S-wave anisotropy estimated by seismic interferometry using ambient noise record in the Nankai Trough subduction zone, Japan

In the Nankai Trough subduction zone, located beneath the Pacific Ocean off the southeast coast of Japan, interplate earthquakes can be generated repeatedly in association with stress accumulation and release cycle. In this study, we aim to obtain the information of S-wave anisotropy beneath the seafloor, which could be interpreted as a proxy of stress and strain field above the subduction zone. For this purpose, we apply the seismic interferometry technique to ambient noise records acquired by seafloor and subseafloor seismometers deployed above the Nankai Trough subduction zone. In this area, we have twenty seafloor seismometers as a part of DONET (Dense Oceanfloor Network System for Earthquake and Tsunamis) and a borehole seismometer installed in the IODP (Integrated Ocean Drilling Program) C0002G observatory at the bottom of the borehole, 900 m below seafloor. Both observatories were designed and installed to monitor the seismic activity and the process of earthquake generation including the stress accumulation. In this study, we apply the seismic interferometry to ambient noise records observed by these DONET and C0002G seismometers. Seismic interferometry is a method to retrieve the impulse response by the cross-correlation of seismic records simultaneously acquired by the two seismometers. Because the horizontal components are dominated by S-wave energy, we expected that auto- and cross-correlation functions (ACF and CCF), calculated from the horizontal components of each seismometer, would provide us the knowledge of S-wave velocity and anisotropy beneath seafloor, as a proxy of strain and stress field, and fluid migration above the plate boundary. We obtained zero offset 4-C ACF and CCFs comprising V11, V12, V21, and V22, calculated form continuous ambient noise records observed by horizontal components of each seismometer. Vij are ACF and CCFs calculated from ambient noise record observed by i- and j-direction receiver components, and represents impulse response which has i-direction source and j-direction receiver of each seismometer. We used each 1 hour dataset for more than 6 months and obtained Vij as 30 s zero offset impulse responses for each seismometer. In the obtained ACF and CCFs, several coherent events are visible. However, the events in each component are not consistent with that of others. It might result from S-wave splitting affected by anisotropy. S-wave split into two orthogonal directions along anisotropy direction in propagating anisotropy media. We then applied the Alford rotation and the layer stripping method to the obtained 4-C ACF and CCFs to estimate S-wave anisotropy direction and amplitude beneath each seismometer in each layer, shallow sediment and accretionary prism above the plate boundary. Obtained results, including the azimuth and magnitude of anisotropy as functions of depth, show good agreement with S-wave anisotropy directions and principle shear stress directions estimated from two of the other methods, i.e., borehole breakout analysis in the IODP C0009 borehole, and P-S converted wave analysis using airgun OBS data. We expect that our method could make it possible to monitor temporal changes in the azimuth and the magnitude in anisotropy, as a proxy of stress field, using real-time ambient noise records in the subduction zone.