[Mitgcm-support] Re: initial w and cyclic bc's

[mitgcm-support at dev.mitgcm.org]

The _EXCH_XYZ_R8(uVel,myThid) method of exchange has been replaced
with CALL EXCH_UV_XYZ_RL(uVel,vVel,.TRUE.,myThid). This call is lower
down in the routine:

c     IF (uVelInitFile .NE. ' ' .OR. vVelInitFile .NE. ' ') THEN
        CALL EXCH_UV_XYZ_RL(uVel,vVel,.TRUE.,myThid)
       _EXCH_XYZ_R8(wVel,myThid)
c     ENDIF

but indeed it is in the wrong place. The exchange block should be
before the call to integrate W.

I've fixed this in code. You can see the latest routine at:
http://mitgcm.org/cgi-bin/cvsweb.cgi/models/MITgcmUV/model/src/ini_vel.F?only_with_tag=MAIN

Good spot. Thanks,

Alistair.

Tor Eldevik wrote:
> Why has the line that updates the overlaps been
> commented out in model/src/ini_vel.F?
> 
> C     Read an initial state for each component if required
>       IF (uVelInitFile .NE. ' ') THEN
>        _BEGIN_MASTER( myThid )
>        CALL READ_FLD_XYZ_RL( uVelInitFile, ' ', uVel, 0, myThid )
>        _END_MASTER(myThid)
> c      _EXCH_XYZ_R8(uVel  , myThid )
>       ENDIF
> 
> The above implies that any cyclic b.c.'s in u,v are
> not accounted for prior to initializing w. Accordingly
> large initial vertical velocities are set up at the
> eastern and northern boundaries.