Hello,
Yes, this is surprising, unless:
* You have tidal flats, in which case the PSI scheme is not guaranteed to work (use scheme 14 instead).
* You have only a partial mass-lumping of depth, and consequently a partial mass-lumping of the tracer diffusion, and this could be the problem.
* it is due to the settling velocity step (so far an upwind scheme without guaranteed positivity, we are working on several others for the next version).
With best regards,
Jean-Michel Hervouet