Abstract: Early warning indicators for mountain flood disasters are the core of mountain flood disaster forecasting and early warning, directly related to the timeliness and effectiveness of mountain flood disaster early warning. Due to the large number of small watersheds and the difficulty in determining model parameters and disaster-causing water levels and flows, to quickly determine the dynamic early warning indicators for mountain floods in small watersheds, a small watershed runoff generation and concentration model was constructed. The Antecedent Precipitation Index (API) model was used to calculate runoff generation, and the unit hydrograph of small watersheds was automatically extracted based on topographic features to simulate the mountain flood process in small watersheds. Through continuous simulation trials, combined with the surveyed disaster-causing flow in small watersheds, a method for determining the disaster-causing flow in small watersheds considering the water exchange between upstream and downstream watersheds was proposed to achieve the dynamic back-calculation of the critical rainfall for disaster-causing. Taking the Lianjiang River Basin in Anhui Province as an example, combined with actual and forecasted rainfall, static and dynamic early warning for mountain floods in small watersheds was achieved, providing an effective means to improve the accuracy of early warning for mountain floods in small watersheds in Anhui Province.