综合多源不确定性的洪水概率预报试验

为探究多种不确定性因素综合影响下洪水概率预报的实现方法及其适用性，利用2003—2020年史灌河流域多场次洪水开展模拟试验，在量化降雨输入、模型参数及结构不确定性程度的前提下，基于MCMC(Markov chain Monte Carlo)方法融合各来源不确定性实现洪水概率预报，并分别基于覆盖率、确定性系数指标评估预报结果的可靠性及预报精度。结果表明:综合多源不确定性洪水概率预报的5%~95%置信区间，能够较为稳定也包络预见期内洪水发展变化过程，有效降低误报、漏报的可能性；均值预报结果所提供确定性的预报洪水略优于原始的单一模型预报结果，进一步提高了洪水预报成果的可靠性及参考价值。

Experiments on flood probability forecasting accounting for multi-source uncertainty

To explore the realization method and applicability of the flood probability forecast under the comprehensive influence of various uncertain factors, the study was carried out in Shiguanhe River Basin from 2003 to 2020, on the premise of quantifying the input of rainfall, model parameters and uncertainty degree of structure. The probabilistic flood forecasting is realized by fusing the uncertainty of every source based on the MCMC method, while the reliability and accuracy of the prediction results are evaluated based on the coverage ratio and the certainty coefficient, respectively. The results show that synthesizing the 5% to 95% confidence interval of the probabilistic flood forecasting method can predict the flood development process in a stable envelope and reduce the possibility of false alarm and missing alarm effectively. The result of mean value forecast is better than that of original single model forecast, which improves the reliability and reference value of flood forecast.