• Japan subduction zone;
  • P wave model;
  • FDM;
  • SEM;
  • P-SV and SH waveforms;
  • low-velocity zone

[1] The tomographic P wave model for the Japan subduction zone derived by Zhao et al. (1994) has two very striking features: a slab about 90 km thick with P wave velocities 3–6% higher than the surrounding mantle and a mantle wedge with −6% low-velocity anomalies. We study three-component seismograms from more than 600 Hi-net stations produced by two earthquakes which occurred in the downgoing Pacific Plate at depths greater than 400 km. We simulate body wave propagation in the three-dimensional (3-D) P wave model using 2-D finite difference (FDM) and 3-D spectral element (SEM) methods. As measured by cross correlation between synthetics and data, the P wave model typically explains about half of the traveltime anomaly and some of the waveform complexity but fails to predict the extended SH wave train. In this study we take advantage of the densely distributed Hi-net stations and use 2-D FDM modeling to simulate the P-SV and SH waveforms. Our 2-D model suggests that a thin, elongated low-velocity zone exists atop the slab, extending down to a depth of 300 km with an S wave velocity reduction of 14% if a thickness of 20 km is assumed. Further, 3-D SEM simulations confirm that this model explains a strong secondary arrival which cannot easily be imaged with standard tomographic techniques. The low-velocity layer could explain the relatively weak coupling associated with most subduction zones at shallow depths (<50 km), generally involving abundant volcanic activity and silent earthquakes, and it may also help to further our understanding of the water-related phase transition of ultramafic rocks, and the nature of seismicity at intermediate depths (∼70–300 km).