Custom Query (964 matches)
Results (232 - 234 of 964)
| Ticket | Owner | Reporter | Resolution | Summary |
|---|---|---|---|---|
| #110 | Fixed | 4DVAR observation processing, shallow water model | ||
| Description |
The processing of the observation in the ad_main2d.F needs to be processed outside of the delayed IO time-step: !
! Avoid time-stepping if additional delayed IO time-step.
!
IF (iic(ng).ne.ntstart(ng)) THEN
...
to allow the correct processing of observation at initialization time. This only applies to ROMS shallow water model configuration. |
|||
| #111 | Fixed | Adjoint radiation boundary conditions | ||
| Description |
The radiation conditions are not trivial in adjoint codes since the radiation celerities for the basic (forward) state need to be saved for every 2D and 3D time-steps. In addition, it is tricky to keep the tangent linear and adjoint model stable under such boundary conditions. Therefore, we have ignored this type of boundary conditions in all adjoint-based algorithms. Since we are not satisfied with this type of boundary conditions in the nonlinear code, we have not pay attention to this option in the TLM, RPM, and ADM codes. Some of the infrastructure for these options has been coded but not tested. We keep getting inquires about this options. We will look at this in the future. Anyway, there are few bugs associated with adjoint radiation conditions. I am correcting some of these codes but still this option is not available! You need to use a different strategy for open boundary conditions in the adjoint-based algorithms. Many thanks to all of you who inquired about this. |
|||
| #112 | Done | Parabolic splines vertical diffusion convolution, 4DVAR | ||
| Description |
I added a new option SPLINES_VCONV to 4DVAR background/model error covariance convolutions using the parabolic splines (implicit algorithm) reconstruction of the diffusion operator. Therefore, there are now three algorithm for vertical convolution (VCONVOLUTION):
1) Explicit |
|||
