open boundary condition problems

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply
Message
Author
smorey
Posts: 5
Joined: Thu Jul 05, 2007 9:20 pm
Location: Florida State University

open boundary condition problems

#1 Post by smorey » Fri Feb 22, 2008 11:50 pm

I am running a realistic northeastern Gulf of Mexico simulation with open southern and western boundaries. I am trying to apply boundary conditions derived from a larger scale model. I have success with baroclinic velocities and scalar fields, but am having trouble with the free surface and barotropic velocity components. If I choose ANY of the following options:
*_FSCHAPMAN
*_M2FLATHER
*_FSREDUCED
Then my model immediately blows up. My domain has land along the northern part of the western boundary and eastern part of the southern boundary. Inspection of the model fields just after blow-up shows NaNs propagating from the last ocean grid cell along each boundary adjacent to the land.
I have double-checked the values I am giving the model for the open boundary conditions and all appear correct. If I choose the NUDGING and RADIATION options for FS and M2, then the model works (although I'm having trouble with it losing or gaining mass). I am not currently applying tides, but just t/s/u/v/ssh fields from a larger-scale model.
Has anyone attempted to apply these boundary conditions along a partially open boundary (one that has some land on one end or the other)?
Thanks for any help
Steve

User avatar
wilkin
Posts: 526
Joined: Mon Apr 28, 2003 5:44 pm
Location: Rutgers University
Contact:

#2 Post by wilkin » Sat Feb 23, 2008 2:43 am

My experience is that the Chapman and Flather combination for zeta and ubar/vbar works well in conjunction with tides (which is what the Flather condition was designed for). I have not used them without tides.

The ADD_FSOBC and ADD_M2OBC flags combine the slowly varying boundary data (from an external model for example) to the tidal variability computed internally from the harmonic data. Perhaps you have omitted these options and the external model data are unused. Does stdout report that it is reading the boundary data?

Without tides, the NUDGING/RADIATION combination makes sense, and will make use of the boundary data you provide. You indicate this works OK. So if that works, I'm not sure what there is to be achieved by using Chapman/Flather options. Save those for when you want to add tides.

My boundaries are partially open, so this has been tested. If you apply the VOLCONS options an adjustment will be made to conserve volume on every timestep. This would not make sense when tides are included, however, because there is no reason to expect the volume of a limited area model to remain constant through all phases of the tide.

John.
John Wilkin: DMCS Rutgers University
71 Dudley Rd, New Brunswick, NJ 08901-8521, USA. ph: 609-630-0559 jwilkin@rutgers.edu

kurapov
Posts: 26
Joined: Fri Sep 05, 2003 4:49 pm
Location: COAS/OSU

#3 Post by kurapov » Mon Feb 25, 2008 5:36 pm

Double-check your open boundary files. We've seen the statements like "I've checked
the boundary files and they were fine". And then, later "sorry, there was a NaN at one point in one snapshot". NaN does not appear for no reason. Something has been set NaN.

You obtained your open boundary files by interpolation from some larger scale product. Check values near the surface, bottom, coast, i.e., in places where the larger scale model was probably undefined. For instance, what did you do when your ROMS depth is larger than that of the large scale product?

If everything is fine, try to use Flather for M2 and Chapman for FS. Do not select the reduced physics option. FL-Ch should work (it should not blow up immediately).

As for what makes sense to use... FL-Ch makes sense to use. Flather conditions are in fact conditions for the first Riemann wave characteristic (there are 3 such characteristics for shallow water equations; to provide well-posed formulation one should use one, two, or three depending on inflow/outflow and the strength of the advective current ; FL conditions must be speciifed in any case - e.g., see a recent paper by Debru and Blayo, Oc. Mod. ??). If anybody can justify the use of radiation + nudging from similar math. positions, drop me a line.

Cheers.

smorey
Posts: 5
Joined: Thu Jul 05, 2007 9:20 pm
Location: Florida State University

#4 Post by smorey » Tue Feb 26, 2008 5:09 pm

Thanks for the advice. I will continue to check the ob data that I am setting up, but ROMS does print the max/min each time it reads a field and if there are NaNs, I know from experience that Iwill see NaNs in the diagnostic messages. I don't see NaNs. I'll keep looking but please keep the suggestions coming!

kurapov
Posts: 26
Joined: Fri Sep 05, 2003 4:49 pm
Location: COAS/OSU

#5 Post by kurapov » Tue Feb 26, 2008 5:14 pm

Good poi`nt, about the diagnostic. Dig it and keep us updated.

drivas
Posts: 29
Joined: Tue Aug 28, 2007 8:31 pm
Location: CICESE

Open boundary condition problems

#6 Post by drivas » Tue Feb 26, 2008 6:23 pm

Dear Steve,
A few months ago, when I started running ROMS, I had the problem that my model blew up after a one-month long running. I had checked my input files and no NaNs were in them. My problem was the following: I'm forcing my model by using realistic wind from QuickScat, but when I pre-processed my forcing files, one bad-flagged survived and few large values appeared after the interpolations. Perhaps any unrealistically data is still in your input files (bad-flagged or the proper scaling wasn't applied), don't you think?

User avatar
jivica
Posts: 135
Joined: Mon May 05, 2003 2:41 pm
Location: The University of Western Australia, Perth, Australia

#7 Post by jivica » Wed Feb 27, 2008 8:36 am

Well I can confirm that model blows up is one have quite large values (or low) for forcing.. What I do is I find nans and replace them with nanmean(field) rather than putting just 0 or some missing_value
... After that all is working oky..
Cheers,
Ivica

Post Reply