Seismological Features of the Subducting Slab Beneath the Kii Peninsula, Central Japan, Revealed by Receiver Functions

We report seismological evidence that the subducting Philippine Sea slab (PHS) beneath the Kii Peninsula, central Japan, can be divided into three segments. Offshore the Kii Peninsula, the "Tonankai" and "Nankai" fault segments suffer mega-thrust earthquakes that repeat every 100 to 150 years. The structure of the young, thin, contorted PHS is important to the seismo-tectonics in this region. We apply the receiver function (RF) analysis to 26 Hi-net short-period and 4 F-net broad-band seismographic stations. In the case that dipping velocity discontinuities and/or anisotropic media exist beneath seismometer, both radial RFs and transverse RFs contain useful information to estimate underground structure. For isotropic media with a dipping-slab interface, back- azimuthal variation in RFs depends largely on three parameters, the downdip azimuth, dip angle and the depth of the interface. We stack both radial and transverse RFs with allowance a time-shift caused by the dipping interface, searching for optimal parameters based on the grid-search technique at each station. At some stations located near the eastern coastline of the Kii Peninsula, the dip angle of the interface inferred from RF stacking is much steeper than that estimated by the local seismicity. This discrepancy arises from the interference of two slab-converted phases, suggesting a layer atop the slab. In these cases we refine the stack to distinguish two slab phases and estimate three parameters of each dipping interface separately. Two interfaces with the same dip direction and low dip angle are estimated at these stations, with depth difference near 6 km. Thus, the shallower interface may be related to the layer within the oceanic crust and the deeper one is the slab Moho. These double-layered interfaces are detected only at stations located up-dip of a belt-like distribution of non- volcanic low-frequency tremor. Comparing the interface dips estimated in this study with the direction of slab motion determined by the GPS observation, we can classify the slab beneath the Kii Peninsula into three segments: eastern, central and southern. Within the eastern part, the RF-estimated dip direction is shifted about 30° clockwise from the GPS- estimated slab motion, and the dip angle is small. The difference of RF-estimated slab dip and GPS-estimated slab motion is small in the central part. Discrepancy between slab dip and motion estimates increases to 40°--90° in the western part. At the some stations located in the western and eastern part of the Kii Peninsula, the dip angle of the interface is much steeper than another model estimated by the local seismicity. Since the local seismicity distribution reflects well the larger-scale geometry of the slab, the RF-estimated dip angle may imply anisotropic wavespeeds within strongly sheared media above the slab. The boundary between the eastern and central regions of the Kii Peninsula corresponds to the region where (1) a topographic high in the slab interface is subducting and (2) the intraslab seismicity shows double-layered activity. The boundary between central and southern regions is coincident with the segment boundary of megathrust earthquakes in the Nankai region. The structural features revealed by RF-stacking may be an important key to the seismotectonics around the Kii Peninsula.