Articles | Volume 12, issue 4
https://doi.org/10.5194/npg-12-481-2005
https://doi.org/10.5194/npg-12-481-2005
13 May 2005
 | 13 May 2005

On deterministic error analysis in variational data assimilation

F. -X. Le Dimet and V. Shutyaev

Abstract. The problem of variational data assimilation for a nonlinear evolution model is considered to identify the initial condition. The equation for the error of the optimal initial-value function through the errors of the input data is derived, based on the Hessian of the misfit functional and the second order adjoint techniques. The fundamental control functions are introduced to be used for error analysis. The sensitivity of the optimal solution to the input data (observation and model errors, background errors) is studied using the singular vectors of the specific response operators in the error equation. The relation between "quality of the model" and "quality of the prediction" via data assimilation is discussed.