Noise on vertical velocities when using UBS momemtum advection scheme

[ctlod]

Context

Under few circumstances, that have to be clarified, noise on vertical velocities arises when using the UBS momentum advection scheme.
It has been observed in several configurations with various horizontal resolution (1/4°, 1/12°). This noise seems to develop
close to sharp shelves and then expands offshore along either i or/and j grid direction. The attached figure illustrates this behavior in
a regional configuration over the Arctic basin. It represents a monthly mean vertical velocity field around 700m depth.
The noise emerges clearly in the Arctic area and is less visible within the north Atlantic region.

Analysis

The origin of this noise is not yet identified and may also not be systematic.
It looks like the implicit viscosity associated to this scheme is not working or is not strong enough; yet the algorithm
itself may not be the source of the problem since 2 configurations using exactly the same module dynadv_ubs.F90
have different behavior: one with noise while the other without.
Below, a list of options that may not be responsible for this noise (deduced from an inter-comparison of 4 configurations namelist and CPP keys) :

• the non linear free surface (VVL) or linear free surface

• Partial or full steps

• ln_logyer

• ln_bfrimp

• advective and/or diffusve BBL

• GLS or TKE

• Non Penetrative Convection (NPC) or Enhenced Vertical Diffusion (EVD)

Fix

No fix for the moment.