5:15 PM - 7:15 PM
[HDS07-P17] Changes in slowness before and after rainfall detected by an active seismic array monitoring test using chirp signals on a slope
Keywords:Active seismic monitoring, Slowness, Surface failure, Rainfall
Slope surface failure can be triggered by rainfall or snowmelt. Seismic monitoring is one of the possible tools to understand the failure process and develop early warning systems (e.g., Whiteley et al., 2019). Mainsant et al. (2012) measured seismic ambient noise during the failure due to rainfall by embedding two seismometers across a slope. They found that the velocity gradually decreased due to rain and subsequent groundwater elevation, followed by a sharp decrease occurring before the initiation of a major failure. However, it is pointed out that the propagation function (Green’s function) is not obtained if the excitation source distribution is anisotropic or changed seasonally (Stehly et al., 2006). Nakayama et al. (2022) embedded a speaker as a stable elastic wave source and accelerometers on a slope and performed an active seismic monitoring test during a period including rainfall. They repeatedly applied linear sweep (chirp) signals (10-310 Hz) to the speaker (Fig. 1). They found that the phase difference from the nearest receiver (AC9) changed after the rainfall. In this study, we analyzed their data and estimated changes in slowness (reciprocal of phase velocity) before and after the rainfall.
We extracted the phase difference of the receivers from AC9 for a thousand shots from the periods before and after the rainfall, respectively. We stacked them every 100 shots to improve the signal-to-noise ratios and obtained ten datasets. We calculated the residual of the phase difference at any given slowness and the weighted sum of the residual for all receivers and all datasets to find candidates of slowness. We used the reciprocal of the root square of the spectral amplitude of stacked datasets as weight. Assuming that slowness was spatially uniform and a little dispersive, we chose three candidates from data before and after the rainfall, respectively. We checked the phase difference for each receiver for each candidate. We estimated the average slowness for each receiver before and after the rainfall, respectively, by accumulating the product of the residual and the weight in an analyzed frequency range (Fig. 2). The average slowness increased by 5% (the average velocity decreased by 4%) after the rainfall, which may be caused by the decrease in coupling among particles due to water infiltration.
Acknowledgments: This work was supported by JSPS KAKENHI Grant Numbers JP15H02996 and 26750135. We used the meteorological data from JMA.
We extracted the phase difference of the receivers from AC9 for a thousand shots from the periods before and after the rainfall, respectively. We stacked them every 100 shots to improve the signal-to-noise ratios and obtained ten datasets. We calculated the residual of the phase difference at any given slowness and the weighted sum of the residual for all receivers and all datasets to find candidates of slowness. We used the reciprocal of the root square of the spectral amplitude of stacked datasets as weight. Assuming that slowness was spatially uniform and a little dispersive, we chose three candidates from data before and after the rainfall, respectively. We checked the phase difference for each receiver for each candidate. We estimated the average slowness for each receiver before and after the rainfall, respectively, by accumulating the product of the residual and the weight in an analyzed frequency range (Fig. 2). The average slowness increased by 5% (the average velocity decreased by 4%) after the rainfall, which may be caused by the decrease in coupling among particles due to water infiltration.
Acknowledgments: This work was supported by JSPS KAKENHI Grant Numbers JP15H02996 and 26750135. We used the meteorological data from JMA.