Wind modeling

[artem_vb]

How the ROMS can be adjusted to take into account wind stress? I do not need to model the air flows, so I do not need AIR_OCEAN switched on, right? I'm also quite happy to specify surface winds analytically, ao ANA_WIND should be on I presume. What else should be adjusted to take wind effects into account?

Thank you for any help.

[arango]

You can specify wind stress in ROMS with analytical expressions in analytical.F using the CPP option ANA_SMFLUX or read wind stress (gridded or point data) from forcing NetCDF file (if not analytical options are specified). Alternatively, you need to provide all the relevant atmospheric fields (analytical expressions or data) to compute the stress and heat fluxes using the air-sea interface boundary layer parameterizations. This is activated with the BULK_FLUXES option.

There is a lot of information in the forum and in ROMS web site about forcing fields.

The CPP option AIR_OCEAN is more complicated and is for full coupled atmosphere and ocean models. It is only used in distributed memory applications where the MPI communicator is split between atmosphere and ocean model nodes. Both models run simultaneously and the coupling can be sequential or concurrrent.

Good luck

[artem_vb]

Thank you very much for your answer. Hope it will help.

[OcGaBy]

I'm forcing with analitical wind, constant all the time over all my domine... I have run my application for 10 days and it never gets to a estationary or quasy-stationary state, and the kenetic energy coninues to grows all the time...

What can be wrong??

Greetings
GaBy

[kate]

What sort of domain are you talking about? I remember running Big Bad Basin (stratified box ocean with constant wind) and it would take over a hundred days to really get going.

[OcGaBy]

After running for 50 days I got quasi stationary conditions for my model with NW wind, so thanks Kate! You were right.

Now I am trying the same but with different wind directions, and it is always blowing up. I have try many different possibilities for NE wind, I mean week winds, strong, parallel to the coast, more perpendicular, with the same characteristics of my successful NW wind, just in the opposite directions and it doesn't work.

I try also a N and S winds with same result Blows Up..... exit_flag: 1

Does the exit_flags have a meaning? what is it for exit_flag: 1??

Any help, comment, suggestion is welcome.

Thanks in advance.
GaBy

[kate]

Does the exit_flags have a meaning? what is it for exit_flag: 1??

exit_flag alone doesn't mean a lot to me. Is there any other useful output? Does it run for 10 steps or 10 days before blowing up? I assume you mean it blows up with the energy growing to NaN or something like that. Can you look at plots from before the blow up to see what is going on? Anyway, if it goes only ten steps, try a shorter timestep.

[OcGaBy]

Thanks for your attention Kate…

Let see, it runs for 11.8 days, 102600 steps… so I don’t think DT is the problem. Kinetic energy starts to oscillate a bit and then goes to NaN:

Code: Select all

STEP	DAY	KINETIC_E	POTEN_E	TOTAL_E	NET_VOL
102300	11.840278	1.54E-01	3.82E+02	3.82E+02	4.57E+10
102330	11.84375	8.41E-02	3.83E+02	3.83E+02	4.57E+10
102360	11.847222	1.53E-01	3.81E+02	3.82E+02	4.57E+10
102390	11.850694	8.14E-02	3.81E+02	3.81E+02	4.57E+10
102420	11.854167	1.56E-01	3.82E+02	3.82E+02	4.57E+10
	WRT_HIS	-	wrote	history	fields
102450	11.857639	8.53E-02	3.83E+02	3.83E+02	4.57E+10
102480	11.861111	1.55E-01	3.81E+02	3.82E+02	4.57E+10
102510	11.864583	8.26E-02	3.81E+02	3.81E+02	4.57E+10
	WRT_RST	-	wrote	re-start	fields
102540	11.868056	1.59E-01	3.82E+02	3.82E+02	4.57E+10
102570	11.871528	8.66E-02	3.83E+02	3.83E+02	4.57E+10
102600	11.875	NaN	NaN	NaN	NaN

Something that disappoints me is that If I restart the model after it blows up (using the last restart file generated) it runs again and for much more time… Why can it be?...I am starting originally from static conditions.

I hope you can help me.

Cheers
GaBy

[jcwarner]

Just jumping in here without really knowing about your exact issue but --

Currently the restarts in ROMS are not exact. The restart is initiated with values of free surface, velocity, tracer fields, etc, but not their gradients (ie the restart only knows about the last time level, not the previous time level(s) required for the predictor-corrector time stepping. This issue has been presented in other discussion threads. So presently a restart does not provide exactly the same answer, but for most applications the difference is most likely not discernable to a simulation that just kept on running. For future releases, I believe that Sasha and Hernan are working on full restarts, but these would not be available for some time.

You say "it runs again and for much more time" So does it eventually crash again? You need to become more 'investigative here." Just keep restarting, running, and restarting until it crashes in just a few steps. Then save every step to a history file and look at the results. There will be something that does not look correct. Is it blowing up at the coast where it may be getting too shallow?? Is it blowing up at the boundary??? What are the tracer fields doing ?? You need to dig into the model results and see what it is telling you.

[kate]

On the issue of restart, most versions of ROMS are at their worst when it comes to vertical mixing. Are you using one of the fancy mixing schemes, such as Mellor-Yamada, GLS, or KPP? They need to reinitialize themselves after restart, so perhaps one of them is taking a while to get into trouble after the restart. Just a thought, but John's advice is good too.

Note that perfect restart is on The List, but Hernan's latest beta code saves more vertical mixing state to help with this problem (or perhaps hurt in your case).

[OcGaBy]

Hi!

Your information about restarting is very interesting...

I have been doing more experiments around it. In some cases (wind direction = NW) when I restart after a blow up the model gets to quasi stationary state, the kinetic energy looks stable, after 30 days and I leave it to run until 50 days...

What do you think about this case, is the problem that I am starting from static initial conditions? Can I take this restart run as a valid result?

In other cases (wind direction = S) I have restarted the model 5 times and it does crash again and again but after 6 or 8 days never immediately…

I am using GLS mixing scheme and KANTA_CLAYSON…

Any comment is welcome.

Greetings
GaBy

[OcGaBy]

Kate:

I would like to ask you if there is a simpler vertical mixing scheme that you can recommend me. I don’t have a strong justification to use GLS, it was a recommendation from another ROMS user that knows closely my case of study.

I will be waiting to read your opinion.
Cheers
GaBy