[MITgcm-support] Freshwater OBCS problem
jklymak at uvic.ca
Sat Dec 22 12:54:25 EST 2012
Does it die after one time step or one save step? You can set your MonitorFreq to equal your time step to get diagnostics each step. I often set dumpFreq equal to my time step as well to save each step if I have something blow up early in the run.
Any idea where your NaN's are? An uneducated guess would be that your salinities go slightly negative and the model doesn't accept negative salinity for some reason and NaN's those salinities. Have you tried setting S=1 psu (or 10 psu) just to see if it can run for a while? I don't think the advection scheme is up to keeping a zero salinity front strictly greater than zero.
Good luck - hopefully someone else with experience putting a river into their model speaks up.
On Dec 22, 2012, at 9:09 AM, Dustin Carroll <spanishcastlemagik at gmail.com> wrote:
> Hi Jody,
> Thanks for your response. I do have the MY82 package turned on.
> I changed my river flow to ramp up linearly over two days (with max flow being 1 m/s), but I still get NaN's from cg2d on my second time step.
> Date: Fri, 21 Dec 2012 10:44:50 -0800
> From: Jody Klymak <jklymak at uvic.ca>
> To: mitgcm-support at mitgcm.org
> Subject: Re: [MITgcm-support] Freshwater OBCS problem
> Message-ID: <68F248F6-B3F9-41DD-B34F-87BD47334E07 at uvic.ca>
> Content-Type: text/plain; charset=us-ascii
> Hi Dustin,
> I've not tried to implement a river, and I bet others have, but...
> Have you considered ramping up your river flow over a day or two? That might give the river plume time to set up more gently in the fjord and allow mixing processes that need to happen happen out in the fjord naturally rather than jamming a fast river into a strong salinity front, which means you would expect a lot of small scale, and likely unresolved turbulence which the model turns into unstable noise.
> Also, what mixing do you have in addition to Kz and Az? I'd think you want something that allowed convective instabilities and/or shear instabilities to mix more than normal. MY82 probably does a good job, unless you are also interested in the energetics of the turbulence. That will suppress a lot of grid noise that might be causing you to go unstable.
> Cheers, Jody
> On Dec 21, 2012, at 9:39 AM, Dustin Carroll <spanishcastlemagik at gmail.com> wrote:
>> I have a model set up to simulate circulation in a fjord. I am using OBCS with my western boundary being forced by the tides (with a 100km sponge layer). The eastern boundary is a small channel through the topography / bathymetry. Through the eastern boundary I am forcing fresh water into the fjord with a u vel of 0.5 m/s.
>> The initial salinity prescribed to the model grid varies between 32 and 35, increasing with depth.
>> I am having a problem with my model blowing up in the first time step. I seems that anytime I turn on the freshwater eastern boundary I get NEGATIVE SALINITY warnings and the cg2d doesn't converge and I get NaN's. Is there anything else I have to consider when having such a strong salinity gradient at a OBCS?
>> I have tried setting freshwater flux to true, increased my cg2d iters / lowered my timestep, and tried both 33 and 77 advection schemes. I am using EXF for forcing my boundary conditions.
>> # ====================
>> # | Model parameters |
>> # ====================
>> # Continuous equation parameters
>> # Elliptic solver parameters
>> # Time stepping parameters
>> # Gridding parameters
>> # Input datasets
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
> Jody Klymak
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the MITgcm-support