Comparaison des versions

Légende

  • Ces lignes ont été ajoutées. Ce mot a été ajouté.
  • Ces lignes ont été supprimées. Ce mot a été supprimé.
  • La mise en forme a été modifiée.

...

   Grd_typ_S='GU'   :Global Uniform grid - not used anymore

Grd_typ_S='GY':Global YinYang grid

Grd_typ_S='LU':Limited areaUniform

Grid rotation (optional)

...

One only needs to specify the core/free number of points in y-direction (Grd_nj) as well as the "overlap" in degrees (Grd_overlap). The model will calculate the number of points in x-direction as well as the size of the grid boxes automatically from that.

   Grd_nj:core/free number of grid points in y-direction
Including the overlap!!!
=> Number of points covering 90°+2*Grd_overlap
Grd_overlap:Overlap in degrees(!)
The overlap size should be at least 3 times the resolution of the grid resolution. In general, for higher resolution grids (ie: 0.25° or finer), an overlap of 8 times would be better for the solver.

Grd_nj:

Core/Free number of grid points in y-direction
Including the overlap!!!
=> Number of points covering 90°+2*Grd_overlap + 1 point

Note that there is no blending/merging area, also called Davies sponge, in Yin-Yang grids, which means that the free area is the same as the core area!

The overlap is needed because the two LAM grids will give different answers in the overlap region and the larger the overlap region the easier it is for the solver to find the common solution.The minimum overlap (degrees) should be 3 to 8 times the grid resolution (delta Y degrees).

To specify a Yin-Yang grid one should only specify 'Grd_nj' and 'Grd_overlap'. 'Grd_ni' will get calculated automatically using the following equation:

...

This means the overlap is 3 times larger on the left and right border of each LAM grid than at the top and bottom! This is done because the solver needs the grids to be 3 times larger in x- than in y-direction (minus 1 point). An overlap is needed because the two LAM grids will give different answers in the overlap region and the larger the overlap region the easier it is for the solver to find the common solution.The minimum overlap (degrees) should be 3 to 8 times the grid resolution (delta Y degrees).

Example to setup a 0.25° Yin-Yang grid

1. First calculate the size of the overlap when covering 8 times the grid resolution:

    Grd_overlap = 8 * 0.25° =

2. Then calculate the resulting total number (including 2 times the overlap) of core/free points in y-direction + 1 point:

    Grd_nj = (90° + 2*) / 0.25° + 1 = 377

That's all!

FYI, Grd_nj does not have to match the FFT criteria.

To learn more about Yin-Yang grids have a look at the following wiki page:  Yin-Yang grid

...

Specify the total number of points of the core(!) grid and the size of the grid boxes in degrees.
If possible, insure ensure that the total number of grid points of the core-grid in x-direction matches the FFT criteria. This will speed up your simulation by up to 10 %.
If you want to use spectral nudging, the number of grid points of the core-grid in x- and y-direction have to match the FFT criteria!

   Grd_ni:core(!) number of grid points of the core-grid in x-direction
(including blending but excluding pilot area)

Grd_nj:core(!) number of grid points of the core-grid in y-direction
(including blending but excluding pilot area)

Grd_dx:grid size in x-direction in degrees

Grd_dy:grid size in y-direction in degrees

...

    Grd_lonr  =  180.0,   Grd_latr  =  0.0

Click here to find out how to set blending and pilot area (halo).

...


Blending and pilot region

Pilot region

Also called "halo" or "nesting" region/area. This is a frame around the core grid (the grid specified by Grd_ni and Grd_nj). It contains the fields of the model used to drive the LAM model, interpolated to the model grid. The size of the pilot region in number of points gets set with the parameter 'Grd_maxcfl' in the namelist 'grid'. 'Grd_maxcfl' is the maximum supported Courant-Friedrichs-Lewy (CFL) number. The default as well as the minimum value is 1.

LAM grid

For a LAM grid, this pilot region contains the fields of the lateral driving data, interpolated to the model grid. The full width of the pilot region get calculated with:

Volet

    Pilot area = Grd_maxcfl + Grd_bsc_base + Grd_bsc_ext1

with:
    Grd_bsc_base = 5    : basic global lateral boundary conditions width
    Grd_bsc_ext1 = 2     : added points for proper de-staggering of u,v at physics interface

For LAM grids we usually set: Grd_maxcfl = 3

Yin-Yang grid

For a Yin-Yang grid this region contains the fields of the respective "other" LAM grid, interpolated to the "current" LAM grid.  The full width of the pilot region get calculated with:

Volet

    Pilot area = Grd_maxcfl + Grd_bsc_base + Grd_bsc_ext1

with:
    Grd_bsc_base = 6    : basic global lateral boundary conditions width
    Grd_bsc_ext1 = 2     : added points for proper de-staggering of u,v at physics interface

For Yin-Yang grids we usually set: Grd_maxcfl = 10

Blending area

Also called "merging area" or "Davies sponge". This area only exists for LAM grids. This is a frame just inside the core grid (the grid specified by Grd_ni and Grd_nj). In this blending area the results of the model are blended with the driving data, following cos2, giving more weight to the model results on the inside and less on the outside. The grid inside the merging area is called the "free" grid.

The size of the blending area in number of points gets set with the parameter 'Lam_blend_H' in the namelist 'lam'. The default value is 10, which is also the minimum that should get used. As a rule of thumb the blending area should be large enough to cover at least 1 point of the driving data.

FFT criteria

Having a number of points fulfilling the FFT criteria can speed up a simulation by up to 10%.

To fulfill the FFT criteria the number of core grid points in x-direction, excluding the pilot zone'Grd_ni', must be a multiple of 2, 3, and/or 5. Use the findfft GEMCLIM script to find the possible You can use the script 'findfft' to list all numbers of grid points fulfilling the FFT criteria .
Again you need to insure that the model environment is set.
Then you can use findfftin a certain range of points.

Arguments:

   findfft  -min  : minimum number of grid points, default 150


-maxmaximum number of grid points, default 250


This command will then print all "good" values you can use for 'Grd_ni' and 'Grd_nj' (can be different) between the range you specified with 'min' and 'max'.

FYI, when using spectral nudging, the number of grid points specified in 'Grd_ni' and 'Grd_nj' have to match the FFT criteria!!!

Verify your grid

After having specified a grid in the file 'gem_settings.nml' you can have a look at the grid you specified using the grille script - see description below.

In the directory in which you have your file 'gem_settings.nml' simply type :

        grille  -xrec
or, if you just want to see the free grid:
    grille  -xrec  -grid free

...