Salinity drifting when there is no salinity forcing

[fabien-roquet]

When there is no salinity forcing, salinity is still drifting.
We believe it happens with the implicit diffusion scheme which produces noise which accumulates over time, but we are not sure exactly why. Because salinity is used in the equation of state, that creates disturbances on the dynamics that will increase over time.
It would be best to enforce/control salinity conservation. Some similar bug might be affecting temperature as well.

[dionhaefner]

Can you share more details about your setup? How can I reproduce this?

[Titouan-Moulin]

neverworld2.zip
We use this neverworld2 setup inspired from the acc setup and we ran it for 1d before plotting a salinity profile

ds_snap = xr.open_dataset(neverworld2.snapshot.nc)
SA = (ds_snap["salt"].isel(Time=0)-35)
SA.isel(xt=30).plot()

[dionhaefner]

Some numerical noise is certainly expected. Your reproducer outputs salinity fluctuations on the order of 10^-14 which is well below the noise threshold and should not lead to significant model drift.