Abstract: To enhance the timeliness of graded early warnings for tailings dam under rainfall conditions, the variability of effective cohesion, effective internal friction angle, and saturated hydraulic conductivity is considered. Based on the "Geological Hazard Prevention—Slope Engineering Structural Reliability Design Code, " the reliability index values for graded stability warnings of tailings dam are determined. For each warning level’s reliability index, corresponding critical rainfall pattern curves are constructed using a uniform rainfall model, forming a graded early warning zoning map for tailings dam stability based on clusters of critical rainfall pattern curves. Real-time graded warnings are achieved by mapping measured rainfall intensity and duration onto this zoning map. Graded early warning analysis was conducted for a generalized tailings dam using rainfall monitoring data from a provincial meteorological bureau. The study shows that ignoring the variability of saturated hydraulic conductivity underestimates the landslide risk of tailings dam under continuous rainfall. Compared to considering only the variability of effective cohesion and effective internal friction angle, incorporating the variability of saturated hydraulic conductivity reduces the areas of stable, basically stable, and less stable zones by 7.8%, 53.0%, and 64.2%, respectively, while increasing the area of the unstable zone by 14.5%. The proposed method supports real-time graded warnings for tailings dam under rainfall, can be easily integrated with real-time online monitoring systems, and shows good potential for engineering applications.