CDO

Hi,

here is the "ncdump -h" of your attached file:

netcdf aod {
dimensions:
nwl = 14 ;
lat = 96 ;
mlev = 40 ;
ilev = 41 ;
time = UNLIMITED ; // (12 currently)

there is no "lon" dimension. And CDO doesn't support [I guess] generic NetCDF files. However, you may try the same by selecting the appropriate variable along with "selindexbox" (see the docs for more info)

for example:

cdo -selindexbox,1,14,1,96 -selvar,aod aod.nc ofile.nc

Cheers,
J.

Thanks, Jaison.

I'd like to average a field over a particular latitude range (e.g. 20N-20S), and the field is on a lat,lev grid as in the previously attached file, so there's no lon dimension. I haven't found a way to do this average in cdo, so I added a lon dimension. Now the original lat-lev field is repeated at every longitude, which should allow me to use "cdo fldmean -sellonlatbox .." on the new field. I've attached the corresponding file (aod_lon.nc). Sellonlatbox won't work on it though, and I wonder why..

cdo sellonlatbox,-180,180,-20,20 aod_lon.nc ofile.nc
cdf_get_vara_float : ncid = 65536 varid = 4
Error (cdf_get_vara_float) : NetCDF: Index exceeds dimension bound

Cheers
Angela

Hi,

I think the problem is that the NetCDF variable attibuted/co-ordinate variables are not in the standard format which cdo recognize. For example, although ncdump -h shows there are 12 time steps, CDO only show one timestep. so I think it might better to edit the file manually and then apply the cdo operator:

1. ncdump aod_lon.nc > my_data.dat
2. open the file "my_data.dat" in an editor and change all (search & replace) the "nwl" (which I guess is the levels) to "lev" 
3. change the "time = 12 ;" (on top) to time = UNLIMITED ; 
4. change "float aod(nwl, lon, lat, time) ;" to "float aod(time, lat, lon, lev) ;" 
5. save file with all changes
6. ncgen -o new_file.nc my_data.dat

now try

cdo sinfov new_file.nc

which shows all the 12 time steps, and

cdo sellonlatbox,-180,180,-20,20 new_file.nc ofile.nc

will work now but of course, check the results.

I'd like to average a field over a particular latitude

you could use "mermean" (see docs: meridinonal stats) in the original file, I guess?

Anyway, hope this helps,

Cheers,
J.

Yes, the order of the array dimensions is not supported in CDO. At least the time dimension have to be the first dimension.