Important configuration#

This section provides a set of details for specifying your own configuration. Many details on the options provided by Autosubmit can be found in this Autosubmit guide (in the comments to the configuration files). Also check out the configuration files page for more details on where to find which files.

All of the configuration should be included in the experiment configuraion file <expname>/experiment.yml (link points to conf/art/experiments/<expname>.yml) if not stated differently.

See also

Configuration files, Input data, Building the ICON model code

Timing of model runs#

Timing of model runs in Autosubmit is done with start date and lead time. Hereby, the lead time can be split into multiple chunks (EXPERIMENT.NUMCHUNKS), with a length of (EXPERIMENT.CHUNKSIZE) units of unit size (EXPERIMENT.CHUNKSIZEUNIT). Allowed chunk size units are year, month, day and hour. Furthermore, multiple start dates can be supplied to trigger independent simulations.

An example shall demonstrate this. Here, three independent model runs are set, starting on the first of January, each, for the years 2000, 2010 and 2020. Each model run consists of one member (called “default”), that runs for one year in total by running four chunks with a length of three months each.

EXPERIMENT:
  DATELIST: 20000101 20100101 20200101
  MEMBERS: "default"
  CHUNKSIZEUNIT: month
  CHUNKSIZE: 3
  NUMCHUNKS: 4

Note

Autosubmit also allows for further splitting chunks into so-called splits. This feature is not yet tested for auto-icon.

Specifying directories#

All input data has to be located in one directory and its subdirectories. The output data is then written into a common output directory. In this output directory, there is a subdirectory structure <startdate>/<member>/ and all chunks of one member are stored there. The input and output directories are specified in the DIRECTORIES section with the INDIR and OUTDIR keys.

By default, these are set to

%DIRECTORIES.PREFIX_INDIR%/input/%EXPNAME%
%DIRECTORIES.PREFIX_OUTDIR%/output/%EXPNAME%

Hereby, %EXPNAME% will be substituted with your chosen experiment name and %DIRECTORIES.PREFIX_<DIR>% is defined specifically for the HPC system. The latter is to make it easier if several experiments are run, as the user does not have to change anything importantly in between. However, this comes with some caveat on how this can be defined. For the machines Levante and Horeka, this is defined as follows.

Levante 
PREFIX_INDIR: '/work/%PLATFORMS.LEVANTE.PROJECT%/%PLATFORMS.LEVANTE.USER%'
PREFIX_OUTDIR: '/scratch/%PLATFORMS.LEVANTE.SCRATCH_PROJECT_DIR%/%PLATFORMS.LEVANTE.USER%'
PREFIX_REFDIR: '/scratch/%PLATFORMS.LEVANTE.SCRATCH_PROJECT_DIR%/%PLATFORMS.LEVANTE.USER%'
HoreKa 
WORKSPACE_NAME: 'auto-icon'
PREFIX_INDIR: '/hkfs/work/workspace/scratch/${USER}-%DIRECTORIES.WORKSPACE_NAME%'
PREFIX_OUTDIR: '/hkfs/work/workspace/scratch/${USER}-%DIRECTORIES.WORKSPACE_NAME%'
PREFIX_REFDIR: '/hkfs/work/workspace/scratch/${USER}-%DIRECTORIES.WORKSPACE_NAME%'

This requires a workspace called auto-icon and all input/output data there.
Alternatively, the variable DIRECTORIES.WORKSPACE_NAME in the configuration file %PROJDIR%/conf/common/platforms/HOREKA.yml shall be set to the name of a present workspace.

Domain specific files#

All domain-specific information and input files shall be provided within this section. The FILELIST section holds a list of files to be fetched for each domain (exceptions can be provided). They will automatically be fetched from the respective pool directories, if not present in INDIR.

Tip

Extensive documentation on all the allowed options is provided in the configuration template file conf/art/experiments/template.yml.

An example for a simulation on a global R2B04 grid with reduced radiation grid, using an extpar file and IFS input data can look as follows:

GRID:
  FILELIST:
    - GRID
    - EXTPAR
    - IFS
  RADGRID: True
  DOM1:
    TYPE: G
    R: 2
    B: 4
    GRID_NUMBER: 12
    EXTPAR_DATE: 20131001
    BASENAME:
      IFS: 'uc1_ei_t255_nc_remap_rev832'

Nudging#

Nudging for a LAM or global run can be turned on by setting the variable EXPERIMENT.NUDGING_INTERVAL to a non-zero value. It then contains the interval in seconds between two nudging steps. For global runs, one can additionally select whether to nudge the entire atmosphere or the default range. All further configurations for the nudging have to be provided in the namelist.

EXPERIMENT:
  NUDGING_INTERVAL: 43200   # -- Maximum 1 day = 86400
  NUDGING_RANGE: full       # -- "default": use namelist defaults, "full": use entire atmosphere height

Input data has to be provided according to the format set by the namelist parameters. With the ERA2ICON remapping of input data, also the driving data can be remapped automatically.

Substituting namelist parameters#

If you need to replace/add namelist parameters for all chunks except the first one, you can do so via the EXPERIMENT.NML_SUB_NOT_FIRST parameters, which holds a list of strings. Each string holds a single f90 namelist parameter, formatted as follows:

<group_name>:<parameter>:<value>

e.g. to change the init mode, you could add the following line:

EXPERIMENT:
  NML_SUB_NOT_FIRST:
    - "initicon_nml:init_mode:7"

Reinitialization#

auto-icon supports functionality for experiments with reinitialization. It is turned on by including the configuration file conf/common/jobs/reinit.yml in the include file conf/art.yml (e.g. with the init script). Reinitializations will happen after each chunk and can be configured with the following section:

REINITIALIZATION:
  #-- Continue with meterology, aerosol and/or chemistry from previous run.
  #-- Set to false if you want to reinitialize or do not neet that data.
  CONTINUE_MET: True
  CONTINUE_AERO: True
  CONTINUE_CHEM: False
  #-- Namelist parameters that shall apply to all chunks except the first one.
  #-- Format: yaml list with entries "<namelist name>:<parameter>:<value>"
  REINIT_SUBSTITUTES:
    - "initicon_nml:init_mode:7"
    - "initicon_nml:lread_ana:.false."
    - "initicon_nml:ltile_coldstart:.true."
    - "art_nml:iart_init_aero:5"

Herby, REINIT_SUBSTITUTES follows the same structure as NML_SUB_NOT_FIRST (see above). Initialization of meteorology, aerosol or chemistry data requires the respective output from the previous run. It will be included automatically in the output if the respective CONTINUE_... switch is set to True.

All input data not continued needs to be supplied by the user for each reinit step. Job-specific Autosubmit placeholders like %CHUNK_START_DATE% can be used to get the naming correct for each reinitialization step. An example for this can be found in the REINIT_CHEM template where the namelist option

cart_cheminit_file: 'camchem-data/camchem-%CHUNK_START_YEAR%-%CHUNK_START_MONTH%-%CHUNK_START_DAY%-remapkx.nc'

specifies the file to be used for each chunk. The directory camchem-data (containing all files) is provided as an input directory and is thus linked to the working directory.