Abstract: Rainfall is recognized as a primary triggering factor for landslides, making early warning of rainfall-induced hazards is essential.Based on rainfall monitoring data from meteorological stations and disaster records in the Qinling-Daba Mountains Area, a probability model for rainfall-induced landslides was established.Subsequently, a comparison of the warning efficiency between the proposed model and the traditional EE-D (Early Effective Rainfall-Duration) graded threshold model was conducted, using the count of disaster sites corresponding to different warning levels as the evaluation metric.The results indicate that the probability of rainfall-induced landslides in the Qinling-Daba Mountains Area follows an S-shaped distribution surface.Under conditions of low rainfall inducement, the warning efficiency of the proposed probability model is comparable to that of the traditional model.However, under conditions of high-intensity rainfall inducement, the probability model demonstrates superior warning efficiency.Moreover, regarding short-duration, high-intensity rainfall, the traditional model holds a certain advantage over the probability model.When it comes to long-duration, high-induced rainfall conditions, however, the probability model shows greater superiority compared to the traditional model.This difference is attributed to their distinct delineation criteria.The traditional model uses the average process of disaster-triggering rainfall as its delineation criteria, which leads to an underestimation of the hazard risk associated with long-duration rainfall.In contrast, the probability model employs the frequency of disaster occurrence as its delineation criteria, making it more consistent with the increasing trend in quantity and severity of hazards.Consequently, the proposed probability model achieves higher warning efficiency for severe geological hazard events, thereby enhancing early warning accuracy and reducing associated costs.