Opened 6 months ago

Last modified 6 months ago

#2344 new Defect

issues with 3D momentum trends diagnostics and split-explicit time-stepping

Reported by: davestorkey Owned by: davestorkey
Priority: low Milestone: 2020 WP
Component: TRD Version: trunk
Severity: minor Keywords: trends, diagnostics,
Cc:

Description

Context

There are various issues with the 3D momentum trends diagnostics (ln_dyn_trd=T) when using the split-explicit solution of the momentum equation (dynspg_ts - default from NEMO 4.0 onwards):

  1. The "SPG" trend is not the trend due to the surface pressure gradient. It is the change in the velocities due to the solution of the barotropic part of the equations in dyn_spg_ts.
  1. More generally, for trends which are allowed to vary over the barotropic subcycling (PVO, BFR), there is a small inaccuracy because currently this variation is not taken into account.
  1. At various points in the code the barotropic part of the velocities is subtracted or added back in (eg. at the beginning of dyn_spg_ts, or at the beginning of dyn_zdf for the implicit bottom friction option). This subtraction/addition of the barotropic component is sometimes incorrectly included in the trend associated with that module.

These problems mean that it is currently not possible to close the momentum budget, as already noted in tickets #1584 and #1984. The 3D momentum trends are used to calculate various other diagnostics such as 2D vorticity trends (ln_vor_trd=T) or kinetic energy trends (ln_KE_trd=T) so the problems also affect these derived diagnostics.

Analysis


Recommendation

  1. Momentum trends diagnostics should correspond to terms in the primitive equations (rather than NEMO modules).
  1. We should have a small number of 3D trends which are sufficient to close the budget:
    1. Each 3D trend should include the barotropic and baroclinic part. (We can output the barotropic part separately if we want). Where the barotropic part of the trend varies over the subcycling this part can be calculated accurately using the time-averaging weights already available in dynspg_ts.F90.
    2. The horizontal pressure gradient trend (HPG) should include the surface pressure gradient (SPG). The SPG trend can be output as an extra diagnostics not required to close the budget.
    3. The vertical diffusion trend (ZDF) already includes the contribution of the wind stress, top stress (in ice shelf cavities) and implicit bottom stress, but it should also include the contribution from the explicit bottom stress terms. The explicit bottom stress can still be output as an extra diagnostic not required to close the budget.
  1. We should make sure that the subtraction or addition of the barotropic component of velocities or velocity trends is not included in the momentum trend diagnostics. This particularly applies to the ZDF trend. (This is similar to a fix proposed but not implemented by Christoph in ticket #1584).



Commit History (0)

(No commits)

Change History (1)

comment:1 Changed 6 months ago by jchanut

And see ticket #1723 also. Seems to be a hot topic ;)

Note: See TracTickets for help on using tickets.