Custom Query (964 matches)

The issue of nudging free-surface, as an assimilation option NUDGING_SSH, has come-up several times in the ROMS forum. This option is very old and part of the legacy code. I am cleaning this option because it doesn't make sense within ROMS kernel. I tried this many years ago and it was not a good idea because it affected the volume and mass conservation. The nudging scales are very difficult since we need to constraint the free-surface at every barotropic time-step. If you change zeta during nudging, you need to change ubar and vbar accordingly. Then, the very direct (and best) problem is to constrain (nudge) only ubar and vbar.

However, the assimilation of SSH (altimetry) is possible in ROMS via OI or 4D-Var. The OI is intermittent and not part of the time-step and forces a cold start. This allows such assimilation to be consistent with ROMS governing equations and its discretization. In 4D-Var everything is consistent with the adjoint-based increment (adjustment). Notice that in 4D-Var we cannot assimilate ubar and vbar. Everything has to be done in terms of u and v. Then, the ubar and vbar are computed by the vertical integral of the estimated u an v. This implies that we need to be very careful with the assimilation strategy and ROMS kernel.

In ROMS, the free-surface (zeta) is is an statement of the vertical integrated continuity equation, which is a function of the vertically integrated velocity (ubar, vbar) divergence. That is, ubar and vbar are the real prognostic variables in ROMS. The free-surface zeta is kind of a diagnostic quantity since it is computed from the vertically integrated continuity equation. Any change in zeta at every time-step needs to be consistent with changes in ubar and vbar.

Corrected a type in inp_par.F around line 1660 when T_PASSIVE is activated. We need to have instead:

             DO ng=1,Ngrids
               DO itrc=1,NPT
                 i=inert(itrc)
                 LtracerSrc(i,ng)=Ltracer(i,ng)
               END DO
             END DO

Also corrected a typo in ana_psource.F around line 267, we need to have instead:

!$OMP BARRIER

Many thanks to Alessandro Coluccelli for ​reporting these typos.

Corrected a couple of problems when calling routines get_state and netcdf_close:

• The get_state routine has the following arguments:

        SUBROUTINE get_state (ng, model, msg, ncname, IniRec, Tindex)
  

  The IniRec argument is declared as intent(inout). This implies that it is illegal to pass a numeric value when calling this routine. We need to use a variable instead in several drivers. Many thanks to Brain Powell for bringing this to my attention. Some compiler may ignore this roule.

• The routine netcdf_close has the following arguments:

      SUBROUTINE netcdf_close (ng, model, ncid, ncname, Lupdate)
  

  The ncname and Lupdate arguments are optional and only used when closing output files to update the bio_file global attribute in ecosystem model applications. This attribute will contain information about the path of the ecosystem model header file used during compilation. Recall that users can have a customized version of this file in their application directory. We need to be sure that Lupdate=.FALSE. when closing input files open for reading only. Otherwise, an error will occur. There were few files that needed this optional argument. Many thanks to Julia Levin for reporting this problem.