CDO

I got this

am@luthien:~/Downloads cdo -V                                                                                                                                                                                             [16:35:30|16-05-12]
Climate Data Operators version 1.7.1 (http://mpimet.mpg.de/cdo)
Compiled: by ram on luthien (x86_64-unknown-linux-gnu) May  6 2016 16:32:01
Compiler: gcc -g -O3 -std=gnu99 -Wall -fopenmp -march=native   
 version: gcc (GCC) 5.3.0
Features: DATA PTHREADS OpenMP4 HDF5 NC4/HDF5 OPeNDAP SZ Z UDUNITS2 PROJ.4 CURL FFTW3 AVX2
Libraries: HDF5/1.10.0 proj/4.92 curl/7.48.0
Filetypes: srv ext ieg grb grb2 nc nc2 nc4 nc4c 
     CDI library version : 1.7.1 of May  6 2016 16:31:43
 CGRIBEX library version : 1.7.4 of Feb 19 2016 11:03:41
GRIB_API library version : 1.14.2
  NetCDF library version : 4.4.1-rc1 of Apr 20 2016 12:06:30 $
    HDF5 library version : 1.10.0
 SERVICE library version : 1.4.0 of May  6 2016 16:31:41
   EXTRA library version : 1.4.0 of May  6 2016 16:31:40
     IEG library version : 1.4.0 of May  6 2016 16:31:40
    FILE library version : 1.8.2 of May  6 2016 16:31:40

ram@luthien:~/Downloads cdo -sinfov -uv2dv atm_amip_ctl_atm_3d_ml_1979-2008_ymonmean_ua_m+va_m+pfull_m_timestep1.nc                                                                                                        [16:35:35|16-05-12]
cdo sinfon: Started child process "uv2dv atm_amip_ctl_atm_3d_ml_1979-2008_ymonmean_ua_m+va_m+pfull_m_timestep1.nc (pipe1.1)".
cdo(2) uv2dv (Warning): U-wind not found!
cdo(2) uv2dv (Warning): V-wind not found!
   File format : NetCDF2
    -1 : Institut Source   Steptype Levels Num    Points Num Dtype : Parameter name
     1 : MPIMET   https://svn instant      47   1     20480   1  F32  : u             
     2 : MPIMET   https://svn instant      47   1     20480   1  F32  : v             
     3 : MPIMET   https://svn instant      47   1     20480   1  F32  : pfull_m       
   Grid coordinates :
     1 : unstructured             : points=20480  nvertex=3
                             grid : number=42  position=1
                             clon : -3.14159 to 3.14159 radian
                             clat : -1.55192 to 1.55192 radian
                        available : cellbounds
                             uuid : f25aa6bc-809a-11e2-839d-15e92cba77d6
   Vertical coordinates :
     1 : generic                  : levels=47
                           height : 1 to 47 by 1 
                           bounds : 1-2 to 47-48 by 1 
   Time coordinate :  1 step
  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss
  2008-01-01 00:00:00
cdo(2) uv2dv: Processed 2887680 values from 3 variables over 1 timestep ( 0.00s )
cdo sinfon: Processed 3 variables over 1 timestep ( 0.00s )
ram@luthien:~/Downloads cdo -sinfov -uv2dv ERA-Interim_monthly-means-of-daily-means_u+v-wind_1979-2008_ymonmean_ICON.nc                                                                                                    [16:35:58|16-05-12]
cdo sinfon: Started child process "uv2dv ERA-Interim_monthly-means-of-daily-means_u+v-wind_1979-2008_ymonmean_ICON.nc (pipe1.1)".

cdo(2) uv2dv (Abort): U-wind is not on Gaussian grid!                                                                                              

Both do not work!!

Where/How did you get it running?

Sorry, I attached the wrong file.
I mapped ERA-Interim_monthly-means-of-daily-means_u+v-wind_1979-2008_timestep1.nc with 'cdo remapcon,n128' before calculating divergence.
Without remapping I get 'U-wind is not on Gaussian grid!', too.
But remapping before calculating divergence does not work with ERA-Interim_monthly-means-of-daily-means_u+v-wind_1979-2008_ymonmean_ICON.nc and with atm_amip_ctl_atm_3d_ml_1979-2008_ymonmean_ua_m+va_m+pfull_m_timestep1.nc.
Is it because of the ICON grid?

I think so, divergence is well defined on a rectangular grid, but not on a triangular one. But ICON has divergence operators - you could compute it internally and store it as regular output.

AFAIK it is already there: div

see https://code.zmaw.de/projects/icon/repository/entry/trunk/icon-dev/src/atm_dyn_iconam/mo_nonhydro_state.f90#L1642

Thanks. But I have finished all of my experiments and don't want to start each of them again only to get 'div'. That would take some weeks. Is there an alternative to calculate divergence on the triangular ICON grid? Why doesn't it work if I change variable names to 'u' and 'v' and remap it to an rectangular grid?

Hi Ludwig!

AFAIK the velocity in ICON is a single field vn which stores the projection of the real velocity vector onto the edge-normal vector of the edge where it is located. This means, that it makes heavy use of the ICON grid structure. without the knowledge of the edge normal direction, you cannot make anything reasonable out of vn itself. That's why u and v velocities are computed internally (knowing all the grid stuff ...) and put on the cell center.

You possibly could remap theses u and v onto something regular and compute the divergence there, but I doubt that this makes sense. You might ask other ICON users/developers who are more familiar with divergence.

One note on recomputing: If you have restart or check point files, you can compute divergence for all checkpoint intervals in parallel - so it might not take weeks but only a day or 2.

Thanks a lot :)
That should solve my problem.