基于数据同化校正参数的河流磷迁移估计研究

通过磷迁移数学模型合理估计磷在河流中的时空分布,对防治水体富营养化,抑制水华暴发具有重要的科学和工程意义。数据同化方法可以将模型和观测两种研究手段有机地结合起来,将观测数据融入模型,优化模型状态变量,校正模型参数,进而提高数学模型的模拟预报精度,并依托物联网技术将传统数学模型发展为实时数学模型。本文将粒子滤波数据同化算法引入到水动力-泥沙-磷迁移模型中,以实测的断面磷含量作为观测数据,在观测时刻优化磷含量估计结果,同时校正模型参数磷相平衡分配系数Kd,构建了水动力-泥沙-磷迁移模型同化系统。将其应用于长江上游寸滩至坝前河段的计算结果表明,所构建的同化系统在真实的河流中计算效果良好,可以有效地优化更新状态变量各相磷含量浓度,并反演出模型参数Kd随水沙水环境条件变化的动态变化过程,同化之后模型模拟预报磷输移过程的精度显著提升,为水质实时模型打下基础。

Estimation of phosphorus transport in rivers with parameters updating based on data assimilation

It is of great importance to enhance the numerical phosphorus transport models to estimate the temporal and spatial distribution of phosphorus in rivers in order to prevent and control eutrophication and algal blooms in water environment. Data assimilation is a new technology to combine the numerical modelling with observation, which is able to improve the accuracy of model output by incorporating the observations into numerical model to update the model states and correct the model parameters. In this study,Particle Filter (PF),a sequential Monte Carlo data assimilation algorithm,is employed to combine the numerical hydrodynamic-sediment-phosphorus model with phosphorus observations to develop a data assimilation system to improve the estimation of phosphorus concentration in rivers. When the phosphorus observation becomes available,the model state variable,phosphorus concentration,will be updated and the model parameter, partition coefficient K_d, will be corrected according to the PF theory. The developed data assimilation system is applied to the Changjiang River segment from Cuntan to Three Gorges Dam to evaluate its performance of estimating phosphorus transport in a real event. The results show that the developed data assimilation system can update phosphorus concentrations and correct K_d effectively and dynamically at the assimilation time. After assimilation,the accuracy of estimation of phosphorus transport can be enhanced significantly due to the effect of assimilation, indicating the developed data assimilation system has a good performance in the real event.