CDO

Hi there,

I have a file which is stored in grib format and data is distributed on a reduced Gaussian grid (specifically the file is from the reanalysis model ERA5 of ECMWF).
Now I want to transform the coordinates of this file into a regular Gaussian, therefore I use the following command:
cdo -setgridtype,regular input_era5_reduced output_era5_latlon.grib

So far, it seems the command does the expected job, when the main variable of the grib file is not a 'forecasted' but an 'analysis'.

Everytime I use cdo -setgridtype,regular ... for era5 forecasted variables it seems like it also changes the "step" and initialization time dimensions. Let me be more specific:

Original file ( input_era5_reduced ):

Contains dim "time" with 62 values, one every 12 hours (e.g. '1990-05-01T06:00:00', '1990-05-01T18:00:00', '1990-05-02T06:00:00', etc.)
Contains dim "step" with 12 values, representing 12 hours of forecast (i.e. 1,2,3,4,5,6,7,8,9,10,11,12)

Post CDO file:

Contains dim "time" with a single value, the first of the dataset (i.e. '1990-05-01T06:00:00' etc.)
Contains dim "step" with 410 values, I assume they are different forecast time, so it kinds of reshapes (i.e. 1,2,3,9, 15, ... ,740,744)

I expect the output CDO file to have 743 values in the "step" dimension with an initialization value (suming 744 values, which is 24hours * 31days for the month of April), yet I just get random "steps" suming up to 410 values. What am I doing wrong?

I am using the CDO version 1.9.10 with gcc64.

Thanks a lot for your help!