一种稳健的河流糙率反演方法

A robust river roughness calibration model

为了提高河流糙率反演的数值稳定性及结果合理性,以糙率空间分布平滑性作为反演正则化所需的先验信息,建立一个兼顾水文观测信息和糙率分布平滑性的河流糙率反演的目标函数,利用广义逆理论和敏度矩阵的概念进一步确定目标函数的下降方向,并采用拟牛顿法来逐步寻优.通过一个由14条河段组成的河网算例,系统分析模型解的唯一性、数值稳定性和结果可靠性等特性.结果表明:引入河流糙率空间分布平滑性可以有效抑制噪声,提高计算的稳定性,且计算精度较高,结果可靠.

For improving the stability of numerical computation and making the calibrated results reasonable,a novel robust model which takes both error minimization and roughness smoothness matrix into account has been established based on the prior information that roughness parameters within the smoothest space distribution.It uses generalized inverse theory and sensitive matrix to analyze the optimization orientation and uses Guasi-Newton method to optimize step by step.In a computing example of a river network composed of 14 channels,cases were tested about the uniqueness of solution,numerical stability and reliability.The results show that the present model can control numerical noise effectively,and the calibrated roughness is accurate and reliable with high numerical stability.