Landslide susceptibility law under the absence of landslide sample and the susceptibility-InSAR multi-source information method considering the potential landslide identification

Landslide sample data is the basis of susceptibility modeling construction,the lack of landslide samples will affect the accuracy of susceptibility modeling and prediction results. In order to reveal the susceptibility law under the different landslide samples missing conditions and propose a potential landslide identification method, the Xunwu County in Jiangxi Province is selected as the case study. The landslide catalog data by field geological survey is assumed as the basic ideal landslide samples condition. And the different landslide sample missing conditions are determined by randomly and evenly selecting and eliminating landslide samples with different proportions of 10%,20%,30%,40% and 50%. Meanwhile,the landslide sample aggregation missing condition is determined by eliminating the landslide samples in the southern part of the study area. Then,the random forest and support vector machine models are selected to construct landslide susceptibility models under different landslide sample missing conditions,and the effect of landslide sample missing conditions on landslide susceptibility prediction is analyzed. Additionally,the Susceptibility-InSAR method has been proposed by combining SBAS-InSAR technology and landslide susceptibility prediction results to identify potential landslides. Finally,based on the landslide sample aggregation missing condition,the potential landslides are used as expanded landslide samples to reconstruct the landslide susceptibility model and to evaluate the performance accuracy. The results show that: (1) the larger the proportion of landslide sample missing data,the stronger the uncertainty of the landslide susceptibility prediction results,and the landslide susceptibility prediction accuracy in landslide sample aggregation missing condition decreases significantly. (2) The Susceptibility-InSAR multi-source information method can accurately identify potential landslides,providing an effective method for expanding landslide samples in areas with missing landslide data. (3) Compared with the landslide sample aggregation missing condition,the landslide susceptibility results in the condition of expanding landslide samples have higher precision and lower uncertainty, suggesting that potential landslide identification can effectively mitigate the significant issue of landslide sample aggregation missing condition. © 2025 Academia Sinica. All rights reserved.