9:30 AM - 9:45 AM
[SSS07-03] On power spectrum of seismic noise excited by a streamflow
Keywords:streamflow, seismic noise, Hi-net, power spectrum
Seismometers can detect not only earthquakes but also various kind of vibrations caused by natural phenomena and social activities. Vibrations caused by streamflow is one of them; which is especially significant in the upstream basins, where high flow speed tends to excite strong oscillation power. This property of streamflow-origin noise is considered to be useful for remote monitoring of river conditions in the upstream basins (e.g., P.C. and Sawazaki, 2021). However, for the effective use of seismograms for the monitoring of streamflow, it is necessary to clarify physical mechanism of seismic noise excitation due to streamflow. For this purpose, in this study, we have investigated the characteristics of the power spectrum of seismic noise excited by a streamflow using Hi-net seismic stations located close to rivers.
We selected in total 238 Hi-net up-down component seismograms that covers the period from October 10 to 17, 2019, when the 2019 Typhoon Hagibis passed through eastern Japan. First, for each Hi-net record, we calculated the power spectral density at 2 min and 0.2 Hz intervals in the time and frequency domains, respectively. Next, in order to eliminate transient signals due to earthquakes and monotonic vibrations at each Hi-net station, we calculated the minimum value of the power spectral density in a time-frequency window of 10 min and 1 Hz widths. Then we calculated the vibration power every hour for the frequency ranges of 2-4Hz, 4-8Hz, 8-16Hz, and 16-32Hz, in which the influence of microseisms is considered to be small.
Then, each Hi-net record is paired with the available water level records obtained in the basin where the target Hi-net station is located. We assumed the relationship between the water level H and the vibration power W given by
H=A(W-W0)b+H0 (W>W0)
H=H0 (W<W0)
and estimate parameters A and b through the least square method. W0 and H0 are vibration power and water level before the flood, which are computed using the average value of W and H during October 10 (calm period before the Typhoon Hagibis approaches). To evaluate the degree of agreement between the calculated and observed water levels, we used the Nash-Sutcliffe index (NSE: Nash and Sutcliffe, 1970).
As a result, there were 21 Hi-net stations which has at least one counterpart water-level record with the NSE-values greater than 0.8 in all four frequency ranges. In hydrology, the reproductivity of the water level is generally evaluated as “good” if the NSE is greater than 0.6. Therefore, the above 21 Hi-net records can be regarded to reproduce the water level quite well in all frequency ranges, and the observed seismic noise at these Hi-net stations is considered to be mostly excited by the nearby streamflow. We found that the power spectrum P of these Hi-net noise due to the streamflow follows the power-law given by
P(f)=P0/(1+(f/fc)n)
with a few exceptions. By fitting this formula to the observed power spectrum in the frequency range of 2 Hz to 40 Hz, we found that the corner frequency fc is roughly proportional to -1/3 power of P0, and the power exponent n tended to take the value of 3.
Acknowledgements: This study is funded by the Taisei cooperation research grant and the Japan Society for Promotion of Science (Grant-No. 22K03742).
We selected in total 238 Hi-net up-down component seismograms that covers the period from October 10 to 17, 2019, when the 2019 Typhoon Hagibis passed through eastern Japan. First, for each Hi-net record, we calculated the power spectral density at 2 min and 0.2 Hz intervals in the time and frequency domains, respectively. Next, in order to eliminate transient signals due to earthquakes and monotonic vibrations at each Hi-net station, we calculated the minimum value of the power spectral density in a time-frequency window of 10 min and 1 Hz widths. Then we calculated the vibration power every hour for the frequency ranges of 2-4Hz, 4-8Hz, 8-16Hz, and 16-32Hz, in which the influence of microseisms is considered to be small.
Then, each Hi-net record is paired with the available water level records obtained in the basin where the target Hi-net station is located. We assumed the relationship between the water level H and the vibration power W given by
H=A(W-W0)b+H0 (W>W0)
H=H0 (W<W0)
and estimate parameters A and b through the least square method. W0 and H0 are vibration power and water level before the flood, which are computed using the average value of W and H during October 10 (calm period before the Typhoon Hagibis approaches). To evaluate the degree of agreement between the calculated and observed water levels, we used the Nash-Sutcliffe index (NSE: Nash and Sutcliffe, 1970).
As a result, there were 21 Hi-net stations which has at least one counterpart water-level record with the NSE-values greater than 0.8 in all four frequency ranges. In hydrology, the reproductivity of the water level is generally evaluated as “good” if the NSE is greater than 0.6. Therefore, the above 21 Hi-net records can be regarded to reproduce the water level quite well in all frequency ranges, and the observed seismic noise at these Hi-net stations is considered to be mostly excited by the nearby streamflow. We found that the power spectrum P of these Hi-net noise due to the streamflow follows the power-law given by
P(f)=P0/(1+(f/fc)n)
with a few exceptions. By fitting this formula to the observed power spectrum in the frequency range of 2 Hz to 40 Hz, we found that the corner frequency fc is roughly proportional to -1/3 power of P0, and the power exponent n tended to take the value of 3.
Acknowledgements: This study is funded by the Taisei cooperation research grant and the Japan Society for Promotion of Science (Grant-No. 22K03742).