How can I decide the most appropriate parameter

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply
Message
Author
jerry19890126
Posts: 13
Joined: Wed Jul 27, 2011 3:35 am
Location: IOCAS

How can I decide the most appropriate parameter

#1 Unread post by jerry19890126 »

For instance, the horizontal diffusion/viscosity coefficient .Although the website https://www.myroms.org/wiki/index.php/Horizontal_Mixing has offered a guideline for coefficient values, the range seems to be too wide.The set up of the coefficient may play a significant role in the quality of the modelling outcome.As I'm about to run a long-time simulation, to set the most appropriate parameter seems to be crucially important.I'm a beginner, can anyone be kind enough to give me some tips?

Another confusion:
In the ocean.in file , I encoutered an option:
VolCons(west) == T
to activate the west boundary volume conservation.Is that necessary for the model?I really appreciate any help. :)

rduran
Posts: 152
Joined: Fri Jan 08, 2010 7:22 pm
Location: Theiss Research

Re: How can I decide the most appropriate parameter

#2 Unread post by rduran »

Anything with a T for true can have an F for false i.e. your choice.

There is no way (that I know of) around sensitivity tests when starting a new application, you should not expect to get it right the first time because you probably wont. You may want to design an experiment to test the parameters you may think are important as well as possibly discover those that you may have missed.

If somebody has some advice on how to narrow down the range of viscosity coefficients I would also like to know.

lanerolle
Posts: 157
Joined: Mon Apr 28, 2003 5:12 pm
Location: NOAA

Re: How can I decide the most appropriate parameter

#3 Unread post by lanerolle »

There are some rough rules of thumb that you can use:

1. What is the minimum diffusivity/viscosity value you need to use to maintain T, S > 0 (assuming that you do not have negative temperatures)? So you can try some values which give T, S > 0 and then pick the minimum,

2. You can calculate the non-dimensional parameter mu=K*dt/(dx*dx) [in 1-D] or mu=K*dt*(1/dx*dx + 1/dy*dy) over your grid and see what its maximum value is. Ideally, it should be quite small - something in the range of 10^-6 - 10^-3 or else, your T, S fields would be totally smeared out in time and,

3. You may also want to use the grid-scaled diffusivity/viscosity option in ROMS so that all grid cells "see" the same mu value - otherwise different cells will see different mu values and experience different levels of spatial smoothing.

laanderson
Posts: 2
Joined: Thu Nov 15, 2012 4:11 pm
Location: Woods Hole Oceanographic Institution

Re: How can I decide the most appropriate parameter

#4 Unread post by laanderson »

Field experiments indicate that horizontal diffusivities decrease with length-scale (Okubo, 1971, Deep-Sea Research, 18:789-802). So the horizontal diffusivity you should use depends on your horizontal grid resolution. Ledwell et al. (1998, Journal of Geophysical Research 103(C10):21499-21529) suggests an isopycnal (~horizontal) "diffusivity of 2 m^2/s at scales of 1 to 10 km" and "on the order of 1000 m^2/s, acting at scales of 30 to 300 km" i.e. if mesoscale eddies are not resolved.

Traditionally, horizontal viscosities are set at 1 to 10 times the horizontal diffusivity. I am not sure if there is a justification for that. Perhaps higher values just seem too viscous (relative to the density gradients), and lower values might allow certain types of instabilities.

One school of thought is to turn the horizontal diffusivity and viscosity down as low as possible, to resolve as much submesoscale structure as possible. I am not sure if that is the correct thing to do. In nature, there is always subgridscale turbulent advection, and the ocean is constantly agitated by the winds, remotely-generated waves, tides and subgridscale rough bathymetry. I think it is more justifiable to start from the observation-based estimates---hopefully from your region and for your gridscale if you can find them. Then, if your jets are too wide and slow (compared to observations), that is justification for lowering the diffusivity and viscosity further; if the jets are too thin and unstable (compared to observed), raise them. But staying within the numerical limits (CFL condition, etc.) Also, it makes more physical sense to adjust your timestep to fit your diffusivity, than to adjust your diffusivity to fit your timestep!

I hope this helps.

Larry

jerry19890126
Posts: 13
Joined: Wed Jul 27, 2011 3:35 am
Location: IOCAS

Re: How can I decide the most appropriate parameter

#5 Unread post by jerry19890126 »

Many thanks everyone.Your suggestions can be of great help to my study.
However, the 'VolCons Option' problem is perplexing me now.The thing is, I added the tides in my model, and I set the Volume Conservation (known as VolCons) option to F as the ocean.in file suggests that 'Usually activated with radiation boundary conditions to enforce global mass conservation, except if tidal forcing enabled.' The model blows up immediately after started, it seems that it encountered some NaN values. When I turned the option to T , the model runs smoothly.
So the option has to be left to be 'T' regardless of the fitness. Am I missing something?

tony1230
Posts: 87
Joined: Wed Mar 31, 2010 3:29 pm
Location: SKLEC,ECNU,Shanghai,China

Re: How can I decide the most appropriate parameter

#6 Unread post by tony1230 »

Two possible suggestions.
1> you may want to double-check your input files like initial,forcing,etc. to refirm that no NaN exists.
2> turn Debug on as expect to get more directory info.

generally speaking, if model blows up right at spin-up, especially with NaN, you may got it in one/some
fields.

Regards

- Shou

lanerolle
Posts: 157
Joined: Mon Apr 28, 2003 5:12 pm
Location: NOAA

Re: How can I decide the most appropriate parameter

#7 Unread post by lanerolle »

Yes, if you have tides - especially with sub-tidal effects also included, its best not to use the VOLCONS option. It applies mainly in cases which include model coupling/nesting where you want the full flow (and hence also the temperature and salinity and any biological fluxes) from the outer domain to come in to the inner/coupled/nested domain. But then again, in the nested domain, you may have additional rivers say and this will also modify the volume balance within it.

rduran
Posts: 152
Joined: Fri Jan 08, 2010 7:22 pm
Location: Theiss Research

Re: How can I decide the most appropriate parameter

#8 Unread post by rduran »

nan in input files or the need to reduce the time step are probable culprits

Post Reply