How to determine the value of dtfast ?

[langlangxie]

Hi
I know dt=Ndtfast*dtfast,and dafast depends on the maximum depth and grid resolution.what's more,dtfast is limited by the barotropic CFL criterion.
So ,I want to know how to determine dafast.
Can you help me !

[kate]

You're asking how to pick ndtfast? I usually have it at 20 since my baroclinic timesteps are short enough for this to be stable. I could probably get away with smaller values, but haven't really pushed it.

[langlangxie]

You're asking how to pick ndtfast? I usually have it at 20 since my baroclinic timesteps are short enough for this to be stable. I could probably get away with smaller values, but haven't really pushed it.

Yes,I am asking help about how to pick Ndtfast.Kate,you choose ndtfast=20 in your every case? This is a universally right? Saying in other words,I can make my ndtfast=20? I want to understand how to compute my ndtfast!
Thanks a lot!

[kate]

I recently heard a quote that goes something like "I used to worry about that, but now I don't".

I have also tried ndtfast=30, but when I need to shorten my main dt for stability, sometimes I reduce ndtfast. Very rarely (never?) is my model blowing up because my dtfast is too long - it finds many baroclinic ways to blow up instead. So pick something and if the model runs, ignore it.

[langlangxie]

I recently heard a quote that goes something like "I used to worry about that, but now I don't".

I have also tried ndtfast=30, but when I need to shorten my main dt for stability, sometimes I reduce ndtfast. Very rarely (never?) is my model blowing up because my dtfast is too long - it finds many baroclinic ways to blow up instead. So pick something and if the model runs, ignore it.

Thanks a lot! Kate
I will try it ,when I make ndtfast=20.In my word,my most important purpose is to keep my ROMS stably.Then I choose to change the Ndtfast,is it?