OBCS tidal forcing and Orlanski interfere

[jessecusack]

I want to force barotropic tidal currents into a domain using obcs. I have been testing out pull request #752 (thank you @jm-c!), as well as the old tidal forcing with only normal velocity components, on a simple model. In either case the currents being forced at the boundary (as diagnosed from the model output) bare little resemblance to the input file specifications when useOrlanskiXXXX (XXXX = East, West etc.) is TRUE. The amplitude of the currents in the model is an order of magnitude too large, the phase is incorrect, and there is significant mean component despite specifying no other boundary forcing. For example:

Input:

Output:

However, with useOrlanskiXXXX set to FALSE, I get the correct output. If there were topography in the domain I might want to use Orlanski to allow radiating waves to exit, so this is a problem....

I created a simple model that reproduces the problem on my system and runs rapidly: obcs_tides.zip. Note that is uses the unmerged #752 version of MITgcm. I would be happy for this example to be added as a verification experiment in the future of obcs tides if you think it could be useful.

[jm-c]

@jessecusack I took a quick look at the code, it does not look like there is an obvious coding problem in Orlanski + tides: the barotropic tides are added to whatever Orlanski predict.
But it's possible that adding (barotropic) tital velocity to the OB causes the Orlanski scheme to predict the wrong OB flow (since it does not filter out the tidal component that are present in the solution when trying to derive the OB).
I don't have much experience with Orlanski but I vaguely remember some issues about the the barotropic component that this scheme produces.

[jklymak]

@jessecusack if you want all internal waves to radiate, I suggest using a sponge, either with obcs or rbcs. If you check, I think you will see that orlanski is not great at preventing reflection. I usually refer to Carter and Merrifield 2007 where they discuss this (though not for MITgcm). Of course what bc you use is somewhat problem dependant.

[jessecusack]

@jm-c in my quest for a solution I read a few exchanges on the MITgcm mailing list from many years back that perhaps hinted at similar problems.

@jklymak thank you for pointing me that reference. I had reached the same conclusion as you to use a sponge.

[jm-c]

@jessecusack Thanks for your contribution and explanations regarding this issue. I think this issue now could be closed,
unless someone would like to add something here.

[jessecusack]

@jm-c sounds good to me.