FAQ

FAQ#

How about if program hangs?#

If the program stalls or appears to stop without producing output, try the following:

  • Rebuild with the Intel toolchain instead of GNU when possible. In practice, the Intel build is often more reliable for this workflow: ./build.sh -p false -c intel.

  • Load the matching environment before running the executable:

source load_modules_intel.sh

  • Enable detailed logging in system.nml:

verbose = .true.

  • Check that the paths in system.nml and the model file list are valid and that the raw input files exist.

  • If needed, reduce the task scope to a small subset of variables to isolate the problem.

How to specify the time length of CMORized file?#

The number of records in the cmorized files are determined by:

  • the selected time period in experiment.nml (year1, yearn, month1, monthn)

  • the output frequency implied by the variable name and CMIP7 data request

  • the native time axis of the raw model output

The CMORized output will contain the number of records (depending on the raw model output frequency, i.e., daily, monthly, or yearly) between year1, month1 and yearn, monthn. Adjust the temporal settings in experiment.nml to specify the number of records in the CMORized files.

Important

When the length of the records is more than 10 years, the cmor4cmip7 program by default automatically close the NetCDF file and start a new NetCDF file.

For example, if the experiment.nml is prescribed as:

 ...
 year1                  = 1346,
 yearn                  = 1395,
 month1                 = 1,
 monthn                 = 12,
 ...

The first file will have the time stamp as *_134601..-135512.nc, and then another file from *_135601-136512.nc and so on.

If one would like to have first file as *_134601-134912.nc and the second one *_135001-135912.nc and so on, one can modify the experiment.nml, launch the cmor4cmip7, and when it finishes the first 5 years, update the year1 and yearn and launch the program again.

How do I add or customize a dataset?#

There are three standard ways to do this:

  1. Add or remove entries in variables.nml.

    • Each line in compound_names is a CMIP7 branded variable.

    • Comment out a variable with ! to skip it.

&variables
    compound_names =
        'ocean.zos.tavg-u-hxy-sea.mon.glb',
        'ocean.agessc.tavg-ol-hxy-sea.mon.glb'
/

  1. Update the mapping information in mapping.json if the variable or metadata needs a custom translation from the model output to the CMIP7 variable definition.

  2. Adjust the model/experiment metadata in model.nml and experiment.nml if the source ID, experiment label, grid label, or branch information needs to be changed.

  3. If the variable/dataset is not a standard CMIP7 dataset, you need to hack the corresponding CMIP7 json table file, e.g., CMIP7_ocean.json to make a dataset entry as the same compound name in the variables.nml and mapping.json.

A good workflow is:

  • add the variable to variables.nml.

  • confirm the metadata in model.nml and experiment.nml.

  • check the mapping in mapping.json.

  • hack the CMIP7 CV json file (CMIP7_*.json), if necessary.

  • rerun the program with verbose = .true. to inspect the generated CMIP7 metadata and output structure.