[MITgcm-support] curvilinear grid

Fri Oct 21 07:35:53 EDT 2011

Hi Madhu,

haven't looked at your files yet, but what you see means that the model is "blowing up" (I think that "SOLUTION IS HEADING OUT OF BOUNDS" is pretty clear, isn't it?).

Usually this happens when your time step / viscosity values violate certain stability criteria. In the monitor output there is an "adv_uvel_max" (and vvel, wvel) that checks the classic CFL criterion, but there are also others, please read the documentation to find out about numerical stability in general, e.g. here
<http://mitgcm.org/public/r2_manual/latest/online_documents/node120.html> (this is one of the tuturial examples that does not necessary apply to your experiment, there are some criteria listed). 

Modify your time steps and viscosity parameters to get a stable model.
Also have a look at the verification/internal_wave example for using the Crank-Nickolson time stepping scheme, which might be interesting for internal wave studies (see also <http://mitgcm.org/public/r2_manual/latest/online_documents/node41.html>).

About the grid parameters:
For a first guess you should be fine with your hack, but I strongly recommend that you generate  more accurate values, as all of the conservation properties of the model rely on these grid parameters. Have a look at ini_cartesian_grid.F or ini_spherical_grid.F to see how this is done in a simpler case. You should be able to get satifactory results by appropriately averaging the availalbe parameters.

Martin

On Oct 16, 2011, at 11:30 AM, Madhu wrote:

> Sorry Martin,
> This time I am sending the entire project folder as 'madhu.tar'. 
> 
> As per your suggestion, after removing the 'useCubedSphereExchange=.TRUE' flag in eedata file the previous warning massage has gone. 
> However, the model had stopped at 7th itteration  with the following massage on the screen
> 
> $ ./mitgcmuv >curvi.txt
> SOLUTION IS HEADING OUT OF BOUNDS: tMin,tMax= -2.356E+10  5.602E+10
>   exceeds allowed range (monSolutionMaxRange=  1.000E+03)
> MON_SOLUTION: STOPPING CALCULATION at Iter=         7
> S/R ALL_PROC_DIE: ending the run
> STOP ABNORMAL END: S/R MON_SOLUTION, stops due to EXTREME Pot.Temp
> $ 
> 
> Also Martin, it will be important to mention here that some of the grid varible used in the input folder are having dummy values. The curvilinear grid used here is derived from SEAGRID tool (SeaGrid Orthogonal Grid Maker For Matlab). Which generates the curvilinear grids for the models like ECOM, POM and ROMS. But it does not calculate the all input varibles required for MITGCM ( the varible calulated are RA,DXF, DYF, LATC, LATC, Z, ANG (angle of the x-axix w.r.t North)). Therefore for initial test i took the dummy values as
> DXF=DXG=DXV
> DYF=DYG=DYU
> LATC=LATG
> LONC=LONG
> RA=RAZ=RAW=RAS
>  
> I also tried to plot the Bathy written from model output, but it is coming as two tiles.
> Do you think this problem will be resolved if we give correct grid information? 
>  
>  
> Regards
> Madhu Joshi
> 
>  
> On Thu, Oct 13, 2011 at 1:38 PM, Martin Losch <Martin.Losch at awi.de> wrote:
> Madhu,
> 
> still not enough, but for a start, you do not need this (in eedata)
> useCubedSphereExchange=.TRUE.,
> unless you use the cubed sphere (which you don't).
> 
> May I ask which experiment you started from/used as a template?
> 
> Please provide the CPP-flags (content of the "code" directory), that are non-default.
> 
> Martin
> 
> On Oct 13, 2011, at 8:11 AM, Madhu wrote:
> 
> > Hi Martin,
> >  Thanks for reply..
> > Attached are the required flies
> > curvi.txt  :  is the screen dump after executing MITGCM
> > list.txt : list of the files in INPUT folder..
> >
> > Regards
> > Madhu
> >
> > On Wed, Oct 12, 2011 at 2:24 PM, Martin Losch <Martin.Losch at awi.de> wrote:
> > Madhu,
> >
> > we need more explanation/error messages in order to help you. E.g. the content of your code directory and namelists (data) files, the content (tail) of your curvi.txt.
> >
> > M.
> >
> >
> > On Oct 12, 2011, at 8:34 AM, Madhu wrote:
> >
> > >
> > > Hi All
> > > I am new user to MITGCM and want to use the model for eastern coast of India (in Bay Of Bengal) to study the internal waves.
> > > I have tested the model for rectangular domain. It has run and results are ok.
> > > But ultimately i have to run the model using orthogonal curvilinear grids (attached as *.pdf) to reduce the computational time. I have generated the curvilinear grids/bathy and associated parameters are LONC.bin, LATC.bin, DXF.bin, DYF.bin, RA.bin, topo.bin etc. When the model is run with new grid setup , it failed to run and gave the Error massage as follows.
> > > ++++++++++++
> > > $./mitgcmuv >curvi.txt
> > > ** WARNING ** INI_MODEL_IO: use tiled-files to write sections (for OBCS)
> > > EXCH1_RS_CUBE: Wrong Tiling
> > > EXCH1_RS_CUBE: works only with sNx=sNy & nSx=6 & nSy=nPx=nPy=1
> > > STOP ABNORMAL END: EXCH1_RS_CUBE: Wrong Tiling
> > > $
> > > ++++++++++++++
> > >
> > > I want to test it first with single tile case hence in 'SIZE.h'  file the setup was as follow
> > > & sNx = 50,
> > > & sNy = 100,
> > > & OLx = 1,
> > > & OLy = 1,
> > > & nSx = 1,
> > > & nSy = 1,
> > > & nPx = 1,
> > > & nPy = 1,
> > > & Nx = sNx*nSx*nPx,
> > > & Ny = sNy*nSy*nPy,
> > > & Nr = 5)
> > >
> > >
> > >
> > > I will be happy to know,  If there is any suggestion with which i can resolve this issue.
> > >
> > > Also if there is any other similar test case please send the information.
> > >
> > > Waiting for reply
> > > --
> > > Regards
> > >
> > >  Madhu
> > > <Grid.pdf>_______________________________________________
> > > MITgcm-support mailing list
> > > MITgcm-support at mitgcm.org
> > > http://mitgcm.org/mailman/listinfo/mitgcm-support
> >
> >
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-support
> >
> >
> >
> > --
> >
> >  Madhu
> > <curvi.txt><list.txt>_______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
> 
> -- 
> 
>  Madhu
> <madhu.tar>_______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support