Abstract: Objectives: Landslide Landslide susceptibility assessment is a crucial method for evaluating the spatial probability of landslide occurrence. However, conventional susceptibility assessment methods mainly rely on static conditioning factors, making it difficult to capture the temporal evolution characteristics of landslide susceptibility. To address this limitation, this study takes Longnan City, Gansu Province as a case study. Methods: A dynamic landslide susceptibility assessment framework coupling "static background-dynamic factors" was proposed. Firstly, eight static conditioning factors, including curvature, slope, and landuse type were selected. Four ensemble learning models, namely Random Forest (RF), Light Gradient Boosting Machine (LGBM), Gradient Boosting Decision Tree (GBDT), and eXtreme Gradient Boosting (XGBoost), were employed to construct static background susceptibility maps and identify the optimal model. Secondly, Interferometric Synthetic Aperture Radar (InSAR) technology was used to derive quarterly cumulative surface deformation for 2024, while quarterly precipitation and normalized difference vegetation index (NDVI) data for the corresponding periods were incorporated as dynamic factors. By integrating static susceptibility classes with quarterly dynamic factor levels through an overlay matrix, dynamic landslide susceptibility maps were generated for different quarters. Finally, the Shapley Additive exPlanations (SHAP) method was applied to interpret the contributions of controlling factors and reveal the underlying landslide mechanisms. Results: The results indicate that the XGBoost model achieved the best performance in static background modeling, with an AUC value of 0.853. After incorporating quarterly dynamic factors, the proportion of landslide points located within the very high susceptibility zone increased from 51.9% in the static background map to 57.4% in the third quarter. The third quarter exhibited the largest surface deformation occurred, corresponding to the rainy season and peak precipitation period. The dynamic susceptibility maps further reveal that areas with high and very high susceptibility are primarily distributed along valley slopes in the northeastern and southern parts of the study area, with susceptibility levels showing pronounced seasonal variations. These results demonstrate that dynamic factors provide critical temporal information that cannot be captured by conventional static susceptibility maps alone. SHAP analysis further indicates that land-use type, distance to roads, and distance to rivers are the dominant controlling factors influencing landslide susceptibility in Longnan City. Conclusions: This study demonstrates that integrating quarterly-scale dynamic surface deformation, precipitation, and vegetation information into conventional susceptibility assessment framework can effectively improve the identification of landslide-prone areas. The proposed approach provides a scientific basis for dynamic landslide hazard prevention and risk mitigation in Longnan City and similar mountainous regions.