|
| 基于多元径流组分与模拟驱动的流量历时曲线构建研究 |
| Construction of Flow Duration Curves Based on Multi-Component Streamflow and Simulation-Driven Hydrological Modeling |
| 投稿时间：2024-10-07 修订日期：2025-08-04 |
| DOI： |
| 中文关键词: 流量历时曲线 Vine Copula结构 DFI基流分割 物理过程 汉江流域 |
| 英文关键词: Flow duration curve Vine Copula structure Delay flow separation Process control Hanjiang basin |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| 摘要点击次数: 54 |
| 全文下载次数: 0 |
| 中文摘要: |
| 流量历时曲线（Flow Duration Curve, FDC）是反映流域水文过程变化和水资源状况的有效方法之一，为水资源工程设计和水库调度策略优化提供重要参考。为实现高精度预测未来时段的FDC，本研究聚焦于改进现有预测方法，以提升FDC对流域水文过程的表征能力。传统方法通常基于流量模拟值直接构建 FDC，忽略了流域中多元径流组分之间复杂的依赖关系和空间异质性，导致流量时序在相位与量级上的刻画偏差，进而影响 FDC 的构建精度。针对这一问题，本文提出一种基于多元径流组分的mixture Copula模型，用于准确构建FDC。先利用水文模型HYMOD模拟汉江流域的日尺度流量数据，为mixture Copula模型提供输入项。通过延迟流指数DFI基流分割方法，将流量模拟值细分为短、中、长和基线延迟流，并使用Vine Copula结构这一多元统计建模方法对径流组分进行联合建模，以捕捉其相互关系。旨在从考虑FDC的物理组分出发，提高FDC构建的精准度，为未来时段FDC的准确预测提供支持。研究结果表明：在率定期应用mixture Copula模型，均方根误差RMSE值平均降低26.5%，有效修正了基于流量模拟值构建FDC的偏差；在低流量相位，与基于流量模拟值构建的FDC相比，RMSE_low值下降约90%。此外，该模型在不同时段内均表现出良好的鲁棒性，为FDC的模拟与预测提供了一种更加精确有效的新方法。 |
| 英文摘要: |
| The Flow Duration Curve (FDC) is an essential tool for reflecting changes in hydrological processes and water resource conditions assessment. Due to the complexity of hydrological processes, accurate prediction of the FDC is a challenge. To address the issue, a process-based mixture Copula model is developed to improve the predictive accuracy of the FDC by analyzing and capturing the influence of multiple streamflow components on the shape of the FDCs. The HYMOD model is taken as an example to predict daily streamflow for the Hanjiang River Basin, which serves as the input for the mixture Copula model. The delay flow separation is employed to divide the predicted streamflow into components representing different water sources, the dependence among these components is captured using an optimal Vine Copula. The results indicate that: During the calibration period, the application of the mixture Copula model resulted in an average reduction of RMSE by 26.5%, correcting biases generated in FDCs constructed directly from predicted streamflow, particularly in the low streamflow, the RMSE_low is reduced by about 90%. In addition, the model exhibits robustness and stability across different periods, and a novel approach for FDC prediction is presented. |
|
查看/发表评论 下载PDF阅读器 |
| 关闭 |