关键词: 信息矩阵方法, 信息分配, 山洪灾害, 风险预警, 淠河流域

Abstract: Identification of the early warning indicators of flash flood disasters risk warning is a key technical issue in the flood control and disaster mitigation. However, because of the short time series and the small number of flash flood disasters, it is difficult to use the traditional statistical model to establish the flood disaster risk early warning method. An algorithm relates to relationships between inputs and outputs and the input-output relationship can be mathematically represented by functions. The function relationship can be established based on information matrix method which is able to get information from a small number of disaster samples. In this paper, a novel approach, called information matrix, is introduced to recognize the function relationship. Based on the rainfall and flow data of the 17 flash flood events between 2003 and 2012, the fuzzy relation matrix is constructed by learning from the samples of the antecedent soil moisture saturation, rainfall of five durations （1, 3, 6, 12, 24 h） before the occurrence of flash flood and the estimates of the peak flow of four return periods （＜5, ≥5, ≥20, ≥50 a）. 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.

Key words: information matrix, information distribution, flash flood disaster, risk warning, Pihe basin