Some numerical enhancements in a data assimilation scheme

In this work we show preliminary numerical improvements in OceanVar, a 3D-Variational Scheme for data assimilation. The original assimilation scheme computes the horizontal error covariances by using a recursive gaussian filter of the first order (1st-RF) with the use of the ghost points in presence of discontinuities, represented by the costs. The new proposed features are based on the implementation of a third order recursive gaussian filter (3rd-RF). An advantage of the 3rd-RF, to compute the horizontal error covariances, is that computational time can be substantially reduced with benefits to large scale applications.