Risk Assessment Of Debris Flow In Jilin Province Based On GIS And Machine Learning

Jilin Province is one of the provinces that are seriously threatened by debris flow in China,especially in the southeast mountainous areas.During the rainy season,the probability of debris flows increases greatly,and their devastating power can cause unforeseeable consequences.Therefore,taking Jilin Province as the research area,the risk assessment of debris flow is carried out in this paper,hoping to provide decisionmaking basis for the government to formulate prevention measures.Based on the project of "comprehensive research on geological disaster investigation and regionalization of counties(cities)in Jilin Province",we collect the basic data of debris flow in the region,and firstly,we carry out the hazard assessment of debris flow.Based on ten factors,collinearity analysis and information gain ratio are used to select high-quality factors.And we use support vector machine model,convolutional neural network model and the coupling model to complete the debris flow hazard assessment and compare the accuracy of the three models.Judging from the results of the receiver operating characteristic curve,the coupling model has the highest accuracy.Secondly,considering the economic,material and population,three factors are used for the final vulnerability assessment.Finally,the vulnerability map of debris flow is obtained by using the analytic hierarchy process.Finally,the risk map of debris flow is made with the help of GIS platform.The main research contents and results of this paper are as follows:1.This paper first collects the regional geological and economic development data of Jilin Province,and preliminarily analyzes the distribution law and development characteristics of debris flow,which lays a solid foundation for further evaluation of debris flow harm.2.Ten factors including elevation,slope,aspect,vegetation coverage,landform,petrofabric,fault density,rainfall,water system density and land use are utilized as the hazard assessment factors of debris flow.Multicollinearity and information gain ratio are used to screen the factors.Supported by Python,support vector machine model,convolution neural network model and the coupling model are utilized to analyze the hazard of debris flow.The quality of three models is quantitatively evaluated by using the receiver operating characteristic curve.3.From the aspect of economic development,GDP density is selected to measure economic vulnerability.From the material aspect,road density is selected to measure the material vulnerability.From the aspect of population distribution,population density is selected to measure population vulnerability.In order to get the weight value more conveniently,this paper uses Python to encapsulate the analytic hierarchy process,and the vulnerability map is obtained on the Arc GIS platform.4.Taking the above debris flow hazard and vulnerability assessment as the basic layer,the final debris flow risk map is made by using the Arc GIS platform,which is based on the risk model proposed by Liu Xilin.5.The above three evaluations are all carried out on the basis of Python and Arc GIS.In this paper,graphical user interfaces are developed to simplify the operation process and present the risk evaluation process in a more systematic and perfect way.