1:45 PM - 2:00 PM
[SSS07-07] Estimation of river discharge using Hi-net seismic noise and the turbulent flow model
★Invited Papers
Keywords:River discharge, Seismic noise, Turbulent flow model, Parameter study, Hi-net
Vibrations excited by streamflow (streamflow noise) is one of the dominant sources in seismic noise in frequencies ranging from 1 to tens of Hz. The excitation process of relatively low frequency streamflow noise has been modeled by the turbulent flow (Gimbert et al., 2014). According to the turbulent flow model, excitation of vibration energy is strongly correlated with the riverbed slope and the water height. Therefore, the streamflow noise is considered a good proxy of river discharge during a flood event and is especially observable in the upstream basin. This property leads to the expectation that seismic noise recorded in a mountainous region, where applying conventional hydrological methods is challenging due to poor accessibility and dangerousness, could be used to indirectly estimate river discharge. From this perspective, elucidating the characteristics of streamflow noise from both theoretical and observational viewpoints is considered one of the highly useful applications of environmental seismology.
In this study, we analyzed seismic noise recorded by NIED's Hi-net during the July 2024 flood in Yamagata Prefecture and the September 2024 flood on the Noto Peninsula, Japan. Durng both flood events, as predicted by the turbulent flow model, a significant increase in noise power was observed in 1-2 Hz particularly in upstream regions. Applying the turbulent flow model to these noise records, we estimated the discharge of rivers flowing nearby the Hi-net station. We first extracted river channel geometry, river width, and local riverbed slope distribution from Google Earth and the DEM data provided by the Geospatial Information Authority of Japan. The typical grain size of riverbed sediments was assigned as a function of riverbed slope. The Rayleigh wave phase velocity, group velocity, and quality factor for describing the Green’s function from the vibration source to the observation point were taken from the values reported by previous studies.
Both the absolute values and temporal variations of the discharge estimated from the noise records showed a good agreement with hydrologically observed discharge. Specifically, the peak discharge estimated from the record of Hi-net N.FGTH, which is located in the upstream region of Mogami river, differed by only 20% compared to the hydrological observation at 8 km downstream. Additionally, the peak discharge estiamated from Hi-net N.MGMH, located about 20 km upstream of N.FGTH, was approximately one-fifth of that at N.FGTH, qualitatively demonstrating the general trend of decreasing discharge in upstream regions. The estimated peak discharge at Hi-net N.YGDH on the Noto Peninsula was even lower, which is reasonable given the narrower river basin area of this region.
However, the parameters used in turbulent flow model are generally highly uncertain, which significantly affects the discharge estimation. In particular, if the S-wave velocity of the medium differs by a factor of two, the estimated discharge can vary by nearly a factor of ten. The uncertainty in the grain size of riverbed sediments and the river width also significantly affects the discharge estimation. This means that parameter uncertainty makes it difficult to determine absolute discharge values with better than an order-of-magnitude accuracy. Nevertheless, the relationship between noise power and discharge follows a power-law in wide ranges regardless of the input parameter values. This means that if a relationship between discharge and noise power can be calibrated using data from past floods, it would be possible to estimate the discharge of ongoing floods using the ratio of the past and current noise powers.
Acknowledgements: This study was supported by Japan Society for the Promotion of Science (Kakenhi Grant-No. 22K03742).
In this study, we analyzed seismic noise recorded by NIED's Hi-net during the July 2024 flood in Yamagata Prefecture and the September 2024 flood on the Noto Peninsula, Japan. Durng both flood events, as predicted by the turbulent flow model, a significant increase in noise power was observed in 1-2 Hz particularly in upstream regions. Applying the turbulent flow model to these noise records, we estimated the discharge of rivers flowing nearby the Hi-net station. We first extracted river channel geometry, river width, and local riverbed slope distribution from Google Earth and the DEM data provided by the Geospatial Information Authority of Japan. The typical grain size of riverbed sediments was assigned as a function of riverbed slope. The Rayleigh wave phase velocity, group velocity, and quality factor for describing the Green’s function from the vibration source to the observation point were taken from the values reported by previous studies.
Both the absolute values and temporal variations of the discharge estimated from the noise records showed a good agreement with hydrologically observed discharge. Specifically, the peak discharge estimated from the record of Hi-net N.FGTH, which is located in the upstream region of Mogami river, differed by only 20% compared to the hydrological observation at 8 km downstream. Additionally, the peak discharge estiamated from Hi-net N.MGMH, located about 20 km upstream of N.FGTH, was approximately one-fifth of that at N.FGTH, qualitatively demonstrating the general trend of decreasing discharge in upstream regions. The estimated peak discharge at Hi-net N.YGDH on the Noto Peninsula was even lower, which is reasonable given the narrower river basin area of this region.
However, the parameters used in turbulent flow model are generally highly uncertain, which significantly affects the discharge estimation. In particular, if the S-wave velocity of the medium differs by a factor of two, the estimated discharge can vary by nearly a factor of ten. The uncertainty in the grain size of riverbed sediments and the river width also significantly affects the discharge estimation. This means that parameter uncertainty makes it difficult to determine absolute discharge values with better than an order-of-magnitude accuracy. Nevertheless, the relationship between noise power and discharge follows a power-law in wide ranges regardless of the input parameter values. This means that if a relationship between discharge and noise power can be calibrated using data from past floods, it would be possible to estimate the discharge of ongoing floods using the ratio of the past and current noise powers.
Acknowledgements: This study was supported by Japan Society for the Promotion of Science (Kakenhi Grant-No. 22K03742).