Cannot reproduce temperature on pressure levels from ECMWF MARS with CDO
Added by Leopold Haimberger about 5 years ago
ECMWF provides ERA5 temperature data on model as well as pressure levels.
We want to create a local archive where we store spectral and Gaussian grid data on model levels
and from those we want to calculate the pressure level data with CDO.
As a test I tried to reproduce the pressure level data in MARS with CDO. I used the following command:
cdo ml2plx,92500,85000,50000,20000 tgp.grb TGPp.grib
tgp contains model level temperatures on 137 levels on a F320 Gaussian grid, as they are stored in ERA5, plus
the surface geopotential and the logarithm of surface pressure, also on the Gaussian grid from ERA5.
However, when I compare the temperatures at 925 hPa the standard deviation of the difference between the two is 0.5K
I was surprised by this difference. Do you know the reason and how it can be reduced?
With best regards,
Leo Haimberger
Replies (2)
RE: Cannot reproduce temperature on pressure levels from ECMWF MARS with CDO - Added by Uwe Schulzweida about 5 years ago
It seems that only the extrapolated grid cells differ. So I guess the method on how to extrapolate this data is different in ERA5 and CDO. In CDO we use a very old implementation from ECHAM.
Cheers,
Uwe
RE: Cannot reproduce temperature on pressure levels from ECMWF MARS with CDO - Added by Leopold Haimberger about 5 years ago
Ok, thank you.
I assume you do not intend to update this procedure to the one currently used in IFS in the near future.
Extrapolation is questionable anyway, and probably there are subtle differences between the postprocessing packages of the various forecast centres.
I will tell the collegues who need it that they should look into the IFS postprocessing manual to fully reproduce the extrapolated data.
This should be reasonable.
If you intend to update the vertical extrapolation package, however, just let me know.
Best regards,
Leo