Early Identification And Risk Assessment Of Debris Flow Pregnant Areas In Danba County

China is a country with frequent debris flow disasters.Especially in recent years,there have been many abnormal weather and extreme rainfall,resulting in frequent debris flow disasters,posing a serious threat to social and economic development and the safety of people’s lives and properties.Early and accurate identification of debris flow disaster-prone areas(ie,early identification of debris flow)can provide a basis for the prevention and control of debris flow disasters,and carry out early warning and monitoring of debris flow to reduce losses caused by disasters.At present,the early identification of debris flow has the following problems: there are many machine learning methods,which need the comparative analysis of multiple models to obtain the optimal recognition effect;when the early identification of debris flow is carried out by basin as a unit,the debris flow cannot be accurately located;there is a lack of risk evaluation and analysis of the identification results.In view of the above problems,this paper first takes the basin unit as the analysis object,and selects the optimal machine learning method to identify the disaster-prone areas of debris flow through the comparison of various machine learning methods.Then the extension theory is introduced to evaluate the risk degree of debris flow in the branch gully in the disaster area,and the debris flow gully is further located according to the evaluation results.This paper takes Danba County as the research area to carry out related research,and the specific work is as follows:(1)According to the topography,geology,rainfall and other factors of Danba County,combined with the catalogue of historical debris flow disasters,analyze the characteristics of debris flow in Danba County,and extract a total of 26 identification factors for debris flow disaster areas,including drainage area,relative height difference,rock formation Hardness,etc.(2)In order to ensure the rationality and scientificity of debris flow disaster area identification,the watershed unit is selected as the analysis object.In this paper,a total of201 watershed units are taken out,and combined with the catalogue data,distribution and formation conditions of debris flow in Danba County,38 watersheds with debris flow are taken as positive samples,85 watersheds without debris flow are taken as negative samples,and the remaining 78 watersheds are taken as unknown watersheds.(3)Use Decision tree,Random forests,Support vector machines,Multilayer feedforward neural network to construct the identification model of debris flow disaster area respectively.Based on the model accuracy evaluation results,the random forest model was finally selected to identify the 78 watersheds to be identified for debris flow disaster-prone areas,and a total of 13 debris-flow disaster-prone areas were identified.After verification by later data,one of them had a mudslide in June 2020.(4)Extract 41 branch ditches from 13 debris flow disaster-prone areas,select 11 evaluation factors such as the length of the branch ditches and the longitudinal slope of the branch ditches,and combine the CRITRIC weighting method to construct an extensibility debris flow risk evaluation model.The debris flow risk assessment was carried out in the branch ditches,and it was concluded that 7 had a very high risk level and 16 had a high risk level.This result can provide a scientific reference for the prevention and control of debris flow disaster risk in Danba County.