Abstract: The geological environment in the upper reaches of the Jinsha River is fragile, where large-scale geological hazards are widely developed and pose severe threats. Restricted by complex topographic and geological conditions, the early identification and monitoring of potential hazards are extremely difficult. This paper adopts new technologies and methods, including optical satellites, InSAR, UAVs, and ground-based monitoring equipment to establish a watershed-scale space-air-ground integrated long-term monitoring system for major landslides. The system was successfully applied and verified in the reservoir area of the Yebatan Hydropower Station, with the typical large-scale ancient Dingba landslide taken as a case study to reveal its deformation characteristics. The results show that by integrating multi-source space-air remote sensing technologies such as optical remote sensing, InSAR and airborne LiDAR, a landslide hazard identification chain of "macro-identification-fine interpretation-dynamic monitoring" is constructed. This method breaks through the technical bottleneck of early identifying concealed landslides under complex topographic and geological conditions, enabling precise localization and classification of river-blocking hazard points and significantly improving identification efficiency. In addition, a trinity early warning model of "macro-phenomena-single parameter-multi parameter" is proposed. Based on a quantitative scoring mechanism and cross-validation of multi-sensor data, this model can effectively reduce the false alarm rate and dynamically upgrade the warning level. The monitoring system has been verified by case studies to have good sensitivity and applicability, which can provide technical support for the emergency monitoring and management of major geological disasters in river basins, thereby enhancing disaster prevention and mitigation capabilities.