[MITgcm-support] Non-Zero Meridional Velocity Kelvin Wave

Sun Nov 19 12:36:19 EST 2017

Hi Martin, 

Thank you very much for these precisions. GRID.h described indeed all that nicely and I should have looked at it much sooner.

My set-up works great now. I still have very small non-zero v values but they decrease significantly when I increase the grid resolution,
So, I expect them to be related to residual error, as pointed out by Yu-Kun.

Thanks again for all the help!

Best,
Bertrand

Le 19/11/2017 5:10 AM, « MITgcm-support au nom de Martin Losch » <mitgcm-support-bounces at mitgcm.org au nom de Martin.Losch at awi.de> a écrit :

    Hi Bertrand,
    Please have a look at the header of GRID.h where the grid layout is explained quite nicely.

    In general, all fields have the horizontal dimensions Nx *Ny. 3D fields have Nx*Ny*Nz. These are the parameters as specified in SIZE.h.
    Because of the c-grid staggering u needs an extra row Nx+1 and v needs an extra column Ny+1, but you never specify them. The are provided by the default periodic BCs.

    The bathymetry field is always 2D. There are many examples of how to generate initial conditions in the verification folder, look for gendata.m or gendata.py

    Nz+1 is always bottom an does not even exist in the model.
    If you want a solid boundary, say in the north, set the bathymetry=0 for j=Ny, as you did. Then your domain has only Ny-1 wet grid cells in the y-direction and the center in this direction is at the the center of cell j=Ny/2 for even Ny and at the interface between cells (Ny-1)/2 and (Ny-1)/2+1. 
    Does that make sense?

    Martin
    > On 17. Nov 2017, at 20:07, Bertrand Louis Rene Delorme <bdelorme at stanford.edu> wrote:
    > 
    > Hi Martin,
    > 
    > Thanks a lot for your suggestion. It did decrease the amplitude of V by a factor of 10, which is extremely positive
    > (as Yu-Kun mentioned, we cannot expect to have a meridional velocity exactly equal to 0, but still it should be pretty small).
    > 
    > I guess that I have been confused by the grid specification in the model and let me know if I got that right now.
    > 
    > Let’s say that I want a domain with a north and bottom boundary, where I specify U, V, T and ETA as IC. 
    > First, I generate a Nx*(Ny+1)*(Nz+1) grid. 
    > My bathymetry file, Bat, is of size Nx*(Ny+1)*(Nz+1), with Bat(:,:,Nz+1)=0 the bottom boundary, and Bat(:,Ny+1,:)=0 the north wall (elsewhere Bat=-H).
    > Then, the U grid is (Nx+1)*(Ny+1)*(Nz+1), V grid is Nx*(Ny+2)*(Nz+1), ETA grid is Nx*(Ny+1) and T grid is Nx*(Ny+1)*(Nz+1).
    > But since I prescribed a north and bottom boundary, I guess that the Y(Ny+1) column and Z(Nz+1) row are not considered by the model, is that right?
    > Now, I need my IC to be set on the relevant grid, ie: U_IC on (Nx+1)*(Ny+1)*(Nz+1), V_IC on Nx*(Ny+2)*(Nz+1), ETA_IC on Nx*(Ny+1) and T_IC on Nx*(Ny+1)*(Nz+1).
    > Yet, when I do that, I don’t get the IC I expect from the model. It seems to me that the model is expecting all IC dataset to be Nx*(Ny+1)*(Nz+1).
    > Does that mean that we need to define the initial velocities at cell boundaries without considering the last (or first?) boundary of the domain?
    > 
    > Thanks again,
    > Bertrand
    > 
    > 
    > 
    > Le 17/11/2017 3:23 AM, « MITgcm-support au nom de Martin Losch » <mitgcm-support-bounces at mitgcm.org au nom de Martin.Losch at awi.de> a écrit :
    > 
    >    Hi Bertrand,
    > 
    >    I didn’t read the thread very carefully but I found this
    > 
    >> I also checked my IC variables and those are perfectly symmetric about the equator (assuming that I am defining my initial U and V velocity at cell center, right?).
    > 
    >    Just to avoid confusion: the initial conditions are actually specified at the corresponding grid points. I.e. dynamic topograph Eta is a the center of the C-cell, Uini and Vini are at the U- and V- points of the grid cell. If you assumed that they are all co-located, that may be the problem?
    > 
    >    Martin
    > 
    > 
    >> On 17. Nov 2017, at 03:58, Bertrand Louis Rene Delorme <bdelorme at stanford.edu> wrote:
    >> 
    >> A little update on my problem (note also that the plot of Vm_dPHdy_Surf in my previous message is wrong):
    >> -          The full baroclinic pressure gradient force (ie: Vm_dPHdy-gravity*(ETAN(i) - ETAN(i-1) )/DYC)  and
    >> the Coriolis force (in Vm_Advec or Vm_Cori -- same) are the only forces occurring in the v-momentum equation 
    >> (AB_gV is non-zero but much smaller).
    >> -          I plotted those at the first time step (see enclosed). We can see that they do seem to have opposite spatial structures.
    >> Yet, looking at the difference of their amplitude ( ABS(f*u) - ABS(dPHI/dy) ), we can see that they do not balance
    >> each other perfectly. The induced acceleration is consistent with the plot of V.
    >> 
    >> So, I managed to find where the nonzero values in V come from, but still struggle to know why.
    >> If anyone has an idea, please let me know.
    >> 
    >> Thanks,
    >> Bertrand
    >> 
    > 
    >    _______________________________________________
    >    MITgcm-support mailing list
    >    MITgcm-support at mitgcm.org
    >    http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
    > 
    > 
    > _______________________________________________
    > MITgcm-support mailing list
    > MITgcm-support at mitgcm.org
    > http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
    _______________________________________________
    MITgcm-support mailing list
    MITgcm-support at mitgcm.org
    http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support