model space where dynamically similar models fit the observations with approximately the same misfit.
Online Material: Figures that compare observed and synthetic waveforms for all available stations and comparison of Fourier spectrum of observed and synthetic seismograms.