Accurate identification of deep-seated slip surfaces (DSS) is fundamental to understanding the kinematics and hazard potential of large deep-seated creeping landslides (LDCLs). Current approaches that integrate InSAR-derived deformation with mass-conservation-based inversion enable regional-scale estimation of landslide thickness; however, they are inherently ill-posed and often poorly constrained in terms of the absolute depth of slip surfaces. To address these limitations, this study proposes an improved mass-conservation-based inversion framework, termed the Deep-seated Slip Surface Inversion Model (DSIM), which couples InSAR-derived surface deformation with sparse borehole information as constraints to enhance the reliability of DSS reconstruction for large-scale landslides. The proposed method was applied to the Jungong landslide in the Upper Yellow River. First, three-dimensional surface deformation was derived using SBAS-InSAR combined with topographic constraints. Second, DSIM was employed to invert the DSS geometry and estimate landslide volume. Finally, a post-instability runout scenario was simulated under the assumption that failure initiates along the inverted DSS. The results demonstrate that DSIM yields a physically consistent DSS geometry that agrees well with borehole observations and independent geophysical evidence, indicating strong robustness under sparse in-situ constraints. The sliding mass associated with the DSS of the Jungong landslide is primarily concentrated in Zones I–IV, with a maximum thickness of 100.2 m and an estimated total volume of approximately 5.525 × 10⁷ m³. The extreme-scenario simulation indicates a potential river-blocking hazard to the Yellow River. The proposed DSIM improves the reliability of DSS reconstruction for LDCLs and provides a practical geometric basis for subsequent hazard assessment.

Article link: https://doi.org/10.1016/j.enggeo.2026.108701