Context

Wrong vertical velocity at subdomain boundaries with ln_bt_fw = .false.

Analysis

There is an issue with vertical velocities when using ln_bt_fw = .False. in the trunk at the subdomain boundaries. Indeed, on the figure ​https://nemo-ocean.zulipchat.com/#narrow/stream/292415-validation-4.2E2.20RC/topic/wrong.20W.20at.20boundaries.20with.20ln_bt_fw.3Dfalse (daily average of the vertical velocity with AMM12 and ln_bt_fw = .False.) we can cleary distinguish the subdomain boundaries on W fields.

This issue is not present when ln_bt_fw = .true. . Note also that the problem is also not present in an older trunk revision (r14365) .

Fix

This feature is also distinguishable on the horizontal divergence hdiv so my guess is that it may be due to wrong scale factors or velocities in halos cells when using ln_bt_fw = .False.

In dynspg_ts, vector optimisation gets wrong lin 259

from Julien Jouanno:
sn_cnf should be sn_frld in namelist_ice_lim2