A study on multiple-model evaluation of landslide susceptibility

HUANG Cheng; DENG Yunlong; YAN Xiangsheng; ZHOU Xincheng

doi:10.11932/karst20240615

Abstract：Landslides are one of the most common geological disasters in China,characterized by sudden occurrence and uncertainty.The evaluation of landslide susceptibility is a complex process.Conventional methods mainly use static factors,making it difficult to achieve dynamic assessment of landslide susceptibility.With the ongoing advancement of science and technology,interferometric synthetic aperture radar(InSAR)has been successively applied to the study of geological disasters.This technology is characterized by its all-weather capability,continuous operation,and extensive coverage,allowing for real-time monitoring of the Earth's surface under varying environmental conditions.InSAR enables a comprehensive understanding of the movement of the surface rocks and soil masses associated with landslide geological disasters.It effectively captures the dynamic deformation characteristics of landslides in the vertical direction,thereby enhancing the identification and dynamic monitoring of surface deformation and improving the accuracy of evaluating landslide susceptibility.In this study,the surface deformation representative factor has been introduced into the conventional evaluation of geological disaster susceptibility.This addition improves the reliability of the evaluation of landslide susceptibility and enhances the overall accuracy. This study focused on Shuangjiang county as the research area.It utilizes evaluation index factors such as Digital Elevation Model(DEM),slope gradients,aspects,curvatures,stratigraphic lithology,faults,land use,annual average rainfall,roads,and rivers.The representative factor of InSAR surface deformation was comprehensively considered to evaluate landslide susceptibility.Through an extensive analysis of InSAR deformation,a dataset of landslides was established,identifying a total of 116 landslide geological disasters.Among them,56 landslide areas exhibited deformation,with some slopes showing significant signs of deformation.The information quantity,certainty factor,and frequency ratio models were employed to evaluate the susceptibility of areas to landslides.The accuracy of the generated landslide susceptibility was evaluated with the use of the landslide density ratio,curve of Receiver Operating Characteristic(ROC),and the Area Under the Curve(AUC).In this study,70%of the landslide events were randomly selected for spatial modeling training,while the remaining 30%were used for model verification.The segment set statistical tool in ArcGIS software was utilized to conduct the mutual independence test on the evaluation factors.Research findings indicate that all the correlation coefficients are less than 0.3,suggesting that the evaluation factors are independent of one another.According to the natural paragraph point method in Geographic Information System(GIS),the susceptibility can be categorized into five intervals:low susceptibility area,relatively low susceptibility area,medium susceptibility area,relatively high susceptibility area,and high susceptibility area.The high landslide susceptibility areas are mainly distributed in the northern part of Shuangjiang county;the low landslide susceptibility areas are mainly concentrated in its northwestern part.In the relatively high susceptibility area and the high susceptibility area,the raster of the inspection samples accounts for 88.69%of the total landslide inspection raster. The experimental results show that the Certainty Factor(CF)model exhibits a relatively high landslide density ratio in both the high susceptibility area and the relatively high susceptibility area,with ratios of 7.77 and 1.10,respectively.Additionally,the model demonstrates the highest accuracy and AUC values,which are 0.822 and 0.879,respectively.The accuracy of Frequency Ratio(FR)model is followed by CF model,and that of Information Quantity(I)model is the lowest.The landslide susceptibility map generated by the CF model provides a more accurate evaluation of slope instability in Shuangjiang county.Therefore,deriving the surface deformation factor based on InSAR technology and the CF model for evaluating landslide susceptibility yields the highest accuracy.