[S-SS30_29PO1] New perspective of great earthquakes along subduction zones

[SSS30-P15] Receiver function analysis using OBS data: modeling 3-D structure of the Philippine Sea plate off the Kii Peninsula

*Takeshi AKUHARA¹, Kimihiro MOCHIZUKI¹ (1.Earthquake Research Institute, University of Tokyo)

Keywords:ocean-bottom seismometer, receiver function, subduction zone

Megathrust earthquakes have repeatedly occurred beneath the southwestern Japan, on the subducting Philippine Sea plate, in cycles of 100-150 years [Ando, 1975]. The rupture boundary of the latest two megathrust earthquakes, the 1944 Tonankai and 1946 Nankai earthquakes, is located at the south of the Kii Peninsula. Although some structural heterogeneity was proposed as factors of the rupture boundary [Mochizuki et al., 1998; Kodaira et al., 2006], the question of why rupture propagation stops there is still open in light of our little knowledge about 3-D geometry of the subducting Philippine Sea plate at offshore region.In this study, we aim to construct 3-D structure model of the subducting Philippine Sea plate by receiver function (RF) analysis, using data of ocean-bottom seismometers (OBSs) deployed from 2003 to 2007 off the Kii Peninsula [Mochizuki et al., 2010; Akuhara et al., 2013]. These OBSs have three-component velocity sensors with natural frequency of 1 Hz, and their orientations were determined in this study from particle motion of regional P-wave. The difficulty of our RF analysis using OBS data is summarized by the following two factors. The first is that noise is dominant within a low-frequency band (1 < Hz), the most stable band for estimating RFs. The second is that the number of teleseismic events is limited because of short observation periods and low S/N ratio.To overcome these problems, we calculated RFs with the aid of multi-taper correlation (MTC) method [Park and Levin, 2000]. The method is resistant to spectral leakage and able to estimate frequency-dependent uncertainties for RF, which is suitable for noisy OBS data and for high-frequency analysis. We binned resultant RFs by back azimuths, and computed time-domain uncertainties of the RFs from the frequency-dependent uncertainties estimated by the MTC method, using jackknife resampling within each back azimuth bin [Leahy and Collins, 2009]. This uncertainty estimation makes the following phase identification more reliable, even though the number of teleseismic events is limited. Our preliminary results show some coherent peaks throughout all back azimuth bins, whose amplitude is larger than one-sigma uncertainties. Some of them have moveout, implying existence of dipping layers, and have arrival times roughly corresponding to the depth of the slab mantle. Although more detailed identification process for these peaks is largely left for our future work, these peaks might be converted phases from the slab mantle.