[RegCNET] PostProc

[Graziano Giuliani]

On 06/16/2011 08:28 PM, Felipe Roque wrote:
> Dear RegCM users,
>
> I´ve been watching many questions about post proc procedures recently. In
> fact, I think it should be created a better document about post proc
> routines.
>
> I have a doubt! I´ve just finished a simulation and I´m using the sequence
> below to generate monthly means of ATM files.
>
> 1) sigma2p ...
> 2) average month ...
> 3) GrADSNcPrepare ...
>
> My regcm.in is set like this (128x138x18 - with 50 km resolution):
>
>   &dimparam
>   iy     = 138,
>   jx     = 128,
>   kz     = 18,
>   nsg    = 1,
>   /
>   &geoparam
>   iproj = 'NORMER',
>   ds = 50.0,
>   ptop = 5.0,
>   clat = -20.00,
>   clon = -57.00,
>   plat = -20.00,
>   plon = -57.00,
>   truelatl = 30.0,
>   truelath = 60.0,
>   i_band = 0,
>   /
> ....
> ....
> ....
>
> The problem is that when I finish post proc sequence above,  GrADSNcPrepare
> returns to me a .CTL file with the following information:
>
> pdef  126 136 bilin sequential binary-big ^AS.coord
> *xdef   273* linear  -86.67   0.22
> *ydef   266* linear  -46.39   0.22
>
> The question is: how does the model works with a 128x138 grid (50 km
> resolution) and post proc returns a 273x266 grid (~25 km resolution)?!
> PS: domain is conserved!
>
[...]

Felipe,

the PostProc chain You are using is it just a possible path, we encourage users 
to share new paths (see NCL, R, GrADS scripts examples in the Tools/Script 
directory).

About the GrADS CTL file. In the pdef line it is clear that "native" model grid 
is 126x136 in output (boundary points used in internal computations are 
stripped). The GrADS rendering engine does not know anything about projection, 
it just maps screen pixels to latitude-longitude pairs in a regular grid.
So interpolation is needed. The choice of the output grid to interpolate on is 
arbitrary, any regular latlon grid with fixed pace is OK.

The algorithm used to "hint" the possible grid resolution is:

   rlatinc = ds/111000.0/2.0
   rloninc = ds/111000.0/2.0

where ds is the grid resolution in the model. The algo was fixed from regcm3 
this way to be a reasonable compromise between the "fidelity" to the original 
grid size and the graphical rendering of the fields by GrADS. Note that if an 
interpolation is needed (and is the case), the field is not plotted at original 
point locations anyway. Graphically the rendering using this setting did seem 
smoother. Feel free to experiment with the code different settings: this is just 
how You tell the GrADS engine to render graphically the data.

Ciao!

Graziano.

-- 
Graziano Giuliani - Earth System Physics Section
The Abdus Salam International Centre for Theoretical Physics
Strada Costiera, 11 - I - 34151 Trieste Italy