Presentation Information
[U15-05]Long-Term Measurement of Water-Level and Channel Vibrations for Debris-Flow Sign Detection★Invited Papers
*Seigo Sakaki1, Yuya Tonegawa1, Teruyuki Kikuchi1, Kyosuke Sasaki2, Koichi Hayashi3 (1.Suwa University of Science, 2.Amber Logix Inc., 3.Research Center for Landslide Disaster Risk Cognition and Reduction, Disaster Prevention Research Institute, Kyoto University.)
Keywords:
Debris-flow,Long-Term Measurement,Anomaly Detection
Abstract
This study aims to quantify experiential knowledge derived from local traditions through long-term measurement (LTM) and the deployment of custom-built sensors, and to explore its use for debris-flow detection in small and medium-sized rivers. In the Suwa area, traditions such as “the river makes a rumbling sound” and “the river smells” have supported early evacuation decisions; moreover, the 2021 debris-flow disaster was accompanied by a temporary drop in river water level immediately before occurrence. In this study, these tradition-based signs are interpreted as two physical variables: (1) water-level fluctuations and (2) river vibrations, and a site-specific monitoring approach is pursued. Specifically, multiple water-level gauges were distributed along small rivers and an LPWA-based system was established for continuous data acquisition, while a simple seismometer was additionally installed to detect vibrations associated with the movement and collision of bedload gravels. Preliminary observations indicate that rainfall-driven background vibrations differ from impulsive waveforms generated by gravel motion, suggesting the potential to physically distinguish the perceived “rumbling” and to support practical warning and early evacuation decisions even in rivers where continuous expert supervision is difficult.
The study area comprises small and medium-sized river catchments extending from the Konan district of Suwa City to the Ankokuji district of Chino City, Nagano Prefecture. The target streams include the Dodogawa River, Dodozawa, and Nakanosawa; upstream reaches have steep gradients and include unstable slopes with landslide topography.
Fig.1 summarizes our monitoring concept. We continuously record rainfall, water level, and ground vibrations and transmit the data in real time via LPWA. Under normal rainfall, vibration levels broadly follow the rise in water level, whereas impulsive vibration bursts (“rumbling”)—validated using a boulder-dragging test as a reference signature—are treated as potential debris-flow precursor signals. In addition, a water-level drop despite increasing rainfall is interpreted as a possible indicator of channel blockage risk.
Along the Dodogawa River, two water-level gauges and a rain gauge captured the largest rainfall event of the year (3 November 2024; total 78 mm; maximum hourly 18 mm): water levels rose nearly synchronously at upstream and downstream sites, and no notable anomalies such as landslide-induced blockage or a water-level drop were observed. Because hourly rainfall did not reach the interview-suggested threshold (>50 mm/h), the “rumbling” sound was not confirmed for this event. Although vibration analysis is ongoing, 1-min data showed an increasing trend in RMS vertical acceleration corresponding to water-level rise during rainfall. In rain-free conditions, the boulder-dragging experiment produced distinct vibration waveforms despite stable water levels, differing from rainfall-related amplitude characteristics.
Consistent upstream-downstream water-level responses confirm the usefulness of distributed water-level surveillance in small rivers. The absence of “rumbling” and pronounced geomorphic change during the largest event is reasonable given rainfall intensity below the suggested threshold. Differences between rainfall-induced vibrations and gravel-dragging waveforms demonstrate a fundamental potential for discriminating sign-like vibrations. Future work will expand monitoring to other regions and attempt to quantify the tradition that “the river smells” as an additional indicator.
This study aims to quantify experiential knowledge derived from local traditions through long-term measurement (LTM) and the deployment of custom-built sensors, and to explore its use for debris-flow detection in small and medium-sized rivers. In the Suwa area, traditions such as “the river makes a rumbling sound” and “the river smells” have supported early evacuation decisions; moreover, the 2021 debris-flow disaster was accompanied by a temporary drop in river water level immediately before occurrence. In this study, these tradition-based signs are interpreted as two physical variables: (1) water-level fluctuations and (2) river vibrations, and a site-specific monitoring approach is pursued. Specifically, multiple water-level gauges were distributed along small rivers and an LPWA-based system was established for continuous data acquisition, while a simple seismometer was additionally installed to detect vibrations associated with the movement and collision of bedload gravels. Preliminary observations indicate that rainfall-driven background vibrations differ from impulsive waveforms generated by gravel motion, suggesting the potential to physically distinguish the perceived “rumbling” and to support practical warning and early evacuation decisions even in rivers where continuous expert supervision is difficult.
The study area comprises small and medium-sized river catchments extending from the Konan district of Suwa City to the Ankokuji district of Chino City, Nagano Prefecture. The target streams include the Dodogawa River, Dodozawa, and Nakanosawa; upstream reaches have steep gradients and include unstable slopes with landslide topography.
Fig.1 summarizes our monitoring concept. We continuously record rainfall, water level, and ground vibrations and transmit the data in real time via LPWA. Under normal rainfall, vibration levels broadly follow the rise in water level, whereas impulsive vibration bursts (“rumbling”)—validated using a boulder-dragging test as a reference signature—are treated as potential debris-flow precursor signals. In addition, a water-level drop despite increasing rainfall is interpreted as a possible indicator of channel blockage risk.
Along the Dodogawa River, two water-level gauges and a rain gauge captured the largest rainfall event of the year (3 November 2024; total 78 mm; maximum hourly 18 mm): water levels rose nearly synchronously at upstream and downstream sites, and no notable anomalies such as landslide-induced blockage or a water-level drop were observed. Because hourly rainfall did not reach the interview-suggested threshold (>50 mm/h), the “rumbling” sound was not confirmed for this event. Although vibration analysis is ongoing, 1-min data showed an increasing trend in RMS vertical acceleration corresponding to water-level rise during rainfall. In rain-free conditions, the boulder-dragging experiment produced distinct vibration waveforms despite stable water levels, differing from rainfall-related amplitude characteristics.
Consistent upstream-downstream water-level responses confirm the usefulness of distributed water-level surveillance in small rivers. The absence of “rumbling” and pronounced geomorphic change during the largest event is reasonable given rainfall intensity below the suggested threshold. Differences between rainfall-induced vibrations and gravel-dragging waveforms demonstrate a fundamental potential for discriminating sign-like vibrations. Future work will expand monitoring to other regions and attempt to quantify the tradition that “the river smells” as an additional indicator.