Hi Julien,
The grid mappings in appendix F are meant to be supplemental to the
latitude/longitude coordinates that all CF conforming files must contain.
The problem you raise concerning the size of the coordinate data can be
overcome if it is possible to store many data variables in a single file,
thus making the coordinate storage only a small fraction of the dataset
size. If that is not possible, then possible solutions include allowing a
grid mapping description to be sufficient coordinate metadata, or allowing
the lat/lon coordinates to be stored in a separate file from the data
variables. Neither of these options is available in CF-1.0.
Brian
On Mon, Mar 20, 2006 at 10:21:23AM +0100, Julien Demaria wrote:
> Brian, Jonathan,
>
>
> Thanks for your answers and help.
>
> The CDL description you give seems handle well my integerized grid, by
> giving all vertexes and area of each point.
> The only counterpart is that it uses a lot of disc space, 6 floats for
> each point...
> Using global bins idexes as in my example
> int bin_num(Data_Bins) ;
> bin_num:long_name = "Index number of each bin in the
> global sinusoidal grid" ;
> uses only an int by point.
>
> Is it really possible, as said Jonathan, to provide a new entry for
> Appendix F for the integerized sinusoidal grid ?
> I can provide the little algorithm and parameters needed to construct
> the grid.
>
>
> Regards,
>
> Julien
>
>
> Brian Eaton a ?crit :
> >Hi Julien,
> >
> >I agree the grid you describe is an example of what CF refers to as a
> >reduced grid. Generally the cells in a given latitude circle of a reduced
> >grid are all the same size, so the cell edges will not be aligned with the
> >cells in adjacent latitude bands that contain a different number of cells.
> >
> >In addition to Jonathans suggestion of a bin_area variable, the vertices of
> >the bins can be attached to the lat/lon arrays using the bounds attribute.
> >That would look like this:
> >
> >dimensions:
> > Data_Bins = 5965 ;
> > nver = 4;
> >variables:
> > short mean(Data_Bins) ;
> > mean:long_name = "Arithmetic mean of the geophysical variable
> > for each bin" ;
> > mean:units = "mg m-3" ;
> > mean:coordinates="lat lon";
> > mean:cell_measures="area: bin_area";
> > float lat(Data_Bins);
> > lat:units="degrees_north";
> > lat:bounds="lat_bnds";
> > float lon(Data_Bins);
> > lon:units="degrees_east";
> > lon:bounds="lon_bnds";
> > float bin_area(Data_Bins);
> > bin_area:units="m2";
> > float lat_bnds(Data_Bins,nver);
> > float lon_bnds(Data_Bins,nver);
> >
> >CF places the restriction that the vertices should be ordered
> >anticlockwise when viewed from above, but places no restriction on the
> >starting vertex.
> >
> >Brian
> >
> >
> >On Sun, Mar 19, 2006 at 08:15:07PM +0000, Jonathan Gregory wrote:
> >
> >>Dear Julien
> >>
> >>
> >>>Its characteristics are that on each row of latitude there is an integer
> >>>number of longitude bins. Poles always represented as 3 triangular bins.
> >>>Bins aren't aligned between rows.
> >>>In SEAWiFS/MODIS Level 3 HDF only bins (pixels) with data are stored,
> >>>and the geolocation information for each bin is only its indice in the
> >>>global integerized sinusoidal grid (first bin is on south, last on the
> >>>north).
> >>>
> >>>The only mecanisms I see in CF to do that are:
> >>>
> >>>- "5.3 Reduced horizontal grid": but:
> >>
> >>Yes, I think your case is most like this case.
> >>
> >>
> >>> Are bins always aligned ?
> >>
> >>You have to specify the latitude and longitude of each bin (grid point).
> >>
> >>
> >>> How can I know/compute the coverage of each pixel (its vertex) ?
> >>
> >>I am not sure what you mean. Do you mean the area of the grid box? This
> >>can
> >>be stored using cell_measures (CF 7.2).
> >>
> >>
> >>> Is this mechanism largely used ?
> >>> Are there free tools to see this kind of CF files ? (FERRET ?)
> >>
> >>I don't know whether Ferret will handle it. The intention of the
> >>convention
> >>is that a generic application will be able to interpret the lat and lon
> >>information without understanding the grid itself.
> >>
> >>dimensions:
> >> Data_Bins = 5965 ;
> >>variables:
> >> short mean(Data_Bins) ;
> >> mean:long_name = "Arithmetic mean of the geophysical variable
> >> for each bin" ;
> >> mean:units = "mg m-3" ;
> >> mean:coordinates="lat lon";
> >> mean:cell_measures="area: bin_area";
> >> float lat(Data_Bins);
> >> lat:units="degrees_north";
> >> float lon(Data_Bins);
> >> lon:units="degrees_east";
> >> float bin_area(Data_Bins);
> >> bin_area:units="m2";
> >>
> >>You would have to obtain the values of lat and lon by looking them up in
> >>the
> >>general grid by using the bin number, I imagine.
> >>
> >>You are right, there is no mechanism for describing your integerised
> >>sinusoidal
> >>grid at present. If you wish to include a description of the grid
> >>construction
> >>in the file, maybe it could be done by providing a new entry for Appendix
> >>F.
> >>
> >>Best wishes
> >>
> >>Jonathan
> >>_______________________________________________
> >>CF-metadata mailing list
> >>CF-metadata at cgd.ucar.edu
> >>http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
> >
> >_______________________________________________
> >CF-metadata mailing list
> >CF-metadata at cgd.ucar.edu
> >http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
>
>
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
Received on Mon Mar 20 2006 - 19:16:58 GMT
