NESTING tools: Bathymetry at bdy of AGRIF-nest is not consistent with parent bathymetry

[andrea.gierisch]

Context

I use release4.0.2/tools/NESTING (revision 12273) to create an AGRIF nest. When running NEMO4.0.2 with the resulting 1_domain_cfg.nc and domain_cfg_updated.nc, an error message says that the bathymetry at the boundary is not consistent between child and parent:

 ERROR bathymetry merge at the southern border ji,jj,jk          44           2          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          45           2          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          46           2          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          44           3          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          45           3          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          46           3          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          44           4          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          45           4          37
   6.70534153693853        34.8481484050435     
 ERROR bathymetry merge at the southern border ji,jj,jk          46           4          37
   6.70534153693853        34.8481484050435     

Analysis

Reason for the error message: e3t_0 differs for 1 coarse grid cell and the according 9 fine grid cells.
The same grid cell was mentioned in the output of agrif_create_bathy.exe saying that an isolated ocean grid cell had been removed at level 37. I suspected that something goes wrong when removing the isolated ocean grid cells in combination with updating the parent grid, and therefore I commented out the respective calls to bathymetry_control:

src/agrif_create_bathy.f90

@@ -326 +326 @@
   ! compute G0%gdepw_ps and G1%gdepw_ps
   CALL get_partial_steps(G0) 
   CALL get_partial_steps(G1)
-  CALL bathymetry_control(G0%Bathy_level)
+  ! CALL bathymetry_control(G0%Bathy_level)
   
   ! ---------------------------------------
   ! Bathymetry at the boundaries (npt_copy)                      

src/agrif_partial_steps.f90

@@ -210 +210 @@
 !!$
 !!$    ENDIF
 
-    CALL bathymetry_control(grid%bathy_level)
+    ! CALL bathymetry_control(grid%bathy_level)
     !
     ! initialization to the reference z-coordinate
     ! 

This work-around fixes the inconsistency between 1_domain_cfg.nc and domain_cfg_updated.nc. However, I don't know what bad side-effect my crude "fix" introduces. Hence, I hope that someone with more insight into NESTING can suggest a better fix.
If it helps, I can share my namelists etc. and a screenshot of the faulty e3t_0 variable.

Recommendation

Just some guesses/ideas:

• Update the parent/child bathymetry after having removed isolated ocean cells from the child/parent grid?
• Or restrict bathymetry_control to run only in the inner domain of the child and not on those grid cells along the boundary which should have the exact same bathymetry as the parent?

[andrea.gierisch]

e3t of child domain

[andrea.gierisch]

e3t of parent domain

[jchanut]

Andrea, could you provide the namelist used by the NESTING tool, please ? Jérôme

[andrea.gierisch]

Sure, Jérôme, I have attached it now.
Thanks for looking into this issue!

PS: Do you know what I need to do in order to tell this trac-system here to send me an e-mail whenever there is a new comment to my ticket? My name is added to to the "CC"-field of this form, but I nevertheless didn't receive a notification about your reply...
Edit: OK, I got a notification about my own comment now, so it seems to work in principle. Maybe the e-mail about your comment just got lost or so.