[CF-metadata] WRF staggered grids and vertical coordinates

[John Caron]

Jonathan Gregory wrote:

>Dear John
>
>It's not clear to me why we need this level of abstraction. If the basic grids
>can be described already by CF, is it really essential to have extra
>conventions to describe the relationships of grids, since this can be deduced
>at run-time? Conventions of this kind would introduce redundancy and hence the
>possibility of inconsistency. Maybe it is worth that price, but I'd be happier
>if we had a clear statement of why - is that possible?
>
>Thanks. Best wishes
>
>Jonathan
>_______________________________________________
>CF-metadata mailing list
>CF-metadata at cgd.ucar.edu
>http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
>

ok heres an abbreviated version of the example:

 float x(x);

 float x_stag(x_stag);


 float lat(t, y, x);

 float lon(t, y, x);

 float U(t, z, y, x_stag);

What are the lat, lon coords of  U? They are missing, because there is no

    float latForU(t, y, x_stag);

    float lonForU(t, y, x_stag);

You cant do the "coordinates" attribute for U with lat and lon because 
they dont share the same dimensions:

  float U(t, z, y, x_stag);
      U:coordinates = "lat lon";

I dont see how you could deduce x and x_stag are related without further 
annotation. The idea of the "stagger" attribute is to explicitly say  
that x and x_stag are values for the same coordinate, so that you can 
use the coordinates attribute:

float U(t, z, y, x_stag);
    U:coordinates = "lat lon";

 float x(x);
 float x_stag(x_stag);
    x_stag:stagger="x";