MOM_interface PR #190 Broke Dimensional Consistency test

[alperaltuntas]

PR #190 Broke Dimensional Consistency test for the T (time) dimension. All other dimensions are still passing.

[alperaltuntas]

A bit more info: The culprit appears to be FULL_DEPTH_KHTR_MIN. Setting it to False fixes the failing dimensional consistency test T. Looking at the checksums of a scaled run with FULL_DEPTH_KHTR_MIN=True, the first diff appears in MOM_diabatic driver, after KPP, when compared to a non-scaled run with FULL_DEPTH_KHTR_MIN=True.

[iangrooms]

On these lines the CS%Khtr_min is compared against Coef_x and Coef_y, but I'm pretty sure that the dimensions of Coef_x and Coef_y do not match the dimensions of CS%Khtr_min.

Looking just a few lines up

Coef_y(i,J,K) = I_numitts * khdt_y(i,J)
Coef_x(I,j,K) = I_numitts * khdt_x(I,j)

I_numitts is nondimensional while khdt_{x,y} are, as the name suggests, a diffusivity times the time step. In light of this, I think the following might fix the dimensional consistency problem:

Coef_y(i,J,K) = max(Coef_y(i,J,K), dt * CS%KhTr_min)
Coef_x(I,j,K) = max(Coef_x(I,j,K), dt * CS%KhTr_min)

This would change the results substantially. I haven't tested it.