The nested model actually involves two simulations running simultaneously. The main run is set up on the whole area to be studied, i. e. a whole oceanic basin or the largest area we can afford. The grid on which this calculation is executed can be relatively coarse and we can afford not to represent some phenomena in the area of interest.

The secondary model is set up on a much smaller area, i. e. the area subject to local study itself. The numerical grid for this run can be finer than the grid of the main simulation and can have as many open sea boundaries as needed. The conditions at those boundaries will be given by interpolating results from the main model at each time step.

For the two models, the horizontal grid must be regular along both zonal and meridional directions. The implementation of the nesting procedure in the numerical model allows for any odd integer ratio between the size of the cells in the main model and their size in the nested model. The vertical discretization in the two models must of course be the the same.

One disadvantage of this implementation is that the equations must be solved twice in the area of interest. This should not be too much of a disadvantage since the coarse model is assumed to be far cheaper to solve than the fine one, due to its larger cells and the relative small size of the nested area. The calculations in the coarse model are so faster than in the nested one and the additionnal cost should be negligible.

The merging grids are presented at figure 4.1

**Figure:** Merging grids;
**+ **denotes a scalar variable in the main model;
+ denotes a scalar variable in the nested model;
arrows denote velocity or fluxes of scalar variables components.

Wed Jun 10 10:53:44 DFT 1998