Project

General

Profile

Conservative interpolation from reduced Gaussian to regular lat-lon grid

Added by Peter Watson over 9 years ago

Hello,

I'm trying to interpolate data conservatively from a reduced Gaussian grid to a regular lat-lon grid. A colleague of mine at another institute says this should be possible using CDO, using the following command (if the output grid is 2.5x2.5 deg):

cdo -t ecmwf -R remapcon,r144x72 file_in file_out

(where the -t option specifies to use the ECMWF GRIB table - removing it makes no difference).

This gives me the following error message:

"cdo remapcon (Abort): Unsupported grid type: gaussian reduced, use CDO option -R to convert reduced to regular grid!"

My colleague seems to think that this works for him, so I was wondering should it be possible, and if so what do I require? I currently have CDO 1.5.3 and grib_api libraries 1.9.18 - do these need to be updated? I can't find information on how to do this specific task in the CDO documentation.

Thanks very much for any help you can give.


Replies (2)

RE: Conservative interpolation from reduced Gaussian to regular lat-lon grid - Added by Uwe Schulzweida over 9 years ago

The CDO option -R works only on GRIB edition 1 files. A more general solution to convert a reduced grid to a regular grid is the CDO operator setgridtype with the parameter regular:

cdo remapcon,r144x72 -setgridtype,regular file_in file_out
This feature is available since CDO version 1.6.2.

RE: Conservative interpolation from reduced Gaussian to regular lat-lon grid - Added by Peter Watson over 9 years ago

Thanks for the reply. I've updated my CDO and have got the interpolation to work. In testing I have done, it doesn't seem to be perfectly conservative - area averages are different to those in the original data set by ~5-10% - is this supposed to be the case? The output from CDO indicates that the conservative interpolation is done from a Gaussian to a lat-lon grid - what interpolation method is used to regrid the data from the reduced Gaussian to the Gaussian grid?

    (1-2/2)