Abstract
The high-altitude Duku Highway is characterized by complex terrain changes and frequent geological hazards, which severely impact the lives of local residents and the sustainable development of the regional economy. The lack of understanding of terrain deformation, coupled with scarce foundational observation data, makes it challenging to apply mainstream susceptibility assessment methods such as slope modeling and causality analysis. Consequently, this study utilizes Sentinel-1 A data and employs the SBAS-InSAR technique to extract and analyze the deformation characteristics of 184 hazard areas along the Duku Highway over nearly three years. Furthermore, the Correlation Clustering Evaluation Model is proposed, attributing hazard properties to unsupervised spatial clustering results, thus enabling the study of hazard susceptibility in data-scarce regions without prior knowledge. The results indicate that the SBAS-InSAR coherence is 0.64, with a validation accuracy of 85%. The high, relatively-high, and moderate susceptibility areas account for 24.7%, 17.1%, and 32.6% of the total area, respectively. The rapid uplift of terrain due to plate compression is a major factor leading to frequent hazards in high and relatively-high susceptibility areas. These regions may spontaneously experience cyclic hazards (minimum of 2 months) without extreme external factors. The research findings offer new insights into regional hazards and provide a basis for the sustainable management of highways.