Problem in allocation of t_west_C*

[mathieu]

The problem is the following in mod_boundary.F:

• t_west_C2 is declared as

     pointer :: t_west_C2(:,:,:)
  

• but it is allocated as

     t_west_C2(ETA_DIM,N(ng))
  

i.e. with dimension 2 instead of 3. Same problem for other variables.

[arango]

Yes. However, notice that radiation boundary condition are very problematic in adjoint solutions. The solution is even more sensitive in the tangent linear and adjoint models than in the nonlinear model.

We use a different strategy in adjoint-based applications with open boundary conditions. We have been successful with the clamped boundary conditions for u, v, T, and S and Flather boundary conditions for ubar, vbar, and zeta. This requires open boundary condition data from other models or climatologies. The adjoint of these type of boundary conditions is trivial and well behaved.