[MITgcm-support] Open boundary conditions for tracers

[Jean-Michel Campin]

Hi Martin,

Thanks for your suggestion relative to ptracers & obcs.

I was thinking that in a first attempt to make ptracers work with obc,
with just 1 call to obcs_apply_ptracer after CALL PTRACERS_IMPLDIFF
would do the job.

> 2) call obcs_apply_ptracer at the right places (where obcs_apply_ts or
> obcs_apply_tloc are called). These places are in
> a. pkg/generic_advdiff/gad_advection.F, TWO times, just add a line such
> as
> >          ELSEIF (tracerIdentity.GE.GAD_TR1) THEN
> >           CALL OBCS_APPLY_PTR( bi, bj, k, localTij, myThid )
> b. model/src/thermodynamics.F (after timestep_tracer and impldiff).
> 

I would not worry too much about gad_advection.F, since most (if not all)
of the advection schemes that use multidim advection need a wide stencil
(5 points) and I am not sure that OBCS work well in this case.

And in thermodynamics.F, I don't think 2 calls are necessary (but may be 
I am wrong): we could remove the 1rst call to OBCS_APPLY_TS (the one that is in 
the reverse order k loop) and always call OBCS_APPLY_TS after CALL IMPLDIFF(T,S).

This would make the code simpler, and the k loop could go into obcs_apply_ts.F
(and obcs_apply_ptracer).

Jean-Michel