Bugs with adaptive vertical advection: 2) Interaction with time splitting

[jchanut]

Context

Adaptive (based on courant number) vertical advection, ln_zad_Aimp=.true.

Analysis

With split explicit free surface, vertical velocity is refreshed after completion of the barotropic loop. A new split between implicit and explicit vertical velocities is computed and the "new" implicit vertical velocities are then used to complete momentum vertical advection in dynzdf. It seems to me that one should rather use the initial linear split of vertical velocities.

Recommendation

In step.F90, instead of:

      IF( ln_dynspg_ts ) THEN                         ! vertical scale factors and vertical velocity need to be updated
                            CALL div_hor    ( kstp )              ! Horizontal divergence  (2nd call in time-split case)
         IF(.NOT.ln_linssh) CALL dom_vvl_sf_nxt( kstp, kcall=2 )  ! after vertical scale factors (update depth average component)
                            CALL wzv        ( kstp )              ! now cross-level velocity 
         IF( ln_zad_Aimp )  CALL wAimp      ( kstp )  ! Adaptive-implicit vertical advection partitioning
      ENDIF
                         CALL dyn_zdf       ( kstp )  ! vertical diffusion

rather do:

      IF( ln_dynspg_ts ) THEN                         ! vertical scale factors and vertical velocity need to be updated
                            CALL div_hor    ( kstp )              ! Horizontal divergence  (2nd call in time-split case)
         IF(.NOT.ln_linssh) CALL dom_vvl_sf_nxt( kstp, kcall=2 )  ! after vertical scale factors (update depth average component)
      ENDIF
                         CALL dyn_zdf       ( kstp )  ! vertical diffusion

      IF( ln_dynspg_ts ) THEN                         
                            CALL wzv        ( kstp )              ! now cross-level velocity 
         IF( ln_zad_Aimp )  CALL wAimp      ( kstp )  ! Adaptive-implicit vertical advection partitioning
      ENDIF

[jchanut]

In 11287:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[jchanut]

In 11288:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found