bugs in hmixing.F in the viscocity coefficient BC

[jiangyuwu]

There are couple of bugs in hmixing.F around line 509. The southern boundary condition for Uvis3d_r and Vvis3d_r uses Jend instead of Jstr like in diff3d_u:

Code: Select all

      IF (SOUTHERN_EDGE) THEN
        DO k=1,N(ng)
          DO i=IstrU-1,Iend
#  ifdef DIFF_3DCOEF
#   ifdef TS_U3ADV_SPLIT
            diff3d_u(i,Jstr-1,k)=diff3d_u(i,Jstr,k)
            diff3d_v(i,Jstr-1,k)=diff3d_v(i,Jstr,k)
#   else
            diff3d_r(i,Jstr-1,k)=diff3d_r(i,Jstr,k)
#   endif
#  endif
#  ifdef VISC_3DCOEF
#   ifdef UV_U3ADV_SPLIT
            Uvis3d_r(i,Jend+1,k)=Uvis3d_r(i,Jend,k)
            Vvis3d_r(i,Jend+1,k)=Vvis3d_r(i,Jend,k)
#   else
            visc3d_r(i,Jend+1,k)=visc3d_r(i,Jend,k)
#   endif
#  endif
          END DO
        END DO
      END IF

I also found that if the MIX_GEO_UV is defined the model will be very unstable, unless the viscocity coefficient alone open boundary is set to zero. Maybe there is a bug in uv3dmix4.h when it comes to open boundary...

[arango]

Yes, good catch. I have posted previously that this option is still experimental and there are some stability issues. It can run fine in coarse applications. In high resolution applications with steep bathymetry, it require a very small time-step or do not run at all. The model will blow-up very quickly.

I will recommned to not use this option until we figure out how to stabilize the geopotential mixing operator. The spit scheme requires a geopotential diffusion operator! Otherwise, this does not make sense. You are welcome to check the provided references for this scheme.

[ilicakme]

Hi all,

I have a quick question about UV_U3ADV_SPLIT option.
I understand that this option seems a delicate issue because of stability.
I'm trying to diagnose spurious mixing in a simple problem and this option
improves the model a lot.
So I was wondering the last situation of this option. Is it ready to use?

I just want to mention that I use UV3 for only momentum not for tracer,
I'm using TSMPDATA for tracer. And I'm using a background laplacian viscosity value.
What I understand from Hernan's last email that this option needs a geopotential diffusion operator.
Does this mean that this option needs a viscosity value?

Thanks a lot,

Mehmet