This study proposes an automatic identification approach for landslide dam formation using a real-time abnormal hydrology record and artificial neural network (ANN). In Taiwan, natural dams occur frequently and have the potential to cause significant disasters, especially in the presence of fragile geological conditions such as steep mountains, earthquakes, intense typhoon, and storm events. During a river blockage event, obtaining immediate real-time information can be difficult or even impossible due to extreme weather conditions, remote locations, and the mountainous terrain of the area. As a result of these issues, there is a strong need for a practical tool that can not only detect the occurrence of this type of natural dam but also estimate the potential real-time risk of dam failure. In order to develop the appropriate safety response, a rapid preliminary assessment of the failure type must be identified, followed by a thorough investigation into the failure process after the event has occurred. In this study, a new method is developed that applies an ANN model (1) to evaluate, in real time, the occurrence of a natural dam and the storage changes behind that dam; (2) to rapidly assess the natural dam failure type and failure process; and (3) to reconstruct the dam height variation during the dam failure process after the event. The proposed methodology is applied to Chi-Shan River Basin located in Kaohsiung, Taiwan, during Typhoon Morakot in 2009. In 3 days, Typhoon Morakot produced 1,911 mm of precipitation. The typhoon event produced a landslide dam and a debris flow that inundated Hsiaolin Village, causing 398 deaths. The tragic results of this case study demonstrate the need for an applicable, accurate, and efficient method to evaluate dam formation and its failure rate.
All Science Journal Classification (ASJC) codes
- Geotechnical Engineering and Engineering Geology