Hi,
I would like to report a problem with ROMS 3rd order "horizontal" advection scheme for tracer advection. This scheme is a very nice compromise between quality (no dispersion, quasi-monotonic) and CPU cost (see Shchepetkin, 1998). However, when applied to sigma coordinate models the issue of the orientation of implicit diffusion contained in this scheme is posed. In particular a certain amount of diapycnal diffusion can be expected. Indeed, we have recently noticed in regional ROMS applications to the south-west Pacific that a substantial amount of diapycnal diffusion occurs, especially at depth (where bottom smoothing is not severely required according to dh/h smoothing parameter). Our AAIW was loosing its low salinity characteristics. Replacing the 3rd scheme by the 4th order centered scheme with geopotential explicit diffusion resolves the problem.
Our surprise is that the 3rd order scheme has only small diffusion, but obviously it can be enough (at least at the range of resolution that we have used: 1 to 1/12 degrees) At this point, I wonder if ROMS users and developers have noticed a similar problem, particularly in large regional domain applications . We could then share remedies ...
Patrick