[MITgcm-support] 3D interpolation of model state by flt floats

[Paola Cessi]

Hi,

the FLT package as it is now interpolates all fields using neighboring points, including land points. This is OK for velocities which are indeed zero on the boundaries, but not for T,S, tracers and SSH, which are set to zero on the land points, which is not their true value. So for T, S, tracers and SSH it is better not to interpolate at all  and just use the value on the grid at the nearest cell point, i.e. using NINT in all three directions. 

Paola

----------------------------------------------
Paola Cessi   
Scripps Institution of Oceanography  
9500 Gilman Drive #0213               
La Jolla, CA 92093-0213
USA               
Tel: +1 858 534 0622   Fax: +1 858 534 8045
e-mail: pcessi at ucsd.edu     Web:  http://www-pord.ucsd.edu/~pcessi

Delivery address (Fed-Ex, DHL, UPS, etc.):
Paola Cessi
University of California San Diego
Keck Center - Room 366
8851 Shellback Way
La Jolla, CA 92093-0213
USA


> On Mar 29, 2020, at 7:42 AM, 钱钰坤 <qianyk at mail3.sysu.edu.cn> wrote:
> 
> Hi all,
> 
> When I tried to sample the model state using 3D floats (flt pkg), I found that the sampled fields are not smooth (like a step funtion in some way) as I expected (I made a comparison with the offline-sampled results). When I looked into flt_traj.F, there are some codes doing the interpolation of model state by floats.  For a 3D type float (not horizontal 2D float), I would expect a trilinear (x, y, z) interpolation of the model state. However, I found the following lines (https://github.com/MITgcm/MITgcm/blob/master/pkg/flt/flt_traj.F#L134 <https://github.com/MITgcm/MITgcm/blob/master/pkg/flt/flt_traj.F#L134>):
> 
>     CALL FLT_BILINEAR  (ix,jy,uu,uVel,  kp,1,bi,bj,myThid)
>     CALL FLT_BILINEAR  (ix,jy,vv,vVel,  kp,2,bi,bj,myThid)
>     CALL FLT_BILINEAR2D(ix,jy,pp,etaN,     0,bi,bj,myThid)
>     CALL FLT_BILINEAR  (ix,jy,tt,theta, kp,0,bi,bj,myThid)
>     CALL FLT_BILINEAR  (ix,jy,ss,salt,  kp,0,bi,bj,myThid)
> 
> For a 3D linear interpolation, I think FLT_TRILINEAR would be much better, with proper vertical position kp. Here kp is a roundoff of the vertical position (https://github.com/MITgcm/MITgcm/blob/master/pkg/flt/flt_traj.F#L106 <https://github.com/MITgcm/MITgcm/blob/master/pkg/flt/flt_traj.F#L106>):
> 
>     kp = NINT(kpart(ip,bi,bj))
> 
> When I change FLT_BILINEAR(...) to FLT_TRILINEAR(ix,jy,kpart(ip,bi,bj),tt,theta,0,bi,bj,myThid), I got almost identical results both online and offline. No step-funciton-like behavior.
> 
> Shouldn't it be changed to FLT_TRILINEAR?
> 
> ------------------
> Best regards 
> 
> Yu-Kun Qian (钱钰坤) 
> 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 <mailto:qianyk at mail2.sysu.edu.cn>
>  
>  
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20200329/7baa151d/attachment.html>