科研之友
微信
新浪微博
Facebook
分享链接
摘要
Landslide is an important typical geological disaster. Under humid and hot conditions in southern China, the formation mechanism of landslides in clastic rock area is different from other areas. A systematic and targeted analysis for landslides identification and early risk warning in clastic rock areas is needed in terms of such various formation mechanisms. This article takes the landslides in the clastic rock area of Laibin Jinxiu Yao Autonomous County in Guangxi as the research object. This article selects 10-formative factors such as lithology, slope, slope direction and degree of relief and couples 4 kinds of models which are (I) information model and logistic regression model (LR), random forests (RF), BP neural network (BP) and convolutional neural networks (CNN). For the research method, a landslide susceptibility zoning was at the end provided after first establishing a landslide susceptibility evaluation model in clastic rock areas, and finishing a critical accuracy level evaluation of the model. Based on the research result, the ROC curve AUC value of the coupled model of information and BP neural network is the largest (0.94). Such a result proved the high accuracy level of the research method, especially when utilizing it in clastic rock areas and indicated that such a method could further be applied in classifying landslide susceptibility in Clastic rock areas. Based on the final evaluation, the landslide-prone areas from the study target are mainly concentrated on both sides of the road network. This paper carried out an effective exploration of the landslide-prone zoning in Clastic rock areas, which supported the local disaster prevention work fundamentally. Besides, this study also provided a specific analysing logic for landslide prevention and control in similar areas. ? 2023 Science Press.
- 单位
