An oceanographic three-dimensional variational data assimilation scheme

This study describes the development and evaluation of an oceanographic three-dimensional variational (3D-VAR) data assimilation scheme based on a novel specification of the background error covariances. The new 3D-VAR scheme allows for regional variability of the background error covariance matrix, complex coastal boundary conditions and variable bottom topography. The error covariance matrix is formed by the successive application of linear operators that can consider vertical EOFs, horizontal covariance functions that consider coastlines, sea level corrections that vary from shallow to deep regions and divergence dumping of velocity corrections near the coasts. The scheme is applied to the Mediterranean Sea and the quality of analysis is assessed by comparing background estimates with observations in the period October 2005-October 2006. (c) 2008 Elsevier Ltd. All rights reserved.