KE-conserving correction to velocity remap

iangrooms commented 2 months ago

This PR does two things:

Adds an option to correct the remapped velocity so that the KE is conserved. This can be used in combination with any remapping scheme. It is not a new remapping scheme, it is a correction to the remapped velocity.
Adds two diagnostics, ale_u2 and ale_v2, that measure the rate of depth-integrated KE loss (or gain) associated with remapping velocity. The rates are computed separately for the u and v components, and the output units are Watts per square meter.

The correction is described in more detail here ke_conserving_remap.pdf

iangrooms commented 2 months ago

Thanks @Hallberg-NOAA for your review! I am still working on mastering the unit scalings and Boussinesq vs non aspects of MOM6. I think my latest commit addresses all your comments, but please take a look.

Hallberg-NOAA commented 2 months ago

With these changes that you just introduced, it is all looking good to me!

gustavo-marques commented 1 month ago

Thanks, @iangrooms, for this contribution, and thanks, @Hallberg-NOAA, for reviewing this PR.

The pr_mom tests are passing. I only have the minor comment raised above regarding a new logical parameter (allow_preserve_variance) in subroutine ALE_remap_velocities.

New MOM_input parameter:

REMAP_VEL_CONSERVE_KE = False !   [Boolean] default = False
                                 ! If true, a correction is applied to the baroclinic component of velocity after
                                 ! remapping so that total KE is conserved. KE may not be conserved when
                                 ! (CS%BBL_h_vel_mask > 0.0) .and. (CS%h_vel_mask > 0.0)

New available diagnostics:

 "ale_u2"  [Unused]
     ! modules: {ocean_model,ocean_model_d2}
     ! long_name: Rate of change in half rho0 times depth integral of squared zonal velocity by remapping. If REMAP_VEL_CONSERVE_KE is .true. then this measures the change before the KE-conserving correction is applied.
     ! units: W m-2
     ! cell_methods: xq:point yh:mean

 "ale_v2"  [Unused]
     ! modules: {ocean_model,ocean_model_d2}
     ! long_name: Rate of change in half rho0 times depth integral of squared meridional velocity by remapping. If REMAP_VEL_CONSERVE_KE is .true. then this measures the change before the KE-conserving correction is applied.
     ! units: W m-2
     ! cell_methods: xh:mean yq:point

iangrooms commented 1 month ago

When I was first implementing this I noticed that the subroutine ALE_remap_velocities gets called in more than one context; e.g. this subroutine is applied to remap diffusive tendencies. If we just control the action with a MOM_input parameter then the correction either happens every time the subroutine is called or not at all, and we don't necessarily want to apply this KE-conserving correction every time the subroutine is called - only when remapping the velocity during the time stepping loop.

At the time I discussed this with Alistair, who suggested that I add the logical parameter ALE_remap_velocities. I then hard-code the value to 'true' but only in the places where I want the correction to be allowed. In other places it is 'false' by default. (Alistair also helped me significantly by providing code snippets for various parts of the algorithm.)

In short, we have 2 logical parameters controlling this feature. Only one parameter can be set by the user, and it turns the feature on or off. The other parameter is only internal, and serves to ensure that the correction only happens in the places we want it to happen.

gustavo-marques commented 1 month ago

Thanks for the clarification!

NCAR / MOM6

KE-conserving correction to velocity remap #277