INVERSE ESTIMATION OF BED ROUGHNESS COEFFICIENTS OF OPEN-CHANNEL BY USING ADJOINT MODEL IN SHALLOW-WATER FLOWS

摘要

This work describes and explains the methodology of an inverse estimation of distributed bed roughness coefficients of open-channels with flood plains. The Lagrange multiplier method is employed for the parameter identification. A shallow-water model is utilized as the constraint condition, and a variational approach enables us to develop the adjoint model. The requisite gradient information to satisfy the optimal condition is efficiently obtained through the adjoint equations. The optimal coefficients are determined by the quasi-Newton method. The identical twin experiments were carried out by using synthetic data in order to verify the validity of the proposed method. The assimilated data consist of the water level and the depth-averaged velocity calculated by the forward model. It is clear that the coefficients can be accurately predicted and that these estimates are stable with respect to random noise in data, provided that sufficient data are available at more observation stations.