Abstract: To address the problems of single monitoring methods, insufficient spatial coverage, and limited early warning accuracy in traditional tailings pond monitoring, a monitoring and early warning platform integrating "Land-Air-Space" multidimensional information and its supporting model were constructed. In the monitoring and early warning platform, ground sensors, UAV photogrammetry, and satellite remote sensing data were integrated. A three-dimensional inclinometer was added to construct a landslide classification matrix for the dam slope. Based on the YOLOv8 model, automatic identification of potential hazards in UAV images was achieved. The PS-InSAR technique was used to identify large-scale deformation points in the tailings pond area (identification threshold >10 mm/a) and to delineate anomalous deformation zones. On this basis, the analytic hierarchy process was adopted to establish a "Land-Air-Space" multidimensional monitoring and early warning model for tailings ponds, and an early warning level upgrading rule based on abrupt changes in key indicators was developed. A demonstration application taking the Gaowanqiu tailings pond as an example showed that, at the platform level, coordinated monitoring by ground, UAV, and satellite enabled multidimensional sensing of millimeter-level internal deformation, centimeter-level surface morphology, and large-scale deformation. The monitoring frequency reached 1 time/min, and the mean average precision of UAV hazard identification was ≥90%. At the model level, by introducing the upgrading rule, the weighted score (blue) was coupled with abrupt signals such as extreme rainfall (orange) to generate an orange warning, thereby avoiding underestimation of compound risks. This study achieves a transition in tailings pond safety monitoring from single-point and static approaches to multidimensional and dynamic diagnosis, and improves the capability for early identification and precise warning of risks in complex environments.