Verification of Green's Function Approximation from Ambient Noise Cross-correlation using Three-year Continuous Seismic Data in South Niigata Prefecture, Japan

A continuous broadband-to-strong motion observation network that consists of 15 stations has been in operation since September 2011 in south Niigata Prefecture (Yoshimi et al., 2012). We applied seismic interferometry to the ambient noise data for the purpose of validating three-dimensional S-wave velocity structure (deep sedimentary structure) models beneath the observation area (Hayashida and Yoshimi, 2012, 2015). We used 37-month ambient noise data (October 2011 - October 2014) to obtain nine component (RR, TT, ZZ, RT, TR, ZR, ZT, RZ, and TZ) cross-correlation functions between two stations (105 station pairs) for distances from 4.3 to 40.7 km, according to the signal processing technique by Bensen et al. (2007). Our results show that signal-to-noise ratios (SNRs) of cross-correlation functions increase logarithmically as the stacking number increase and the wave trains are clear especially for TT and ZZ correlations. Some station pairs show large SNRs even for RT and TR correlations, indicating complicated velocity structure beneath the area. To extract Rayleigh-wave Green's functions efficiently, we applied body and surface wave separation technique by Takagi et al. (2014). We obtained clear dispersion curves of Rayleigh and Love waves in the frequency range between 0.1 and 1.0 Hz that correspond well to theoretical curves from existing velocity structure model of Sekiguchi et al. (2009). We found that maximum detectable wavelength of surface wave is nearly equivalent to station-to-station distance for near station pairs (< 20 km), whereas the maximum wavelength is about a half of station-to-station distance for further station pairs. We also investigated the effectiveness of uniform noise source distribution assumptions by comparing hourly stacked complex coherence functions of ambient noise between neighboring two stations and theoretical Bessel functions of the of first kind of zero order with the existing crustal (Takeda et al., 2012) and deep sedimentary structure models (Sekiguchi et al., 2009). The comparisons show good agreements in the frequency range between 0.05 and 0.3 Hz, indicating uniform noise source distribution in the observation period. Acknowledgements: Continuous observations until March 2014 were funded by Japan Nuclear Safety Organization (JNES).