14:45 〜 15:00
[SSS09-15] Extraction of P-wave reflections of mantle discontinuities from ambient seismic noise beneath Japan
キーワード：地震波干渉法
Seismic interferometry (SI) is the technique that extracts the Green's function between a station pair by calculating the cross-correlation function (CCF) of the random seismic wavefield. Microseisms, which are excited by ocean swell activities, and earthquake coda, multiple scattered by the heterogeneous structure, are typical examples of a random seismic wavefield. Because surface waves dominate microseisms, the extracted surface waves are used to infer the S wave velocity structure of the crust and upper mantle (e.g., Shapiro et al., 2005). Recent studies also start to utilize information on the body waves as well as the surface waves extracted by SI. For example, some groups reported reflected P waves at the 410/660 km discontinuities (P410P/P660P) (Poli et al., 2012; Feng et al., 2017) using the continental stations. This study aims to extract the P410P/P660P by applying SI to microseisms and earthquake coda, recorded by Hi-sensitivity Seismograph Network Japan (Hi-net), and discuss the effect of the source distribution of wavefield on the retrievals.
The procedures of the CCF calculation of microseisms are as follows: The dataset in this study is the vertical components of 240 Hi-net stations in Southwest Japan and 168 stations in Northeast Japan (2007–2018). We down-sampled the original 100 Hz data to 2 Hz. The 1-day-length waveform was split into 1024-second-length segments and selected based on the mean square amplitude in 0.05-0.1 Hz and 0.1–0.2 Hz. We then calculated CCFs for every pair of stations by stacking over all the selected segments with spectral whitening from 0.1 to 0.2 Hz.
The method of the CCF calculation of earthquake coda is mostly the same as that of microseisms. The time-windows are the 1500-4800 s from the origin time of large earthquakes (Mw>6.5; 534 events). The running absolute mean method (Bensen et al., 2007) was applied to all earthquake coda waveforms.
In SW Japan, both CCFs of microseisms and CCFs of earthquake coda show P410P, but no or weak P660P. The lack of extraction of P660P despite CCFs calculated from independent wavefield suggests that the structure near 660 km discontinuity in SW Japan makes the extraction of P660P difficult. The CCFs of earthquake coda show P410P/P660P in NE Japan, partly because of the coincidence between the arrival direction of wavefield and the direction of stations, suggesting that many station pairs in NE Japan have incident waves with the adequate direction for the extraction. The extraction of P410P/P660P is limited by the distribution of wave sources and the geometry of stations, especially from microseisms.
As another analysis for obtaining the information of mantle discontinuities from microseisms, the receiver function analysis of microseismic P wave was performed. We focused on the P wave excited by the weather bomb in the Atlantic Ocean (Nishida and Takagi, 2016). For the estimation of the source time function, all the vertical components were stacked over all the stations after the back-propagations to the source. The deconvolution of the radial component with the source time function was calculated using 691 Hi-net stations. The preliminary result shows the extraction of converted S waves by 410 km discontinuity (P410s) beneath Japan. However, the result shows no or weak peak corresponding to P660s. The extraction of P410s and no or weak P660s is consistent with the result of SI, which shows the P410P and weak P660P.
The procedures of the CCF calculation of microseisms are as follows: The dataset in this study is the vertical components of 240 Hi-net stations in Southwest Japan and 168 stations in Northeast Japan (2007–2018). We down-sampled the original 100 Hz data to 2 Hz. The 1-day-length waveform was split into 1024-second-length segments and selected based on the mean square amplitude in 0.05-0.1 Hz and 0.1–0.2 Hz. We then calculated CCFs for every pair of stations by stacking over all the selected segments with spectral whitening from 0.1 to 0.2 Hz.
The method of the CCF calculation of earthquake coda is mostly the same as that of microseisms. The time-windows are the 1500-4800 s from the origin time of large earthquakes (Mw>6.5; 534 events). The running absolute mean method (Bensen et al., 2007) was applied to all earthquake coda waveforms.
In SW Japan, both CCFs of microseisms and CCFs of earthquake coda show P410P, but no or weak P660P. The lack of extraction of P660P despite CCFs calculated from independent wavefield suggests that the structure near 660 km discontinuity in SW Japan makes the extraction of P660P difficult. The CCFs of earthquake coda show P410P/P660P in NE Japan, partly because of the coincidence between the arrival direction of wavefield and the direction of stations, suggesting that many station pairs in NE Japan have incident waves with the adequate direction for the extraction. The extraction of P410P/P660P is limited by the distribution of wave sources and the geometry of stations, especially from microseisms.
As another analysis for obtaining the information of mantle discontinuities from microseisms, the receiver function analysis of microseismic P wave was performed. We focused on the P wave excited by the weather bomb in the Atlantic Ocean (Nishida and Takagi, 2016). For the estimation of the source time function, all the vertical components were stacked over all the stations after the back-propagations to the source. The deconvolution of the radial component with the source time function was calculated using 691 Hi-net stations. The preliminary result shows the extraction of converted S waves by 410 km discontinuity (P410s) beneath Japan. However, the result shows no or weak peak corresponding to P660s. The extraction of P410s and no or weak P660s is consistent with the result of SI, which shows the P410P and weak P660P.