Hazard Assessment of Potential Large-Scale Landslides in the Watershed of the Chenyulan River

Abstract

Taiwan’s mountains are steep, geologically dispersed, and their hillsides are over-utilized. Serious debris flow disasters are prone to happen whenever monsoons or typhoons bring a lot of rain. Typhoon Morakot’s large-scale landslide in 2009, which devastated half of Xiaolin Village and claimed around 500 lives, is one such instance. Large-scale landslides have the potential to seriously harm people, property, and the economy. Large-scale landslides were less frequent in the past, but they have recently increased in frequency. An important topic is how to avert disasters through efficient risk management. Eight influencing factors were chosen after multivariate analysis screening, including three terrestrial factors (total catchment area size, average catchment area slope, and total curvature), three material factors (stratum type, distance from fault, and average normalized-difference vegetation index prior to the event), and two trigger factors (maximum daily rainfall and maximum hourly rainfall). Our overall analysis’ accuracy rankings were as follows: neural network, 93.7%; logistic regression analysis, 92.2%; and discriminant analysis, 89%. The area under curve (AUC) values for the neural network (0.819), discriminant analysis (0.824), and logistic regression analysis (0.732) all demonstrated high discriminative abilities based on the receiver operating characteristic curve, indicating that all three of our models are adequate for this purpose. Large-scale debris flow disasters are more likely to happen when the catchment region’s total area, total curvature, and average slope are higher, according to a cluster study. These findings can help to decrease the effects of disasters of this kind by helping to create early warning systems.