Inversing source intension and longitudinal dispersion coefficient of 1-D stream water quality model using adjoint data assimilation method

The adjoint data assimilation method is proposed to improve the numerical model accuracy by inversing parameters. The adjoint data assimilation method includes: (1) defining the difference value between simulation and observation values as objective function; (2) using the Lagrange operator method to transform the inversing problem with constraint equation into unconstrained control variables optimization problem through minimizing the objective function; (3) applying the steepest descent method to solve the control variables optimization problem. The performance of the adjoint data assimilation method is evaluated by means of test examples, including inverse pollution source intension and longitudinal dispersion coefficient in one-dimension stream water quality model. The case of pollution source intension inversing indicates that the adjoint data assimilation method can fmd multi-variables successfully with great robustness and applicability. In the second case, the adjoint assimilation method is applied to predict the longitudinal dispersion coefficient in one-dimension stream water quality model. The observed data is from the experiment of pollutants concentration by Guymer. The results of the adjoint data assimilation method agree better with the observed values than that of the artificial trial method. It is demonstrated that the model parameters inversed by adjoint data assimilation are closer to the precise values.