Abstract: Identification of the early warning indicators of meteorological risks is a key technical issue in the forecast of flash flood disasters. The antecedent precipitation is used to characterize the degree of antecedent soil moisture saturation of a river basin, which is divided into four levels based on the quartile method. Frequency analysis based on the Pearson Ⅲ distribution (P-Ⅲ) model yields the estimates of the peak flow of four return periods (< 5, ≥5, ≥20, and ≥50 a) and rainfall of five durations (1-, 3-, 6-, 12-, and 24 h) before the occurrence of flash flood. The four return periods of peak flow are used to characterize the four meteorological risk levels (Ⅳ-, Ⅲ-, Ⅱ-, and ⅠLevel) of flash flood disasters. The information diffusion technique is used to establish the matrices of meteorological risk levels, antecedent soil moisture saturation degrees, and rainfall of different durations to form a combined indicator for the warning meteorological risks. Based on historical records of rainfall and extreme flow in the Pihe basin, 17 flash flood events between 2003 and 2012 are analyzed with the proposed method to derive the four-level meteorological risk warning indicators for flash flood at five time steps. The tests of the derived risk warning indicators in 40 independent historical flash flood events at different time steps have yielded a correct warning rate of over 70%, which demonstrates the feasibility of using the method for flash flood disaster warnings.