Abstract: The debris flows at small watersheds are mostly triggered by short-term heavy rainfall, but the studies regarding the rainfall thresholds and hazard assessment still remain some uncertainties. To overcome this issue, the Baishegu debris flow in Jizhou District of Tianjin City in 2024 was taken as an example. The probability of failure (PoF) at the watershed was calculated by using a physically-based model, and the short-term rainfall thresholds under different antecedent rainfall conditions were determined through the PoF threshold at 0.5. Moreover, the rainfall threshold curves were also established for classifying four-level warnings. The movement process of the Baishuigu debris flow was simulated using the FLO-2D software, with the discharge in the source area determined by the calculated surface runoff and a correction coefficient accounting for on-site conditions. Based on the simulated spatial distributions of flow depth (h) and velocity (v), debris flow hazard zonation when using four indicators (hv, hv², impact force F, and decision matrix) were analyzed and compared. Additionally, the response patterns of debris flow hazard to three monitoring indicators (short-term rainfall, soil saturation, and runoff) were quantified. The results showed that the short-term rainfall triggering threshold of debris flows at small watersheds had a strong negative linear correlation with antecedent rainfall (R²≥0.92). Under the very high warning level, the short-term critical rainfall decreased from 81.6 mm (antecedent rainfall: 0.1 mm/d) to 44.4 mm (antecedent rainfall: 0.5 mm/d). The combined FSLAM (Fast Shallow Landslide Assessment Model) and FLO-2D method effectively simulated the kinematic process of the debris flow, and the velocity and depth at specific locations coincide with field observations. The hazard assessments based on impact force and decision matrix yielded relatively "conservative" and "radical" zonation outcomes, respectively, while those based on hv and hv² were more reasonable. Debris flow hazard responded distinctly to all indicators, but was more sensitive to soil saturation changes when antecedent rainfall was low. These findings reveal key uncertainty sources and optimization directions for debris flow hazard assessment, providing crucial references for the dynamic identification of rainfall thresholds and meteorological early warning of debris flows in small watersheds.