5:15 PM - 6:30 PM
[SSS09-P02] Estimation of instrumental clock errors at 46 active volcanoes in Japan based on seismic interferometry
Keywords:seismic interferometry, clock error, volcano monitoring
The correct absolute timing of seismograms is crucially required in many seismological analyses. In recent years, seismic interferometry [e.g., Curtis et al., 2006; Shapiro, 2004] has been widely used to detect structural changes related to large earthquakes and volcanic activities [e.g., Brenguier et al., 2008; Wang et al., 2019; Hirose et al., 2020; Nishida et al., 2020]. Some previous studies have investigated the use of seismic interferometry to correct seismometer clocks [e.g., Stehly et al., 2007; Sens-Schönfelder, 2008; Hable et al., 2018]. Estimating instrumental clock errors are also important for continuous seismic monitoring of active volcanoes. I develop an estimation method of instrumental clock errors based on seismic interferometry and apply this method to 46 active volcanoes in Japan.
I computed seismic ambient noise cross-correlation functions (CCFs) using continuous seismograms at V-net and JMA stations of 46 active volcanoes in Japan between Jan. 1, 2020, and Jan. 31, 2021. Continuous seismic records on vertical component were divided into 10-minutes-long segments and applied spectral whitening [e.g., Shapiro et al., 2006; Bensen et al., 2007] and one-bit normalization [Larose et al., 2004]. Daily CCFs (DCCFs) were obtained by stacking CCFs every 10 minutes. I measured delay times between a reference 3-days-stacked CCF and a current 3-days-stacked CCF for each station pair and lag time at a 0.2-4 Hz band using the windowed cross-correlation method [e.g., Snieder et al., 2012]. A clock error will cause the same delay times between the reference and current CCFs at all lag times. Therefore, I fitted a straight line to the delay times at each lag time and assumed the straight line's intercept as the clock difference for that station pair. On the straight-line fitting, I used L1 norm optimization to suppress the effect of outliers. Finally, I estimated clock errors at each seismic station by linear least-squares inversion with the estimated clock differences.
In this study, clock errors were estimated for the time period between Jan. 1, 2020, and Jan. 31, 2021. Significant clock errors were detected at some volcanoes: At Nasu volcano, the clock of station N.NSOV started to drift with about 0.009 s/day in early Jan. 2021. This drift was caused by a malfunction of the GPS antenna from Jan. 8, 2021. At Asama volcano, a clock jump of about 1.0 s occurred at station V.ASME during Apr.-Jun. 2020, and the clock of station V.ASMG had a drift of about 0.005 s/day during Aug.-Oct. 2020. At Miyakejima volcano, the clock of station V.MYKO repeatedly drifted with 0.002 to 0.006 s/day during May-Jul. 2020, Aug.-Nov. 2020, and Dec. 2020-Jan. 2021. Repeated clock drifts (0.003-0.096 s/day) also occurred at Kusatsu-Shirane volcano. Maximum clock errors of station V.KSHA at Kusatsu-Shirane were estimated to be 1.55 s in Aug. 2020 and 0.84 s in Nov. 2020. Estimated clock errors in this study could improve the results of various seismological analyses at the 46 volcanoes, for example, hypocenter determination of volcanic earthquakes and seismic interferometry analysis.
Acknowledgment: I used seismograms recorded by Japan Meteorological Agency (JMA).
I computed seismic ambient noise cross-correlation functions (CCFs) using continuous seismograms at V-net and JMA stations of 46 active volcanoes in Japan between Jan. 1, 2020, and Jan. 31, 2021. Continuous seismic records on vertical component were divided into 10-minutes-long segments and applied spectral whitening [e.g., Shapiro et al., 2006; Bensen et al., 2007] and one-bit normalization [Larose et al., 2004]. Daily CCFs (DCCFs) were obtained by stacking CCFs every 10 minutes. I measured delay times between a reference 3-days-stacked CCF and a current 3-days-stacked CCF for each station pair and lag time at a 0.2-4 Hz band using the windowed cross-correlation method [e.g., Snieder et al., 2012]. A clock error will cause the same delay times between the reference and current CCFs at all lag times. Therefore, I fitted a straight line to the delay times at each lag time and assumed the straight line's intercept as the clock difference for that station pair. On the straight-line fitting, I used L1 norm optimization to suppress the effect of outliers. Finally, I estimated clock errors at each seismic station by linear least-squares inversion with the estimated clock differences.
In this study, clock errors were estimated for the time period between Jan. 1, 2020, and Jan. 31, 2021. Significant clock errors were detected at some volcanoes: At Nasu volcano, the clock of station N.NSOV started to drift with about 0.009 s/day in early Jan. 2021. This drift was caused by a malfunction of the GPS antenna from Jan. 8, 2021. At Asama volcano, a clock jump of about 1.0 s occurred at station V.ASME during Apr.-Jun. 2020, and the clock of station V.ASMG had a drift of about 0.005 s/day during Aug.-Oct. 2020. At Miyakejima volcano, the clock of station V.MYKO repeatedly drifted with 0.002 to 0.006 s/day during May-Jul. 2020, Aug.-Nov. 2020, and Dec. 2020-Jan. 2021. Repeated clock drifts (0.003-0.096 s/day) also occurred at Kusatsu-Shirane volcano. Maximum clock errors of station V.KSHA at Kusatsu-Shirane were estimated to be 1.55 s in Aug. 2020 and 0.84 s in Nov. 2020. Estimated clock errors in this study could improve the results of various seismological analyses at the 46 volcanoes, for example, hypocenter determination of volcanic earthquakes and seismic interferometry analysis.
Acknowledgment: I used seismograms recorded by Japan Meteorological Agency (JMA).