Prediction And Initiation Simulation Of Rainfall-induced Debris Flow In Yingxiu Town, Wenchuan County, Sichuan Province

Prediction of debris flow based on rainfall conditions is a research hotspot at home and abroad in recent years.In this paper,Yingxiu is taken as the study area,the prediction model of rainfall-induced debris flow is established,and the initiation process of debris flow is simulated.The purpose of this paper is to predict the possible time of debris flow,so as to reduce the casualties and economic losses in the dangerous area of debris flow,and to reduce the damage caused by debris flow.Considering comprehensively the topography and geomorphology,provenance and rainfall conditions needed for the formation of rainfall-induced debris flow,The basin area,the length of the main channel,the average longitudinal slope of the main channel,the bending coefficient of the main trench bed,the average slope of the basin,the shape coefficient,the lithology,the area ratio of collapses and landslides,the Normalized Difference Vegetation Index,the accumulated rainfall,the rainfall duration and the average rain intensity.There are 12 factors that have influence on the occurrence of debris flow.Boruta algorithm and Mann-Whitney U test were used to screen predictors of rainfall-induced debris flow.70%of the data from these samples were selected as training set by random sampling method,and the remaining 30%of the data were used as validation set to verify the prediction effect of the model.The prediction model of debris flow is established based on logistic regression model.The prediction effect of the model is quantitatively evaluated by using the indexes of presion,accuracy,false negative rate,false positive rate,F₁ value and AUC value.In order to further analyze the possible initial position,LHT model is used to simulate the initiation process of debris flow.The main results are as follows:?1?By comparing the influence of factors on model accuracy and determining the importance of factors,Boruta algorithm is of great help to improve model accuracy and simplify the model.Mann-Whitney U test takes significance as the standard of factor screening,and can also be well used in the selection of prediction factors of rainfall-induced debris flow.?2?The area ratio of collapses and landslides,the Normalized Difference Vegetation Index and the average rainfall intensity have important effects on the occurrence of debris flow,while the other factors are of low importance and significance.So they are used as predictors of rainfall-induced debris flow in Yingxiu Town.?3?Based on the logistic regression model,the debris flow prediction model is established.The prediction effect of the model is better,the AUC values of the training set and the verification set are respectively 0.865 and 0.823.It shows that the logistic regression model,as a generalized linear statistical analysis model,has been well applied in the process of establishing the rainfall-induced debris flow prediction model in Yingxiu Town.?4?By comparing with the I-D threshold model,Fisher discriminant model and ID-Fisher model,it is found that the prediction effect of the logistic regression model is better than other models,indicating that the logistic regression model is advanced and applicable in the debris flow prediction model.?5?Using LHT model to analyze the initiation process of debris flow,the simulation results are basically consistent with the actual debris flow initiation time and the average rain intensity at initiation,indicating that the simulation effect is better.