The vulnerability assessment of rainfall-induced debris flows in Taiwan

A debris flow vulnerability assessment which incorporates topographic and rainfall effects is developed. Rainfall at a scale compatible with the resolution of digital elevation model is obtained using a neural network estimation method with a wind induced topographic effect and rainfall derived from satellite rain estimates and an improved inverse distance weight method. The technique is tested using data collected during the passage of typhoon Tori-Ji on July 2001, which caused massive debris flows in central Taiwan. Numerous debris flows triggered by the typhoon were used as control for the study. The results show that the proposed wind-topography neural network (WTNN) technique outperforms other popular interpolation techniques, including inversed distance weight method (IDW), ordinary kriging (OK), co-kriging method, and multiple linear regression method.; Multiple fuzzy-logic-based debris flow susceptibility factors are used to characterize watersheds. Self-organizing maps (SOM) was adopted for the debris flow vulnerability assessment by incorporating estimated rainfall and debris flow susceptibility factors. The result examined by contingency table agrees to the assessment proposed by Soil and Water Conservation Bureau of Taiwan and National Science and Technology Center for Hazard Reduction of Taiwan. An index of vulnerability representing the degrees of hazard is implemented in a GIS-based decision support system which decision maker can use to manage debris flow environmental issues.; Key Words. Debris flow, spatial interpolation, vulnerability assessment, satellite rainfall, neural network, GIS.