基于Copula函数的多模型多时段洪水联合概率预报及风险预警研究

The commonly used flood warning paradigm at present is the flood characteristic (e.g., peak discharge) warning based on the deterministic forecast result. However, there are some limitations in the evaluation of flood warning risk due to not fully considered uncertainty of flood forecasting. Therefore, it is of great significance to quantify and reduce the uncertainty of flood forecasting and produce flood warning considering the forecast uncertainty. In this project proposal, the ensemble precipitation forecast information will be adopted to analyze the uncertainty of quantitative precipitation forecast, and bias correction and multi-member information integration in a probabilistic manner are used to improve the performance of the raw ensemble precipitation forecast. The corrected ensemble precipitation forecast is randomly generated from the aforementioned precipitation probability distribution, and the spatial-temporal correlation will be reconstructed. Then, the multi-model multivariate hydrological uncertainty processor based on Copula function (Copula-MMHUP) will be derived to quantify the hydrological uncertainty of joint flood forecasting for multiple lead times under the condition of multi-hydrological model combination forecasting. On this basis, the integrated joint probabilistic flood forecast for multiple lead times simultaneously considering two uncertainty components will be obtained by coupling the corrected ensemble precipitation forecast and the Copula-MMHUP. In addition, the exceeding probability distribution function of the peak discharge within the concerned time interval will be derived to construct the risk-based flood warning paradigm based on the probabilistic forecast information. Research results can not only reduce the uncertainty of flood forecasting, improve the accuracy of flood forecasting and the reliability of flood warning, but also provide scientific basis for flood risk assessment, flood control and disaster reduction.

目前常用的洪水预警模式是基于确定性预报结果的洪水要素预警，没有考虑洪水预报结果的不确定性，在评估洪水预警可靠性方面存在局限性。因此，如何量化和降低洪水预报不确定性，实现考虑预报不确定性的洪水预警具有重要意义。本项目针对洪水预报不确定性和洪水预警面临的关键科学问题，引入降水集合预报分析定量降水预报不确定性，采用贝叶斯模型平均（BMA）方法对原始降水集合预报进行偏差订正和概率集成，并重建校正后降水集合预报的时空相关性。基于Copula函数建立多模型多变量水文不确定性处理器，量化多模型组合预报条件下多时段洪水联合预报的水文不确定性；将两种不确定性进行耦合得到考虑总不确定性的多时段洪水联合概率预报。在此基础上，推求预见期时段内洪峰流量的超过概率分布函数，构建概率预报条件下的洪水风险预警模式。研究成果可以降低洪水预报不确定性，提高洪水预报精度和洪水预警可靠性，为洪水风险评估和防洪减灾提供科学支撑。

目前常用的洪水预警模式是基于确定性预报结果的洪水要素预警，没有考虑洪水预报结果的不确定性，在评估洪水预警可靠性方面存在局限性。因此，如何量化和降低洪水预报不确定性，实现考虑预报不确定性的洪水预警具有重要科学意义和应用价值。本项目针对洪水预报不确定性和洪水预警面临的关键科学问题，评估了定量降水预报在典型研究区域的预报性能，构建了基于Copula-贝叶斯处理器的定量降水概率预报模型，量化了降雨径流模拟不确定性，揭示了降雨径流模拟中不确定性的季节性规律及Copula函数选择和相关系数的影响，实现了集合多水文模型信息的洪水概率预报，发展了概率预报条件下的洪水风险预警模式。结果表明，基于Copula-贝叶斯处理器可以很好地量化定量降水预报和降雨径流模拟的不确定性，选择合适的Copula函数和有效地捕捉实际依赖关系非常重要。集合多水文模型信息的洪水概率预报可靠性高，综合指标连续概率排位分数结果相较于确定性预报明显减小，改善效果明显。基于Copula函数的多变量水文不确定性处理器为分析水文预报不确定性在时间上的演变特征提供了有效工具，基于概率预报的洪水风险预警模式考虑了洪水预报的不确定性，可以给出洪峰流量超过任一级别预警阈值的可能性大小，能在一定程度上减少漏报和空报，据此作出的洪水预警决策更稳健、可靠性更强。研究成果有助于延长洪水预见期、提高预报精度和洪水预警的可靠性，为洪水风险评估和防洪减灾提供科学支撑。