RESUME / ABSTRACT  


Issues for next generation DA for NWP


Andrew Lorenc
Met Office, UK



In this talk I aim to:

*Show the linear and adjoint models in 4D-Var are just part of a 4D covariance model.

*Discuss important properties of error covariances in statistical assimilation. Use analogy with geostrophic constraint in 3D-Var to understand role of model in 4D-Var.

*Understand important properties to be retained when approximating.

*Review the use of dynamical ensembles to represent forecast error covariances, in the EnKF.

*Discuss the strengths and weaknesses of the EnKF compared to 4D-Var.