[CF-metadata] "grid relative" vector components revisited

[Rich Signell]

All,

Last fall I asked about how to represent "grid relative" vector components and 
the answer was to use the "standard_name" attribute to indicate this.  I'm 
beginning to wonder whether this will get us into trouble. Here's why;

For u and v velocity components oriented along "i" and "j" directions we specify:

u:standard_name="sea_water_x_velocity"
v:standard_name="sea_water_y_velocity".

Great.  That's fine as long as we only have a few of these standard_names, but 
what about grid-relative fluxes?  We often store fluxes of material in the 
NetCDF file because we cannot store the velocity and tracer fields at sufficient 
intervals to compute accurate fluxes a posteriori.

The problem is we have *lots* of these.  We have many tracer variables and we 
therefore have endless grid-relative fluxs with components like "200 micron sand 
concentration *  sea_water_x_velocity".   We don't really want to invent 
standard_names for all these, and require that tools like IDV parse the whole 
standard_name list just to figure out whether to rotate the vectors or not when 
it puts arrows on a map.  I think it would be much simpler to have some sort of 
property that just indicates "grid_relative".   Perhaps via the linkage to a 
"grid" variable, as I suggested below?

-Rich


Rich Signell wrote:
> All,
> 
> The U and V staggered velocities computed in Arakawa C grid
> models are oriented relative to the grid in the "I" and "J"
> directions, not true east and north (unless the grid happens to
> be oriented N/S, E/W).  On a general curvilinear orthogonal grid,
> the orientation of these components change at every grid cell.
>                                                                                
> 
> It seems the proper way to specify this in CF is through the
> use of
>                                                                                
> 
> u:standard_name="sea_water_x_velocity"
> v:standard_name="sea_water_y_velocity",
>                                                                                
> 
> Is this right?
>                                                                                
> 
> One could alternatively have this be inferred for U and V
> variables that have Arakawa C grid defined (via the proposed
> "grid" attribute and "grid_type"="Arakawa C")
>                                                                                
> 
> -Rich
> 
Jonathan Gregory wrote:
 > Dear Rich
 >
 >
 >>It seems the proper way to specify this in CF is through the use of
 >>u:standard_name="sea_water_x_velocity"
 >>v:standard_name="sea_water_y_velocity",
 >>Is this right?
 >
 >
 > Yes.
 >
 > The file might also contain northward and eastward velocity components.
 > Velocity components relative to lat and lon and relative to the grid are
 > different quantities so need different standard names, I think, just like
 > northward and eastward components themselves have different names.
 >
 > Cheers
 >
 > Jonathan
 > _______________________________________________
 > CF-metadata mailing list
 > CF-metadata at cgd.ucar.edu
 > http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
 >
-- 
Richard P. Signell
U.S. Geological Survey       Phone: (508) 457-2229
384 Woods Hole Road          Fax:   (508) 457-2310
Woods Hole, MA  02543-1598