Due to the strong spatial and temporal heterogeneity of storm-flash flood process, accurate assessment of rainfall and flood variation characteristics and flood hazard is of great significance for flash flood prevention and control. Six rainfall characteristic indices and six flood characteristic indices were used to describe the rainfall and flood variability of flash flood events, and China Flash Flood hydrological model was combined with flood frequency indices to simulate the flood processes and assess the corresponding flood hazard in Kouqian catchment. Results showed that the modulus of flood peak, skewness of flood peak time, proportion of high pulse duration and rates of flood rising and declining processes were significantly correlated with rainfall characteristic indices, such as total rainfall depth, average rainfall depth, maximum rainfall intensity, rainfall peak location coefficient, and Gini coefficient. Specifically, the rainfall events of three catastrophic flash flood events were characterized by great magnitude, high intensity and short duration. In the calibration and validation periods, the average relative errors of runoff depth were within 9%, the average relative error of peak flow were within 11%, the average error of time to peak flow were within 1.7 h, and the average Nash-Sutcliffe efficiency coefficients were 0.80 and 0.76, respectively. For all the flood events, the flood magnitudes of 0.0%~93.3% of the river reaches were at the general hazard grade and above, and the hazard grades were the highest for the three catastrophic flood processes, with the flood magnitudes of 80.0%, 35.0% and 1.7% of river reaches at high hazard grade above respectively. The research can provide technical supports for flood hazard assessment, disaster response and review of storm-flash floods in small mountain catchments.