Opened 17 years ago
Closed 17 years ago
#71 closed bug (Fixed)
Variables t_east_C*
| Reported by: | mathieu | Owned by: | arango |
|---|---|---|---|
| Priority: | major | Milestone: | Adjoint Based Algorithms |
| Component: | Adjoint | Version: | 3.1 |
| Keywords: | Cc: |
Description
In mod_boundary.F there is are the variables t_east_C2, t_east_Cx and t_east_Ce.
But in
- tl_t3dbc_im.F,
- rp_t3dbc_im.F and
- ad_t3dbc_im.F
the variable t_east_C is used, while undeclared.
The solution is maybe to replace it by t_east_C2.
The nonlinear t3dbc_im.F seems to be correct in this respect
Note:
See TracTickets
for help on using tickets.
Yes, good catch. We haven't used this strategy in long time. Radiation conditions in the tangent linear and adjoint model are very tricky since the basic state needs to be saved at each barotropic and baroclinic time-step. In addition, there is not guarantee of stable solution. These models are more sensitive than the nonlinear model to the open boundary conditions. And we all know how troublesome are radiations conditions in the nonlinear model...
As I mentioned in https://www.myroms.org/projects/src/ticket/67 is better to use open boundary conditions from other models (clamped conditions). This is working very well in our regional adjoint applications.