[MITgcm-support] stagger grid and obcs

Wed Apr 27 12:11:50 EDT 2016

Hi Yu-Kun,

On Wed, Apr 27, 2016 at 10:20:22AM +0800, ?????? wrote:
> Hi Jean-Michel,
> 
> 
> I've not used exch2 pkg.  After reading the source code, I understand that OBCS_uvApplyFac is used for 'backward compatibility'.  I am also pretty sure that the there are some mistakes in section 6.3.1.4 in the manual (latest version checked yesterday) because the description is inconsistent with the source code (e.g., OB_Jnorth(Ny) should be OB_Jnorth(Nx) etc).

There is good chances that some mistakes in the manual regarding size of OBCS index arrays
have been fixed this morning (Thanks to Martin); new (fixed) version of the manual should 
appear on-line after 4.pm EST. But if you find other mistakes, please let us know.

> I still confused on applying the initial condition (IC) and OB to the grids.  In my example, the IC I prepared has Nx=3 values uIC1 uIC2 uIC3.  How to assign these values to the grids?  Is it like u1=uIC1, u2=uIC2, u3=uIC3 and u4=uIC3 so that every IC values is used?

Must be that I was not clear: 
> (and the model tracer value will be overwritten by pkg/obcs); 
this applies to all main state variables U,V,T,S (& W if NH), at any time
including Initial conditions. And consistently, this should also appear in the output
file as mentioned here:
> And regarding model output, in the output file you will find at the OB location
> the OB value that pkg/obcs imposes/overwrites.

> In my example, I also have to prepare two OB values uOBw and uOBe.  If what you said is true, then in my example there is no any IC value is used because OB values overwrite the four velocity grids.

I am afraid but the domain you pick (Nx=3) is a little bit too small to generalize
to a larger domain: indeed in this case, none of the initial U velocity component
will be kept, and all will be replaced by what pkg/obcs imposes.

> The last question is should uOBw equals to uIC1/uOBe equals to uIC3 in preparation?
It's not strictly necessary, since the IC @ OB location will be "lost" (more
precisely, overwritten by pkg/obcs).

Cheers,
Jean-Michel

> 
> Thanks a lot.
> 
> ------------------
>  Best regards 
>  
> Yu-Kun Qian (??????) 
> Center for Monsoon and Environment Research 
>  Department of Atmospheric Sciences
> School of Environmental Science and Engineering 
>  Sun Yat-sen University 
> No. 135 Xingang West Road, Haizhu District 
> Guangzhou, 510275, P.R. China 
> Tel; 020-84115227 
> Email: qianyk at mail3.sysu.edu.cn      
> 
> 
>  
>  
>  
> ------------------ Original ------------------
> From:  "Jean-Michel Campin"<jmc at mit.edu>;
> Date:  Wed, Apr 27, 2016 01:01 AM
> To:  "mitgcm-support"<mitgcm-support at mitgcm.org>; 
> 
> Subject:  Re: [MITgcm-support] stagger grid and obcs
> 
>  
> Hi Yu-Kun,
> 
> You could find some documentation in MITgcm manual regarding OBC positions
> (section 6.3.1.4).
> 
> In your simple example, c1 & c3 will be tracer point OB, respectively Western 
> & Eastern OB (and the model tracer value will be overwritten by pkg/obcs); 
> same for tangential velocity component v1 and v3.
> Regarding normal component, the true OB values are u2 (Western OB) and u3 (Eastern OB),
> but for historic reason, the OB conditionis is also applied to u1 and u4, but this
> should not have any effect on the solution (to test this, we can set OBCS_uvApplyFac=0
> and check that the results are the same as with the default setting: OBCS_uvApplyFac=1).
> 
> And regarding model output, in the output file you will find at the OB location
> the OB value that pkg/obcs imposes/overwrites.
> 
> Finally, regarding how "halo" regions are filled, there are some ways to disable 
> the X or Y periodicity of the domain (when using pkg/obcs at both ends), but:
> a) without pkg/exch2: it's not well maintained (and not tested) and not sure if 
>  it works well.
> b) it does works with pkg/exch2 but involves some careful settings.
> c) it's not necessary to prevent "periodic domain" rules (even with OB at both
>   ends) and it works nicely with the default "periodic domain" assumption.
> 
> Cheers,
> Jean-Michel
> 
> On Fri, Apr 22, 2016 at 10:34:30PM +0800, ?????? wrote:
> > Dear MITgcm user,
> > 
> > 
> > Although I've asked related questions about stagger grids here, I am still a little confused about them.
> > 
> > 
> > Here let's make a concrete example in x direction only.  Assume that there are 3 zonal grids (Nx=3), then the zonal velocity points and mass points are laid out as:
> >       u1        c1        u2        c2        u3        c3        u4
> > 
> > 
> > When OBCS is off, periodic OB is used so that u1=u4.  In this case, we only need to prepare 3 zonal velocity values and 3 tracer values for u(1-3) and c(1-3).
> > 
> > 
> > When OBCS is on, OBwU is applied to u1 and u2 (Martin told me this is very important) and OBeU is similarly applied to u3 and u4.  Then the effective values of u is only 2 (Nx-1), which means that the one column of initial u field (Nx=3) will be overwritten by OBu.  I'm a little confused whether the outer most of initial velocity fields should be exactly the same as those of OBs?  If yes, then the effective velocity in the model interior should be Nx-1 and the output velocity field should contain at least one slice of OB velocity.  If true, which OB should be output? East or west?
> > 
> > 
> > Can anyone clarify these details?  Thanks in advance.
> > 
> > ------------------
> >  Best regards 
> >  
> > Yu-Kun Qian (??????) 
> > Center for Monsoon and Environment Research 
> >  Department of Atmospheric Sciences
> > School of Environmental Science and Engineering 
> >  Sun Yat-sen University 
> > No. 135 Xingang West Road, Haizhu District 
> > Guangzhou, 510275, P.R. China 
> > Tel; 020-84115227 
> > Email: qianyk at mail3.sysu.edu.cn
> 
> > _______________________________________________
> > 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

> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support