Wrong time axis when converting from GRIB to netCDF

Added by Joakim Kjellsson about 7 years ago

I've got some output from the OpenIFS model, in grib format. The output is mixed grib1/grib2, but I've used grib_set -s editionNumber=1 to convert all to grib1.
When I use CDO to convert from spectral to grid-point space, and from grib to netcdf, I sometimes don't get the right time axis. It works for 95% of the files, but not for some, and I have no idea why.

I'm enclosing the two grib files "wrong_time.grb" and "right_time.grb". They are outputs from hours 3066 and 3072 (i.e. 127 days 18hours and 128 days into the run).
If I do

cdo -t ecmwf -f nc sp2gpl right_time.grb right_time.nc
ncdump -v time right_time.nc

I see that the time is 3066 hours since 1982-01-01, which is correct.
However, if I do

cdo -t ecmwf -f nc sp2gpl wrong_time.grb wrong_time.nc
ncdump -v time wrong_time.nc

I find that it's 128 days since 1982-01-01, which is correct, but why is CDO listing time in units "days since 1982-01-01" in this file, but in almost all other files it has units "hours since 1982-01-01"?
Is there a setting that I'm missing?
I've tried with CDO version 1.9.2 on our cluster and version 1.9.0 on my laptop with the same results.

Best wishes
Joakim

Download all files

right_time.grb (13.9 MB) right_time.grb
wrong_time.grb (13.9 MB) wrong_time.grb
right_time.nc (53.9 MB) right_time.nc
wrong_time.nc (53.9 MB) wrong_time.nc

Replies (2)

RE: Wrong time axis when converting from GRIB to netCDF - Added by Karin Meier-Fleischer about 7 years ago

Hi Joakim,

the GRIB files have different reference time units.

cdo sinfo right_time.grb

   File format : GRIB
    -1 : Institut Source   T Steptype Levels Num    Points Num Dtype : Parameter ID
     1 : ECMWF    unknown  v instant       1   1     25760   1  P16  : 152.128       
     2 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 131.128       
     3 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 132.128       
     4 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 130.128       
     5 : ECMWF    unknown  v instant       1   3     25760   1  P16  : 152.128       
     6 : ECMWF    unknown  v instant       1   3     25760   1  P16  : 129.128       
   Grid coordinates :
     1 : spectral                 : points=25760  nsp=12880  truncation=159  complexPacking
   Vertical coordinates :
     1 : pressure                 : levels=1
                             plev : 100000 
     2 : hybrid                   : levels=91
                              lev : 1 to 91 by 1 
                        available : vct
     3 : hybrid                   : levels=1
                              lev : 1 
                        available : vct
   Time coordinate :  1 step
     RefTime =  1982-01-01 00:00:00  Units = hours  Calendar = proleptic_gregorian
  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss
  1982-05-08 18:00:00
cdo sinfo: Processed 6 variables over 1 timestep ( 0.01s 13MB )

Have a look at the line: RefTime = 1982-01-01 00:00:00 Units = hours Calendar = proleptic_gregorian

cdo sinfo wrong_time.grb

   File format : GRIB
    -1 : Institut Source   T Steptype Levels Num    Points Num Dtype : Parameter ID
     1 : ECMWF    unknown  v instant       1   1     25760   1  P16  : 152.128       
     2 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 131.128       
     3 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 132.128       
     4 : ECMWF    unknown  v instant      91   2     25760   1  P16  : 130.128       
     5 : ECMWF    unknown  v instant       1   3     25760   1  P16  : 152.128       
     6 : ECMWF    unknown  v instant       1   3     25760   1  P16  : 129.128       
   Grid coordinates :
     1 : spectral                 : points=25760  nsp=12880  truncation=159  complexPacking
   Vertical coordinates :
     1 : pressure                 : levels=1
                             plev : 100000 
     2 : hybrid                   : levels=91
                              lev : 1 to 91 by 1 
                        available : vct
     3 : hybrid                   : levels=1
                              lev : 1 
                        available : vct
   Time coordinate :  1 step
     RefTime =  1982-01-01 00:00:00  Units = days  Calendar = proleptic_gregorian
  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss  YYYY-MM-DD hh:mm:ss
  1982-05-09 00:00:00
cdo sinfo: Processed 6 variables over 1 timestep ( 0.01s 13MB )

Have a look at the line: RefTime = 1982-01-01 00:00:00 Units = days Calendar = proleptic_gregorian

You can change the reference time unit with the settaxis operator. E.g.

cdo -settaxis,1982-01-01,00:00:00,1hour wrong_time.grb ex_wrong_time.grb

-Karin

RE: Wrong time axis when converting from GRIB to netCDF - Added by Joakim Kjellsson about 7 years ago

Dear Karin

Thanks for your reply, and for locating the issue in the GRIB files.
The fix you posted did not work for me, but I found something similar that does:

cdo -settunits,hours wrong_time.grb ex_wrong_time.grb

This converts 128 days since 1982-01-01 to 3072 hours since 1982-01-01. Heureka!

Thanks for your help!
/Joakim

(1-2/2)

Project

General

Profile

CDO