Landslide Risk Assessment In The Eastern Part Of Jiangnan Orogenic Belt Based On Multiple Models

In recent years,affected by extreme climate,in the eastern part of the Jiangnan orogenic belt,heavy rainfall often occurs for many days every spring and summer.This has caused frequent landslides and geological disasters in the region,posing a huge threat to the safety of local people’s lives and property.Therefore,it is extremely urgent to carry out a comprehensive study on the landslide geological disasters in this region and provide appropriate disaster prevention suggestions.In this paper,the artificial visual interpretation of landslides in the eastern section of the Jiangnan orogenic belt is carried out through Google Earth platform,and the interpreted landslides are imported into GIS software.Ten influencing factors,including slope gradient,slope direction,slope position,elevation,distance from water system,distance from fault,normalized difference vegetation index(NDVI),distance from road,formation lithology and20-year average rainfall,are selected to study the distribution law of landslides based on the obtained landslide database.Based on the analysis of landslide distribution law,five methods:information value model,analytic hierarchy process-information value model,logistic regression model,logistic regression-information value model and weight of evidence model,are used to evaluate the risk of landslide,and ROC curve and confusion matrix are used to verify the rationality and accuracy of the hazard evaluation results.On the basis of the hazard assessment results and the vulnerability assessment results of the study area,the risk zoning map of the study area is obtained.The main research contents are as follows:(1)Based on the high-resolution satellite image of Google Earth platform,a total of5047 ancient landslides were interpreted by manual visual interpretation,with a total area of527km~2.According to the classification of landslides according to the size of the area,it is found that the landslide is mainly concentrated in the area of 10~4～10~5m~2,which indicates that there are many landslides in this interpretation,but the landslide area is generally small.(2)Select 10 influencing factors,including slope,aspect,location,elevation,distance from water system,distance from fault,NDVI,distance from road,formation lithology,and annual average rainfall,and classify each influencing factor.Count the landslide number(LN)and landslide area(LA)within each grading interval of each influencing factor,and calculate the landslide number density(LND)and landslide area density(LAP),analyze the influence of each influencing factor on landslide development,and explore the spatial distribution law of landslide.(3)The variance expansion factor and Pearson Correlation Coefficient(PCC)are used to determine that the 10 influence factors selected for evaluation are mutually independent,meeting the requirements of landslide risk assessment.Five methods including information value model,analytic hierarchy process-information value model,logistic regression model,logistic regression-information value model and weight of evidence model are used to evaluate the landslide risk,and the hazard index map is obtained.The hazard index map is divided into five categories:extremely low hazard area,low hazard area,medium hazard area,high hazard area and extremely high hazard area.The rationality of the zoning results is tested,The number and area of landslides in each division are counted,and the density of landslide points and landslide areas are calculated.It is found that the density of landslide points and landslide areas increase with the improvement of the division grade.The results show that the division result is reasonable.(4)The ROC(Receiver operating characteristic curve,ROC)curve and confusion matrix are used to test whether the landslide risk assessment results of the five models are reasonable and accurate,and the assessment effects of each model are compared.The results show that the AUC value and accuracy of the logistic regression-information value model are the highest,and the evaluation results are most consistent with the actual distribution of landslides in this area.(5)Select population,buildings and roads to carry out vulnerability assessment of the study area,and combine the logistic regression-information value model to carry out the hazard zoning map of the study area.Multiply the two assessment results and zoning,and obtain the risk zoning map of the study area.Figure 45 Table 26 References 99...