1 IAG-IASPEI J Secular and co-seismic changes of the seismic velocity in the Tokai region detected by ACROSS Shuhei Tsuji 1) , Ryoya Ikuta 2) , Koshun Yamaoka 1) , Takahiro Kunitomo 1) , Toshiki Watanabe 1) , Yasuhiro Yoshida 3) , Akio Katsumata 3) 1) Nagoya University , 2) Shizuoka University, 3) Japan Meteorological Agency I am Shuhei Tsuji, of Nagoya University. I’d like to talk about this study. The title is Secular and co-seismic changes in seismic velocity in the Tokai region detected by ACROSS. We have discovered secular and co-seismic temporal velocity changes with anisotropy in the Tokai region by artificial seismic source. 16:45-17:00 Thursday, August 3, 2017
2 Introduction of ACROSSACROSS generate elastic waves which has quite accurate phase and amplitude with centrifugal force by rotating eccentric mass Made of a motor and an eccentric mass High stability Linear oscillation We continuously control the phase and the frequency of the eccentric mass We can continuously generate quite accurate elastic waves We rotate the eccentric mass clockwise and counterclockwise alternately We can synthesize linear oscillation At first, I’ll introduce ACROSS. Across is an artificial seismic source, that generate elastic waves with centrifugal force by rotating eccentric mass. It is composed of a motor and an eccentric mass. This is the photo of the motor and this is the photo of the eccentric mass. ACROSS has 2 characteristic point. One is high stability. We continuously control the phase and the frequency of the eccentric mass So, we can continuously generate quite accurate elastic waves. The other characteristic point is linear oscillation. We rotate the eccentric mass clockwise and counterclockwise alternately. So, we can synthesize linear oscillation.
3 The ACROSS source and Hi-net stationsStudy area The ACROSS source and Hi-net stations :ACROSS : Hi-net stations Map around Japan Next, I’ll show the location of the study area. This is a map around Japan and red square shows the location of the Tokai region. The Tokai region is located in the central part of Japan. Our ACROSS source is located middle part of the Tokai region. This is a map around the source. Star shows location of ACROSS source, and green dot shows location of the Hi-net stations. In this study, we use Hi-net as receivers. Hi-net is a seismograph network managed by NIED. We use Hi-net as receivers :Tokai region Hi-net: seismograph network managed by NIED
4 Experimental conditionsFrequency: 3.5 – 7.5 [Hz] Processing term: 2006/08/ /9/14 We can detect the ACROSS signal in 150 km from the source! We use daily transfer function in near station with high S/N In this study, we generate 3.5 to 7.5 Hz elastic waves by ACROSS, and we analyzed transfer function from 16th Aug 2006 to 14th Sep 2014 This is the stacked transfer function of all signals at each station. Horizontal line shows time from oscillation and vertical one shows distance and normalized amplitude. As shown from this figure, we can detect ACROSS signal as far as 150km from the ACROSS source, when we stack all transfer functions in the processing term. But in this study, we tried to monitor daily transfer functions. So we use near stations with High Signal to Noise ratio. This is the map of the used stations in this study. Stacked transfer functions of all data
5 Processing to calculate temporal variations1.Calculate reference wave average the transfer function of the all dates at each station 2.Select direct S wave Select direct S wave with Hanning window appropriate length. 3.Remove artificial change Owing to Data logger ・Replacement ・Clock error (Kunitomo, 2014, ZISHIN) ・Malfunction Next I’ll talk about how to calculate temporal variations of the transfer functions. This slide shows the processing to calculate temporal variations. First, we calculate reference waves for each station by average the transfer functions of all dates at each station. Second, we select the direct S wave from reference wave with hanning window appropriate length. This is reference wave and selected S wave of each station. Horizontal line shows time from oscillation, and vertical one shows distance of the stations and normalized amplitude of reference waves. Third, we remove artificial effects, due to replace of the datalogger, clock error of the datalogger and trouble of the datalogger. Then, we calculate daily variation of the transfer functions. 4.Calculate daily variation of the transfer function Reference transfer functions and selected S wave (Normalized with max amplitude of each station)
6 Processing to calculate temporal variations4.Calculate daily variation of transfer function 𝐷 𝑘,𝑙 = 𝐺 𝑘,𝑙 𝐺 𝑘 𝐺 𝑘 : Reference transfer function 𝐺 𝑘,𝑙 : Daily transfer function Travel time change in N.MRIH station (the nearest station D=2.91 [km]) Received in Transverse – Oscillate in Radial 𝛿 𝑡 𝑙 = 1 𝑘 𝑊 𝑘 𝑘 𝑊 𝑘 arg 𝐷 𝑘,𝑙 𝜔 𝑘 𝜔 𝑘 : Angular frequency 𝛿𝑡 𝑙 : Daily travel time change k,l : Frequency number, Date number 𝑊 𝑘 = 𝐺 𝑘 To detect variation of transfer functions, At first, we deconvolved the daily transfer functions with the reference transfer function in the frequency domain. This is travel time change calculated from this equation at Mori station, which is nearest station from ACROSS source. Horizontal line show date, and vertical line shows daily travel time changes. year
7 Secular and co-seismic changesTravel time change in N.MRIH station (the nearest station D=2.91 [km]) Received in Transverse – Oscillate in Radial 2011/03/11 Tohoku-Oki Earthquake This is the same figure as the previous slide. From this figure, we can find sudden delay at the time of the Tohoku-Oki earth quake. year
8 Secular and co-seismic changesTravel time change in N.MRIH station (the nearest station D=2.91 [km]) Received in Transverse – Oscillate in Radial 2011/03/11 Tohoku-Oki Earthquake And secular changes can be also find, like this red arrow. year
9 Secular and co-seismic changes at the several stationsN.TNRH (Δ=9.79[km]) N.KNEH (Δ=17.4[km]) year N.KGWH (Δ=8.40[km]) And the similar pattern can be seen in the all stations. Here, we show the example of the travel time changes at the several stations. No data year
10 Secular and co-seismic changesTravel time change in N.MRIH station (the nearest station D=2.91 [km]) Received in Transverse – Oscillate in Radial Secular change Co-seismic change At the time of the Tohoku-Oki earth quake We characterized this pattern by the secular change and the co-seismic change at the time of the 2011 Tohoku-Oki earth quake. year
11 Estimate secular and co-seismic changes𝛿𝑡 𝑙 =𝑎 𝑇 𝑙 +𝑏∙𝐻 𝑇 𝑙 − 𝑇 𝑒𝑞 +𝑐 year Model: 𝑀 𝑘,𝑙 ≡ 𝐴 𝑘,𝑙 𝑒 𝑖 𝜔 𝑘 𝛿 𝑡 𝑙 𝐴 𝑘,𝑙 ≡1 𝑀 𝑘,𝑙 : Model of daily deconvolution 𝛿𝑡 𝑙 : Model of travel time change H : The Heaviside step function k,l : Frequency number, Date number To estimate the secular change and the co-seismic change, we make a model of travel time changes as this equation. Using previous model of travel time changes, make a model of transfer functions as this equation. And in this study we define amplitude as 1. Using this two, we estimate the pattern with nonlinear least squares method so that to minimize the variance. Estimate a, b and c of 𝛿𝑡 𝑙 in Complex plane so that 𝐷 𝑘,𝑙 − 𝑀 𝑘,𝑙 →𝑚𝑖𝑛.
12 Results of the estimationSH wave at N.MRIH by radial excitation :daily data :estimated Polarization anisotropy year Sv wave at N.MRIH by transverse excitation This is a result of estimation. We show the result of SH wave and SV wave at the MORI station. Then, we found the difference between the component of transfer function. So, we consider the difference as a polarization anisotropy. year
13 Estimation of anisotropy in transfer functionsCalculate transfer function in each combination of the source azimuth q and the receiver azimuth f by 5 deg. N Receiver f Uf (ω) Source Gq, f (w)= E Fq(ω) n q To estimate the direction of anisotropy, We calculate the transfer functions in each combination of the azimuth of the source oscillation and the azimuth of the receiver oscillation by 5 degrees. And we estimate the secular and the co-seismic changes for each combination with previous method 0°≦ q ≦180° 0°≦ f ≦180° e Estimate the secular and the co-seismic changes for each combination with previous method
14 Daily and estimated travel time changes (azimuth: 0°, 45°, 90°, 135°)Travel time changes of several combination of the transfer functions with azimuth of the source and the receiver Daily and estimated travel time changes (azimuth: 0°, 45°, 90°, 135°) receiver azimuth [deg.] source azimuth [deg.] These are example of travel time changes of several combination of the transfer functions with azimuth of the source and the receiver. In each figure, blue dot shows daily travel time changes and red line shows the estimated travel time change. There are 16 example of the travel time change, whose azimuth are 0, 45, degrees for both direction.
15 Estimate principal axes of anisotropyThe effects by the source and by the receiver are independent Average these changes by the source and the receiver azimuth, respectively. And estimate principal axes by fitting with sinusoidal function advance [s] This is the secular changes at Mori station of all combination of receiver azimuth and source azimuth. Horizontal line shows receiver azimuth and vertical one shows source azimuth. Color show the magnitude of advance. We think that the effect by the source and by the receiver are independent. Then, we average these changes by the azimuth of the source, and by the azimuth of the receiver, respectively. And estimate principal axes by fitting with sinusoidal function.
16 Estimate principal axes of anisotropyThe effects by the source and by the receiver are independent Average these changes by the source and the receiver azimuth, respectively. And estimate principal axes by fitting with sinusoidal function This is the principal axes in the receiver azimuth, which is estimated by the averaged travel time changes by the receiver azimuth. This may show the anisotropy around the receiver. Horizontal line show the receiver azimuth, and vertical one shows advance of the travel time change. Blue line with error bar shows averaged travel time changes and brown one shows fitting line. Red and blue vertical line shows estimated principal axes. Receiver azimuth This may show the anisotropy around the receiver Max axis : Data :Fitting Min axis
17 Estimate principal axes of anisotropySource azimuth This may show the anisotropy around the source The effects by the source and by the receiver are independent Average these changes by the source and the receiver azimuth, respectively. And estimate principal axes by fitting with sinusoidal function Min axis Max axis And this is the principal axes in the source azimuth, which is estimated by the travel time changes averaged by source azimuth. This may show the anisotropy around the source. Like this way, we estimate principal axes of the anisotropy for each station. Receiver azimuth This may show the anisotropy around the receiver Max axis : Data :Fitting Min axis
18 Estimated principal axes of the secular changesΔ=5.33[km] Δ=8.40[km] Δ=9.79[km] Δ=21.2[km] Δ=32.6[km] Δ=17.4[km] This is estimated principal axes of the secular changes of the all stations except Mori station.
19 Estimated principal axes of the co-seismic changesΔ=5.33[km] Δ=8.40[km] Δ=9.79[km] Δ=21.2[km] Δ=32.6[km] Δ=17.4[km] And this is estimated principal axes of the Co-seismic changes of the all stations except Mori station.
20 Axes of the anisotropy around receiversTo show the tendency of the anisotropy around receiver, we plot cross mark at the receiver location. Secular change around receivers Co-seismic change around receivers Finally we plot principal axes of the anisotropy on the maps. This slide shows the magnitude and direction of principal axes of the secular and the co-seismic changes around receivers. To show the tendency of the anisotropy around receiver, we plot cross mark at the receiver location. The direction of the axis of the cross mark means principal axes of each stations. Blue cross shows delay and Red cross shows advance. The length of axis means magnitude of advance or delay. As shown from this slide, the sense of the changes has tendency. The secular changes are advance at the most of the stations and the co-seismic changes are delay. In the other hand, direction of the anisotropy has no systematic tendency. Delay Secular changes: advance Co-seismic changes: Color :Advance or Delay Length of axis: Magnitude of the travel time changes Direction: Axis of the direction of the anisotropy at each station
21 Axes of the anisotropy around the SourcesTo show the tendency of the anisotropy around receiver, we plot cross mark at the receiver location. Secular change around the source Co-seismic change around the source Same direction This is the result of the secular and the co-seismic changes around the source. To show the tendency of the direction of the anisotropy around the source, we plot all cross marks at the source location for the all stations. As shown from this slide, secular advance and co-seismic delay are also found. And in the co-seismic change around the source, the direction of the principal axis has systematic tendency. Secular changes: advance Delay Co-seismic changes: Color :Advance or Delay Length of axis: Magnitude of the travel time changes Direction: Axis of the direction of the anisotropy at each station
22 Conclusion AcknowledgementsWe analyze ACROSS signals received by Hi-net stations from 2006/08/16 to 2014/9/14 in Tokai region, Japan. We discover both secular advance and co-seismic delay in travel time with anisotropy. Travel times advanced with time in most of the stations. Travel times delayed at the 2011 Tohoku-Oki EQ in all stations. There are no tendency in the direction of the anisotropy except for the co-seismic delay around the source. Acknowledgements This is conclusion. Thank you for your attenuation. We use continuous seismic data in Hi-net, National Research Institute for Earth Science and Disaster Resilience.