Hi, I am trying to run OpenIFS at the ECMWF on a T511L60 reduced Gauss grid initial state that I created from era-interim. I am using 16 MPI tasks. The program aborts with the message: SLCSET: IPERIOD AND YDSL%NDLSUR ARE NOT CONSISTENT. Apparently, setting up the interpolation weights from the radiation grid to the physics grid checks whether NDLSUR-NDLON=3, which is definitely not the case in my NODE.001_01 (NDLON=512,NDLSUR=1026). Is NDLSUR supposed to be this big?

7 Comments

  1. Filip Vana

    Hi Gijs,

    Looks like you are messing up the model grids. The OpenIFS based on CY40R1 is only recognizing the old reduced Gauss grid. So when you are trying to use it with the new files generated with the current octahedral grid, you are necessarily running to problems. Could you please specify how your initial state was created?

    Cheers,

    Filip

  2. Gijs van den Oord

    Thanks, the problem was a mismatch between the gridpoint and the spectral data I downloaded with MARS, where the former should have been on the N256 grid and the latter with resol = 511. Do you by any chance know how to download the spectral representation of the geopotential height zg with MARS? Is it necessary to have it in the ICMSHINIT file for OpenIFS or could I put it in the ICMGGINIT on gridpoints too?

  3. Glenn Carver

    Hi Gijs,

    The model initial files are normally created with a spectral representation of geopotential (surface orography) in the spectral file ICMSH. However, it is possible to put a geopotential field in the gridpoint file (ICMGG). If fields exist in both, the model will take the spectral version as default. I'm not sure if there is a namelist variable to override this (without checking), but one quick way would be to remove the geopotential from the ICMSH file. Note that the field in the ICMGG file will be transformed to spectral space and therefore spectrally filtered for the spectral computations. It would cause problems if not. There is a message in the model output file NODE_001.01 that should say the orography is initialized from a spectral field.

    As for the MARS command, in case you mean, "how to retrieve the geopotential suitable for the model initial file ICMSH", it would be for example:

       retrieve, type=an,levelist=1,step=0,param=129.128,resol=av,target="inishml",repres=sh

    Note the 'resol=av' option. This tells MARS to retrieve at the native stored resolution. Otherwise MARS will do a spectral truncation as part of the retrieve. I think it truncates to T255 by default, which is fine for ERA-Interim, but not for other analyses.

      Cheers,  Glenn

    1. Glenn Carver

      Gijs,

      I am happy to create the initial files you need for you on our system, or I can send you some scripts that might be useful.

      Glenn

  4. Gijs van den Oord

    Hi Glenn,

    Thanks for the suggestions, I can see the message OROGRAPHY INITIALISED FROM SPECTRAL INPUT in my log file. However, a bit later the process aborts because some data (albedos, nonstandard grib codes) cannot be found in the gridpoint init-file:

    GRIDPOINT 2D FIELD MISSING:           66

     GRIDPOINT 2D FIELD MISSING:           67

     GRIDPOINT 2D FIELD MISSING:           43

     GRIDPOINT 2D FIELD MISSING:           15

     GRIDPOINT 2D FIELD MISSING:           16

     GRIDPOINT 2D FIELD MISSING:           17

     GRIDPOINT 2D FIELD MISSING:           18

     ABOR1 CALLED

     I suspect some setting (radiation?) in fort.4 is causing this, as the example model does not contain this data and happily (printing some warnings) carries on without. Perhaps it is the default value for 3D fields that I should adopt in the namelist? Anyway I would like to attach the namelist + NODE.001_01 here, but the popup says I have to ask permissions.

     

    1. Glenn Carver

      Hi Gijs,

      Better to email attachments, we were concerned about the impact of too many large attachments in the forums.

      The model is missing the climatological fields, they come from the ICMCL* file. e.g. grib codes 66 & 67 are leaf-area index (low/high). These fields are created from montly climatologies, interpolated to daily fields.  The University of Oslo have an up-to-date list of the fields in the model initial files here: https://wiki.uio.no/mn/geo/geoit/index.php/OpenIFS#Step-1:_get_meteorological_input_fields (updated from the IFS  Technical manual part 6). I will send you the scripts they are using.

         Glenn

  5. Gijs van den Oord

    Thanks, that would be very helpful. Perhaps it would be a good idea to adopt their setup scripts into the openifs module at ecmwf too?