Evaluation Of The Risk Of Loess Earthquake Landslides Based On Statistical Model

Loess Landslides in the Loess region are one of the most dangerous geological hazards.According to historical disaster statistics,several significant earthquakes in the Loess Plateau region have generated a huge number of landslides,inflicting great danger and loss of life and property,including road damage,bridge collapse,burial of villages and households,and a large number of casualties.Therefore,to improve the ability to prevent and treat earthquake-induced landslides in loess zones,as well as to guide the selection of town construction locations.A thorough study of the impact factors of earthquake-induced landslide hazards in loess under the effect of powerful earthquakes is vital,as is the establishment of a hazard evaluation index system.This paper takes as an example the historical major seismic landslide hazard affected areas of Dingxi,Longnan and Tianshui in Gansu Province,conducts a survey of seismic hazard cases in the study area,compiles historical landslide hazard data,uses statistical analysis models,screens and classifies seismic hazard evaluation index factors for landslide areas,designs and completes large shaking table model tests for different impact factors,and verifies the typical model was developed to assess the risk of loess seismic landslides..The main results achieved are summarized as follows:(1)Based on the analysis of available data on the geographical and geological environmental conditions of the research area,combined with on-site research on typical loess earthquake landslides in the region: Jingning landslide,Yongguang village landslide,and Baozi village landslide in Weixin Township.Statistical analysis was conducted on the relationship between various influential internal and external factors during the process of earthquake induced landslides and the distribution of historical earthquake induced landslide disasters in the region.Based on the association between the two,a database of loess earthquake landslide affecting elements was initially screened out.Elevation,slope,slope direction,NDVI,stratigraphic lithology,landform type,distance from rivers,and distance from fracture zones were among the factors considered.(2)Using the multiple linear regression function of SPSS software,the correlation analysis of the nine factors obtained from the initial screening was performed.The results showed that their VIF values were low and the covariance problem was negligible,allowing the subsequent evaluation step to be completed.The distribution characteristics of landslides in the area were also determined using Arc GIS software to determine the density of landslide points in various factor intervals.(3)The slope model was designed using the control variable method,and shaking table model tests were carried out to simulate the slope damage of different factors under the action of earthquakes by conducting research on the index system for evaluating the risk of loess earthquake landslides.It was verified that factors such as slope,slope height and rainfall can all induce landslides under the action of earthquakes,while the effect of slope direction factor is not obvious.(4)Landslide hazard assessment method based on statistical models is compared and optimised.By comparing and analysing the principles and assessment results of the informativeness model and the logistic regression model,the coupled application of the informativeness and logistic regression models is proposed.Comparing the accuracy of the assessment results of the three models,and shows that the coupled informativeness-logistic regression model performs best,the informativeness model is second and the logistic regression model is lower.The specific AUC values based on the ROC curves were 0.745 for the informative model,0.715 for the logistic regression model and 0.754 for the coupled informative-logistic regression model.