张苹苹 童菊秀 廖卫红
摘要:数据同化方法可提高数值预报的时效性和准确性,且该方法已在水文领域得到应用,并得到快速发展。为了提高新安江模型径流模拟预报精度,采用集合卡尔曼滤波方法同化径流数据,对参数和状态变量进行同步校正估计。通过对三水源新安江模型进行理想条件下的数值实验,在同时考虑模型自身、模型参数以及观测数据的不确定性的情况下,分析了参数均值和方差改变、集合大小、同化参数的敏感性以及相关性分析对同化过程的影响。结果表明:集合卡尔曼滤波算法具有可行性,且参数均值越接近真值、方差适当增加,集合大小适中,同化参数敏感性较低以及参数与变量间相互独立时,能在一定程度上增加径流同化精度。该研究可为同类型参数同化估计提供一定参考依据。
关键词:新安江模型;径流;集合卡尔曼滤波;同化;同步校正
中图分类号:TV124文献标志码:A文章编号:
16721683(2018)02005108
Abstract:
The data assimilation method can improve the timeliness and accuracy of numerical forecasting,and has been applied and developing rapidly in the field of hydrology.In order to improve the accuracy of runoff forecast of Xinanjiang model,we adopted the ensemble Kalman filter method for synchronous correction of the model parameters and state variables.We designed a numerical experiment of the threecomponent Xinanjiang model under ideal conditions,and analyzed the effects of the mean and variance of parameters,the ensemble size,and the sensitivity and correlation of parameters on the data assimilation with consideration to the uncertainty of the model itself,model parameters,and the observation data.Results showed that the ensemble Kalman filter algorithm is feasible.Moreover,the accuracy of data assimilation can be improved when the mean value of the parameter is closer to the true value,the variance is increased appropriately,the ensemble size is proper,the sensitivity of parameters
is lower,and the correlation between parameters and variables is small.This study can provide a reference for similar research of data assimilation.
Key words:
Xinanjiang model;runoff;ensemble Kalman filter;data assimilation;synchronous correction
水文模型是描述自然界水文系統非常重要的工具,但模型本身具有不确定性,这些不确定性主要来源于模型结构、模型输入数据和模型参数[1]。数据同化方法经过近几十年的发展,在模型多源不确定性方面已取得较多成果。其中卡尔曼滤波(KF)[2]作为统计估值理论的代表性序列方法,已在模型参数实时校正中得到良好的应用[34]。KF作为线性最小方差估计方法,主要用于具有或近似满足高斯分布误差的线性系统。现实中各物理模型大多是非线性系统,为此,扩展性卡尔曼滤波(EKF)[5]在KF的基础上发展而来。……