In landslide-prone areas, spatial gaps in InSAR-derived deformation maps caused by incomplete SAR coverage hinder continuous surface deformation assessment and limit reliable landslide analysis. To address this problem, we propose an explainable AI (XAI) framework that integrates SBAS-InSAR, ensemble machine learning, and Shapley Additive exPlanations (SHAP) to estimate surface deformation in SAR-scarce regions. Geological and engineering factors, including protective measures, distance to roads, and land use, were combined with remote sensing and field data to build a comprehensive dataset. Four ensemble models (LightGBM, XGBoost, Random Forest, and CatBoost) were trained and evaluated, with XGBoost achieving the best performance (R² = 0.816, RMSE = 6.85 mm, MAE = 4.27 mm). Validation against two GNSS benchmarks confirmed sub-millimeter accuracy (0.6 mm and 0.3 mm). Both XGBoost and CatBoost delineated continuous deformation patterns consistent with field-observed damage. SHAP analysis provided model interpretability, highlighting elevation and human-engineering factors as key drivers: areas farther from roads and under cultivation were more prone to downslope movement, while damaged protective works exhibited greater deformation. By coupling InSAR with XAI, this study achieves accurate and interpretable surface deformation estimation in data-scarce regions, advancing landslide assessment and early warning applications.

Physically-based model is an important method for refined assessment of landslide hazard at large scale. The traditional infinite slope model homogenizes the slope’s structure and morphology, discretizing the slope into grid units and neglecting the interactions between different parts of the slope. However, slope units constitute the fundamental elements for stability analysis of natural slopes. Moreover, landslide bodies or slopes prone to landslides exhibit significant spatial heterogeneity. In order to realize the landslide hazard assessment in slope units, this study proposes a physically-based model called representative profile model (RPM). RPM takes the slope unit as the assessment unit and couples the slope surface morphology, Quaternary deposits thickness and ground water level. In order to represent the information of the slope unit within a single cross-section, the elevation range of the slope unit is divided with a uniform interval into some elevation segments. Each segment is assigned the average grid values of its respective elements. Then, a representative profile can be generated, consisting of ground surface, sliding surface, and ground water level. RPM also integrates the slices method and the Monte Carlo method to calculate the failure probability, allowing a physically-based hazard assessment in slope unit at a large scale. This study automates the process of RPM model through secondary development of ArcGIS. RPM model were applied in Tiefeng Township, Chongqing, China. The results validated by ROC curves and field investigation represent good performances, which could provide evidence of the potential of RPM for the landslide hazard assessment at regional scale.