how to choose horizontal diffusion coeficients ?

Discussion of how to use ROMS on different regional and basin scale applications.

Moderators: arango, robertson

Post Reply
Message
Author
cecilia
Posts: 41
Joined: Thu Aug 16, 2007 9:10 pm
Location: DGF - UChile
Contact:

how to choose horizontal diffusion coeficients ?

#1 Unread post by cecilia »

What criterions I could use in ROMS to choose horizontal diffusion coeficients of momentum and tracers (temperature and salunity) for different grid spatial steps (spatial resolutions) ?

Thanks

Cecilia

User avatar
kate
Posts: 4088
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

#2 Unread post by kate »

For this (and the timestep question), it's often a matter of try it and see if it runs. If you use the 3rd order upwind advection, you can possibly get away with no explicit horizontal smoothing at all, especially on tracers. What are your criteria for success? Fields looking smooth in the mean? Instantaneous vertical velocities looking smooth at mid-depth? Tides of the right amplitude? Those could give you three quite different "best" values.

cecilia
Posts: 41
Joined: Thu Aug 16, 2007 9:10 pm
Location: DGF - UChile
Contact:

#3 Unread post by cecilia »

Kate,

But there is no more objective criterions ? as the Courant-Friedrich-Lewis number in the princeton Ocean Model (POM) for the barotropic mode ? and spatial scales in function of dx and timescale (here I suppose the time scale should some characteristic for the subgrid process) for harmonic and biharmonic diffusion ? Here I do not manage information, then I am looking for the necessary to choose the horizontal diffusion coefficients trying to get velocity and tracer fields with physical meaning.

When I run POM I try to mantain a high resolution to solve upwelling with a physical meaning, following the example of Allen and Nerbwerger (1995). But now I will have to use a coarser grid, and I want to do the same..... (if that is possible) ... damping the subgrid scales and do not damping the physical modes of scales greater than 2*dx.

Bye and thanks

User avatar
kate
Posts: 4088
Joined: Wed Jul 02, 2003 5:29 pm
Location: CFOS/UAF, USA

#4 Unread post by kate »

ROMS, like POM, uses a split explicit timestepping scheme. The exact same ballpark estimates will apply. The details of the timestepping differ between POM and all the ROMS branches, plus your options, so the exact timestep you can get away with is not something I try to predict - I just do a bit of shooting for it. You can always get hit with a nasty shock like adding tides and suddenly needing a three times smaller baroclinic timestep.

For the horizontal smoothing, we try to keep it as small as possible to give a smooth enough solution. ROMS is known to have very small implicit smoothing (other than in that upwind advection scheme), allowing for less damping of your primary signals than in some other models. Try the upwind scheme and see if it provides enough implicit smoothing. Other than that, compute the damping timescale from your viscosity on 2-dx noise and on 10-dx signals. Are they acceptable?

Post Reply