How to generate refined grids?

[jihene]

I'm currently using a two way nesting in ROMS for modeling internal tide. I have used the coarse2fine.m scripts to generate the child grids and contact.m to generate the contact file. The coarse grid is 1/30° of resolution and the refined grids are 1/90°. When I run the model I found that there are almost not change in the barotropic -baroclinic energy conversion. But an enhanced conversion occurs when I run the model without nesting with 1/60° resolution. I think that the problem is with the use of the matlab script (coarse2fine.m) since we just interpolate the coarse to the refined grid and so we can't resolve bathymetric details which may lead to an underestimation of conversion.
So,I tried to build nested grids by generating an intermediate grid at the finest resolution (1/60)° and from it I have extracted the coarse grid (1/20°) (using fine2coarse.m) and the refined grids (by using the script grid_extract.m). After I tried to relate the nested grid to the coarse grid (parent) by finding indexes of the corners for each refined grid and I added these values in the global attributes of each refined grid. The contact file is successfully generated and all plots generated by contact.m script are ok, but unfortunately when I run the model it blows up very early!! So, I am wondering if my method of such building is correct!! Is there another way to build properly refined grids?
Thank you very much for your interest and help,
Jihene

[nmolem]

Your fine and coarse grid topography may not match at the boundaries of the fine grid. If that is the case, you should use a merging function to ensure that they do. See also Mason et al., Ocean Modelling, 2010.

[jihene]

Thanks, I will check the topography at the boundaries.

[jihene]

In fact there are significant depth mismatches between child and parent grids along the open boundaries. I have used the Pierrick Penven matlab script ( Thanks Pierrick !) in order to connect topography between child and parent grids and the model runs well right to now!
Thanks nmolem for your help!