CDO

Hi Julia!

the gme grid was developed by DWD I think (https://journals.ametsoc.org/view/journals/mwre/130/2/1520-0493_2002_130_0319_togihg_2.0.co_2.xml) and is part of the GRIB standard becaus GRIB is what all weather services are supposed to use for data exchange.

Thats why the coordinates are not part of the grib file and a simple linking of values and locations is not part of it.

If you want to have that link I recommend switching to netcdf format, where the coordinates are part of the data file (if you stick to the CF-convention). CDO can do this conversion by using a corner-case of the reducegrid operator. see here for more examples on this.

1. create a mask file with your target grid, which is 1 everywhere

   cdo const,1,gme24 mask1.grb

2. _fake_reduce the input data on this grid with

   cdo -f nc reducegrid,mask1.grb data_gme24.grb data_gme24.nc

   This will create a CF-conform version of your input data. As a consequence the data is on an unstructured grid, which means tools like nvciew will not be able to plot it. unfortunately there not many tools to plot data like this

The result should look like this:

netcdf t_gme24 {
dimensions:
    ncells = 6250 ;
    vertices = 6 ;
variables:
    double lon(ncells) ;
        lon:standard_name = "longitude" ;
        lon:long_name = "longitude" ;
        lon:units = "degrees_east" ;
        lon:bounds = "lon_bnds" ;
    double lon_bnds(ncells, vertices) ;
    double lat(ncells) ;
        lat:standard_name = "latitude" ;
        lat:long_name = "latitude" ;
        lat:units = "degrees_north" ;
        lat:bounds = "lat_bnds" ;
    double lat_bnds(ncells, vertices) ;
    float var1(ncells) ;
        var1:table = 255 ;
        var1:CDI_grid_type = "unstructured" ;
        var1:coordinates = "lat lon" ;

Variable var1 has coordinates lon and lat which are given in the respective coordinate variables in the file. their bounds (cell corners) are saved in lat_bnds and lon_bnds.

Another way of getting data values and their coordinates is to use the outputkey operator:

cdo -outputkey,lon,lat,value data_gme24.grb

Will print 3 columns of the corresponding values. So if you prefer working with text rather than netcdf, this might the better choice.

The problem gets a bit worse in case you have an additional time dimension in your input, but let's save this for later ;-)

cheers
ralf

Julia Kaltenborn wrote:

Hi there,

I have a bunch of sea-level pressure netcdf files on a longitude/latitude grid that I am remapping to an icosahedral-hexagonal grid with cdo.

The relevant command I use is this one here:

[...]

See here for the full shell code I am using: https://github.com/liellnima/icosahedral/blob/main/remap2icosahedral.sh

WOW - shell scripts with comments, default values and consistent spacing. I really wish more ppl would pay attention to things like u

Hi Ralf,

Thank you so much for your super helpful response!

The outputkey solution worked perfectly, I used

cdo -outputkey,xind,lon,lat data_gme24.grib

and got exactly the mapping I was looking for.

Your first approach is a bit more elegant though and I think I'd prefer using it like that for the future. I didn't get that running though, I get a "No match for shortName=const" error. Do you know what is going on there?

And thanks for the compliment regarding the shell scripts, glad someone appreciates it

Thank you for your great help!
Julia

Hi again!

Glad you have a solution now. For debugging the remaining error, I'd need the real data and the CDO version.

On the other hand we can do this any time you decide to switch from text to netcdf - as you like

Have a nice Monday!
ralf