Instabilities in LF-AM3 caused by wetting/drying?

[stef]

I have a grid with 12km resolution and dcrit=0.1. Is this reasonable?

I have been looking at a specific blowup and I'm wondering if there is an instability caused by the combination of wet-dry flux-inhibitor and the multiple time levels used in the predictor-corrector scheme.


The "three-time" Adams-Moulton corrector has a negative sign for the oldest contribution, in step2d_LF_AM3.h we have:

Code: Select all

            vbar(i,j,knew)=(vbar(i,j,kstp)*                             &
     &                      (Dstp(i,j)+Dstp(i,j-1))+                    &
     &                      cff*(cff1*rhs_vbar(i,j)+                    &
     &                           cff2*rvbar(i,j,kstp)-                  &
     &                           cff3*rvbar(i,j,ptsk)))*fac

where

Code: Select all

        cff1=0.5*dtfast*5/12
        cff2=0.5*dtfast*8/12
        cff3=0.5*dtfast*1/12

Note the minus sign before cff3 in the formula for vbar. If the leading terms (vbar(kstp), rhs_vbar, rvbar(kstp)) are small or zero but cff3*rvbar(ptsk) is large, its contribution determines the sign of vbar(knew). This can cause unforced steep upward flow into a dry cell that has been previously wet and discharged water downhill (hence rvbar(ptsk) is nonzero, it was allowed to discharge when wet), but then became dry (which caused some of the other terms to have been artificially set zero by the flux delimiter)

Is this possible, and has anybody ran into this issue?

Link to time-stepping schemes:
https://www.myroms.org/wiki/Time-steppi ... B_Feedback

[jcwarner]

the wet-dry scheme is a very sensitive approach and we spent a large effort to get this into roms. It is not solving an exact equation, instead wet-dry prevent flow out of a cell if the total water depth is less than Dcrit. that is all. but it is complicated due to the multiple time levels.

In practice, it is intended for highly resolved coastal scales. we have many applications of tidal and storm flows on beached, barrier island breaching etc and it works well. I think at 12km your spacial scales are probably not resolving those wet-dry processes. what is your minimum depth in your application? at 12 km the grid cells near the coast are most likely not resolving wet-dry processes.

-j

[stef]

I think at 12km your spacial scales are probably not resolving those wet-dry processes.

Thanks for the reply, that makes sense.

I have a nested toy setup and tried to keep the processing of the grids identical as far as possible. So I have not used a minimum depth at all, even in the coarsest grids (although I did use a land mask at the boundary so that tidal forcing is not applied over land).

In the coarser grids, do you set the min. depth relatively large (say 1-2m?) and then apply land masks?

At what horizontal resolution would you say it makes sense to start wetting/drying?

[jcwarner]

i am not sure the exact horiz resolution to start using wet-dry, but i would think on the order of 100m and less, more like 10 m or less ?? really depends on the application. think of the beach, and the tide (or surge) coming in and out. what is the horizontal distance from the high tide line to the low tide line - that is the what you should think about. For some places that is 100m, for others it is 5m. so it depends on the application.

[stef]

think of the beach, and the tide (or surge) coming in and out. what is the horizontal distance from the high tide line to the low tide line - that is the what you should think about.

Good point, thanks again for your help!