[Mitgcm-support] GMREDI updated

[mitgcm-support at dev.mitgcm.org]

Hello Stephanie,

Sorry for this "#define GM_NON_UNITY_DIAGONAL" that was not right.

I fixed few bugs and changed few other things in the GMREDI package
(in fact, all the sources have been modified), + themodynamics.F 
I checked the adiabatic property of the isopycnal diffusion only 
(Redi tensor) in a 2.D test case, so that it should be OK now.

Concerning the GM_NON_UNITY_DIAGONAL part, there is a new parameter 
"GM_Kmin_horiz" that can be of interest to you. It sets the mimimum 
value of the horizontal diffusion where the tapering effect is strong:

GM_Kmin_horiz=GM_background_K   ==> GM_NON_UNITY_DIAGONAL has no effect,
  and then you should get the same results as before.

GM_Kmin_horiz=0. (the default)  ==> the horizontal diffusion can 
  be very small where the tapering is strong. For stability reasons, 
  a small value might be better than 0.

I put a value of 50 m^2/s. in global_ocean.90x40x15/input/data.gmredi
but Alistair says that MOM is currently used with a minimum 
value of 100 m^2/s in a similar configuration. 

Other thing: All the tensor arrays scale now like diffusivity,
and consequently, all the GM output files changed (don't need any 
more to *GM_background_K). I changed this to get real mean-time 
diagnostic when I use the Visbeck-variable-K option.

See you.

Jean-Michel