BDY is not compatible with high order schemes

[smasson]

Context

BDY is not compatible with high order schemes (bilaplacien, fct4...) that requires informations from the neighbor of your neighbor.

The treatment specific to the rim should be applied not only to the rim 1 but also top the rim 2.
This modification would insure that for any point of the inner domain (the free ocean not affected by BDY rim) the neighbor of your neighbor has the appropriated value (and not 0 for land)

Analysis

We should first apply the rim treatment to rim 2 (mixing free ocean values and data values) and next to rim 1 (mixing corrected rim 2 values en data values).

In dorder to facilitate the compatibility with previous versions and in order to keep the possibility to have only 1 rim for simulations using only first order scheme, we propose to add a rim 0 instead of using the rim 2.
We would in any case work on rim 1 and if rim0 is existing we would also modify rim 0.

• Old case adapted to 1nd order schemes:

• New case adapted to 2nd order schemes:

Recommendation

Another side point is the difference of point of view in the definition of the rim in BDY and AGRIF:

• in BDY, we fix the land mask, and add the rim points in the ocean. The larger the number of rim the smaller is the domain of the “free ocean”. The model domain remains constant.
• in AGRIF we fix the limits of the free ocean domain. These limits are the ones through which we want to control the fluxes. We next add the equivalent of the rim by extending the domain outside the free ocean. The larger the number of rim the larger is the domain. The free ocean domain remains constant.

This changes de definition of the rim in the convex corners...