Automatic debris flow detection using geophones

Abstract

Ground vibration recorded using geophones can provide information on debris flows. This information can be used for debris flow detection and warning. However, real debris flow signals change not only from site to site but also during different times at the same location. Thus, a constant threshold calibrated with historic data at one location may not accurately detect debris flow at the same location, let alone at other locations. Therefore, a debris flow detection method requires a criterion that can transform current environmental conditions into a detection criterion. The characteristic of a debris flow tremor was discussed and transformed into two mathematical indices. We used accumulation energy and the rate of energy change in a specific spectrum domain as the representative indicators. To distinguish debris flows from noise, we assumed debris flow energy to be at least an order of magnitude larger than noise before debris flow arrival. To avoid a short-term spike event confusing the detection of debris flow, accumulation energy at an interval longer than 2 s was used. This method was tested on three debris flow events with complete records at the Ai-Yu-Zi Creek, Taiwan. The detection of arrival time for debris flows using geophone data had a delay of less than 10 s compared with the detection of arrival time through video recording.