一维水沙耦合模型参数敏感性分析——以2020年黄河下游洪水演进为例北大核心CSCD

一维水沙输移及河床冲淤模型中有较多理论上尚未解决的参数,而这些参数对一维模型精度的影响缺少定量分析。本文首先介绍了一维水沙耦合动力学模型,并将其应用于2020年汛期黄河下游的水沙过程模拟。计算结果表明:该模型能准确模拟水位和流量的变化,纳什效率系数基本大于0.85,均方根误差比相应的实测特征值小一个量级;计算沙峰虽然偏大,但计算的含沙量过程与实测过程整体符合,且模型计算的分河段冲淤量与断面地形法实测值较为接近。为分析模型中经验参数对模型计算结果的影响,以水文断面特征水沙参数的纳什效率系数作为目标函数,采用Sobol全局敏感性分析方法计算了各经验参数的敏感度指标。计算结果表明:糙率的总敏感度指标较高,是模型的高敏感参数;目标函数的选取对参数敏感度分析有一定影响,以含沙量纳什效率系数为目标函数时,挟沙力公式的系数及淤积时恢复饱和系数公式中指数的一阶敏感度和总敏感度指标较高;各参数组合对不同目标函数的影响差别较大,但对于同一目标函数存在影响作用相似的参数组合。全局敏感性分析方法刻画了模型参数的敏感性特征,有助于优化水沙数学模型参数取值。

Parameter sensitivity analysis of dynamically coupled one-dimensional morphodynamic model. Case study of lower Yellow River

Some parameters used in the numerical simulations of fluvial process are usually determined empirically, but their effects have not been examined quantitatively in previous studies. This paper describes a 1D morphodynamic model based on the dynamical coupling of flow and sediment transport, and discusses its simulations of flood routing processes in the lower Yellow in the 2020 flood season. Simulation results indicate (i) time variations in water level and discharge are accurately reproduced, with Nash-Sutcliffe Efficiency Coefficients (NSEs) larger than 0.85 and RMSEs less than 15% of the corresponding mean values. (ii) Although the simulated peak sediment concentration is larger than its measured value, an insignificant difference exists in the patterns of simulated and measured hydrographs. (iii) The calculated volume of riverbed deformation is close to its value measured using the method of cross-sectional riverbed topography. The global sensitivity analysis method proposed by Sobol is adopted to evaluate the effects of some input parameters on the simulations, by taking the NSEs of hydrological indicators at the hydrometric stations as the objective functions. Our findings indicate the importance of an input factor depends on the choice of objective functions. Comparison of sensitivity indexes calculated using different objective functions reveals comprehensive roughness is most sensitive among all the input parameters. When the objective function is the NSE of sediment concentration, the parameters of higher first-order index and total index values are the coefficient of the sediment transport capacity formula and the exponent for recovery coefficient calculation. Generally, a certain parameter combination has quite different sensitivity behaviors under different objective functions, while different parameter combinations could generate comparable effects under the same objective function. Global sensitivity analysis demonstrates the individual effects of different parameters and their interactive effects on the model results, helping improve model calibration for simulations of fluvial process.