暴雨山洪灾害预警方法研究进展

暴雨山洪灾害预警是中小流域山洪灾害防控体系的薄弱环节，也是决定山洪灾害防控成败的关键。论文围绕山洪灾害预警的核心问题，从中国山洪灾害区域差异特征、山洪灾害预警技术方法、山洪灾害概率预警现状3个方面进行了综述。中国山洪灾害分布存在明显的时空差异，因此有必要根据山洪灾害的区域差异发展有针对性的预警方法。以临界雨量为指标的雨量预警是目前中国中小流域暴雨山洪灾害预警的主要技术手段，但常规方法仅给出一个(组)确定的临界雨量阈值，导致预警结果存在突出的不确定性问题。概率预警可以定量评估诸多不确定性，给出山洪灾害概率预警结果，因此具备很好的理论优势与潜在应用价值。论文展望了山洪灾害概率预警未来的研究重点与方向：(1)充分挖掘暴雨洪水样本信息，开展山洪灾害概率预警基础方法与技术集成研究；(2)加强非平稳性条件下的临界雨量阈值估算与山洪灾害概率预警研究；(3)综合考虑预警阈值发生概率及其致灾概率，优化“多级预警、多级响应”技术方法，推进山洪灾害综合预警业务系统建设与应用。

Progress of research on the methods for the early warning of mountain flash flood disasters

Accurate and timely warning is vital for the prevention and control of mountain flash flood disasters in small- and medium-sized basins, but it is the most challenging task. Focusing on the key issues of mountain flash flood warning, this study conducted a review on the spatial differences of mountain flash flood disaster characteristics in China, techniques and methods for flash flood warning, and development of probabilistic warning methods for mountain flash flood control. The review indicated that the temporal and spatial differences of the characteristics of mountain flash flood disasters are clear across China, thus it is necessary to develop suitable early warning systems in specific areas based on their special characteristics. The indicator of critical rainfall is an important basis of mountain flash flood warning in China. A fixed critical rainfall estimated by statistical methods or hydrological models, as commonly done in practice, cannot meet the needs because it does not perform well sometimes and cannot adequately address the uncertainty of warning results. Alternatively, probabilistic warning of mountain flash flood disasters, which can quantitatively evaluate the uncertainty of warning results, has theoretical advantages and potentially wide applications, and thus could be a more adaptive approach. In order to develop the probabilistic warning methods of mountain flash flood disasters, the following key issues should be addressed: 1) Mining detailed information from different sources of rainstorm and flash floods data, and developing effective methods for the probabilistic warning of mountain flash flood disasters. 2) Dealing with the nonstationarity issue commonly encountered in hydrological studies, and exploring its influences on the estimation of critical rainfall and the accuracy of probabilistic warning results. 3) Establishing an integrated flash flood warning system by considering both the occurrence probability of critical rainfall and its resulting probability of mountain flash flood disasters for the practical prevention and control of mountain flash flood disasters.