黄河流域水沙变化趋势多模型预测及其集合评估

黄河水沙变化情势深刻影响着黄河流域水沙调控工程布局、流域内外水资源配置和跨流域调水工程建设等黄河保护与治理开发重大问题的决策。但受研究时段、方法及其边界条件等影响,黄河水沙变化预测成果差异大,难形成共识。本研究基于水沙变化归因和预测结果的不确定性解析,构建了多模型预测成果集合评估技术,预测了黄河流域未来50 a水沙量。结果表明:既有流域水沙变化归因与预测成果差异大,原因在于不同方法因数据输入、变量构成及精度评价方法差异带来的不确定性;提出了基于输入-结构-输出的模型适用性判别准则和评价技术,并基于标准化的数据输入,从数据需求、物理基础、应用效率、输出尺度和预测精度等五个维度,集合评价了现有水沙变化预测模型的适用性;构建了流域水沙变化多模型集合评价-多结果加权融合-BMA集合预测的集合评估技术,根据9类模型预测成果,集合预测了黄河流域未来50 a年径流量为240亿m³,年均输沙量为2.5亿t。

Multi model prediction and its ensemble evaluation of runoff and sediment change trend in the Yellow River Basin

The variation of runoff and sediment in the Yellow River profoundly affects the decision-making of major issues in the Yellow River protection and development,such as the layout of runoff and sediment regulation projects,the allocation of water resources inside and outside the basin,and the construction of cross-basin water transfer projects.However,the attribution analysis and prediction results of runoff and sediment variation in the Yellow River vary greatly from research to research,since they are affected by many factors such as research period,datum period,and research methods.After analyzing the results of runoff and sediment change attribution and prediction,this study constructs a multi-index and systematic model accuracy evaluation system and proposes a technical framework of ensemble assessment to predict runoff and sediment changes in the Yellow River in the next 50 years based on the 9 types of model prediction results set.The results show that the large difference in the attribution analysis results of runoff and sediment change in the typical Yellow River basin is caused by different data requirements,constituent variables,and the simulation accuracy of the evaluation methods.Thus,the multi-index and systematic model accuracy evaluation system includes the input-structure-output assessment.The applicability of the existing runoff and sediment change prediction models is evaluated from five aspects including data requirements,physical basis,application efficiency,output scale,and prediction accuracy based on the standardized data input.The technical framework of ensemble assessment is finally constructed based on multi model applicability evaluation-multi scenario set-results ensemble prediction,and the average runoff and sediment amount at Tongguan Station of the Yellow River are predicted to be about 24 billion and 300 million tons per year respectively in the next 50 years based on ensemble assessment of 9 types of models.