Real-time flood forecasting requires accurate and reliable estimates of the uncertainty to make efficient flood event management strategies. However, the accuracy of flood forecasts can be severely affected by errors in the estimates of sediment yield in the loess region. To improve the accuracy of sediment-laden flood forecasts generated using streamflow-sediment coupled (SSC) model, an error feedback correction method based on the dynamic system response curve (DSRC) is proposed. The physical basis of the system response curve is the sediment concentration of the hydrological model. The theoretical basis of the method is the differential of the system response function of the sediment yield time series. The effectiveness of DSRC method is evaluated via an ideal case and three real-data cases with different basin scales of the Yellow River. Results suggest that the DSRC method can effectively improve the accuracy and stability of sediment transport forecasts by providing accurate estimates of the sediment yield errors. The degree of forecast improvement is scale dependent and is more significant for larger basins with lower rain gauge densities. Besides, the DSRC method is relatively simple to apply without the need to modify either the model structure or parameters in real-time flood forecasting.

实时洪水预报需要准确可靠地估计不确定性，以制定有效的洪水事件管理策略。但是，黄土地区泥沙产量估算的误差会严重影响洪水预报的准确性。为了提高利用流沙耦合模型（SSC）生成含沙洪水预报的准确性，提出了一种基于动态系统响应曲线（DSRC）的误差反馈修正方法。系统响应曲线的物理基础是水文模型的沉积物浓度。该方法的理论基础是沉积物产量时间序列的系统响应函数的微分。通过一个理想案例和三个黄河不同流域规模的实际数据案例，评估了DSRC方法的有效性。结果表明，DSRC方法可以通过提供泥沙产量误差的准确估算来有效地提高泥沙输送预测的准确性和稳定性。预报改进的程度与规模有关，并且对于具有较低雨量规密度的较大盆地更为重要。此外，DSRC方法的应用相对简单，无需在实时洪水预报中修改模型结构或参数。