Landslide Susceptibility Evaluation Integrating Machine Learning and SBAS-InSAR-Derived Deformation Characteristics: A Case Study of Yining County, Xinjiang.

Against the background of intensified climate change and enhanced human activities, the occurrence mode of landslides is becoming more complex and changeable, showing a trend of clustering, contiguous, and frequent occurrences. Yining County is located in the middle of the Yili River Valley, where the geological conditions are fragile, neotectonic movement is active, and landslide disasters are widely developed and frequent, posing a serious threat to the population, buildings, and infrastructure. Based on multi-source data combined with machine learning models and SBAS-InSAR technology, this paper realized refined landslide susceptibility evaluation. Firstly, through correlation analysis and other methods, 12 landslide evaluation factors were selected, and the ChiMerge method was used to discretize the continuous factors to build the landslide susceptibility evaluation system. Four machine learning models were used to predict landslide susceptibility, and the RF model performed best. Using the dynamic timeliness advantage of SBAS-InSAR technology, the optimized regional landslide susceptibility evaluation results were constructed, which improved the precision of the landslide susceptibility evaluation results. The purpose of this study is to improve the accuracy and timeliness of landslide sensitivity assessment, improve regional disaster prevention and emergency management planning ability, and provide theoretical and data support for local sustainable development.