基于参数不确定性的概率预报研究

Study on Probabilistic Forecasting Accounting for Parametric Uncertainty

水文模型参数的选取通常依靠经验判断或者依赖历史库中的不完备数据集进行自动优选,所选参数并不一定能够准确反映流域降雨径流特点,更不足以反映不同洪水涨落阶段洪水特征的变化。基于水文模型参数存在显著不确定性的客观事实,以随机参数驱动水文模型,并结合数值模型实现概率预报。通过东湾流域36场洪水模拟试验,揭示了水文参数不确定性对洪水预报结果的显著影响,提出的概率预报算法能够给出精确、可靠的预报结果,说明该算法能够降低水文模型参数所带来的洪水预报不确定性。

In the actual flood forecast, the optimization of parameters is often determined by empirical judgment or dependent on the incom- plete data set of history library. However, due to the fact that the flood characteristics cannot be accurately predicted, even there is a big difference for the parameters required at different stages of flood, so the selected parameters are not necessarily suitable for the flood forecast of the current flood. In this study, the objective facts of the uncertainties based on the hydrological model were analyzed, the hydrological model was driven by the randomly generated parameters and the ensemble forecast was realized by combining the numerical model. This study revealed the significant effect of hydrological parameters uncertainty on the flood forecast results through 36 events of flood simulation test in the Dongwan basin. The results show that the probabilistic prediction algorithm based on the uncertainty of the parameters is accurate and the reliability is high, which indicates that the algorithm can reduce the uncertainty in flood forecasting caused by parameters of hydrological model.