Customizing WRFtamer

This secion explains how to customize WRFtamer to your needs.

Each experiment created with WRFtamer is configured using the following files:

  1. configure.yaml, where an experiment is configured by defining variables diverging from the namelist.template. This will, most likely, be different for each experiment.
  2. namelist.template, where the standard settings of your namelist are defined. This file will propably not change once it has been adapted to your needs.
  3. submit.template, defining the standard submit file (optional) most likely, you won't have to change this file, once it has been customized for your cluster.

configure.yaml

is a yaml file collecting settings for each experiment. It has five sections:

  • paths
  • namelist_vars
  • submit_file
  • link_grib
  • pp_protocol

In paths, the four main paths are defined, namelist vars lists all lines in your namelist that diverge from the namelist.template, under submit_file, you may change the number of nodes and the walltime for your submit files and link_grib modifies the script that links your gribfile to the wrf-run directory. By default it is set to 3 and you do not need to make any changes. If the number of grib-files you want to link exceeds 26³=17576 files, you need to add another letter to the GRIBFILE.XXX names. This option allows to do that. However, make sure that you make appropriate changes in the WPS/ungrib code as well.

The section pp_protocol defines the postprocessing protocol that should be executed after a run is finished. The execution is done automatically if the watchdog is set or manually via the commandline or the gui.

Example:

pp_protocol:
  move: 1
  tslist_processing:
    timeavg: [5,10]
  create_maps:
    list_of_model_levels: [5]
    list_of_variables: ['WSP']
    list_of_domains: ['d01']
    poi_file: '/path/to/poi_file'
    store: True

Here, data is moved after completition of the run. Then, tslists are processed and time averages over 5 and 10 minutes are calculated. Finally, maps for WSP of domain 1 and the 5th model level are created and stored (as an intermediate file, for dynamic map plots). Set store to false to plot maps as png for static plots.

namelist.template

This is a combined file for namelist.input and namelist.wps. Variables in curly brackets will be set by WRFtamer and need no changes. All the other variables should be set to the values you are using most by default. WRFtamer will compare the variables set in configure.yaml with this template and replace any lines that are different from the template. This approach keeps your configure.yaml fiels as concise as possible.

submit.template

This is a template for a SLURM submit file for your cluster. As of now, WRFtamer is only able to create SLURM submit files, and only with the build-in template.

If there is a demand by the userbase, more options will be added to WRFtamer later on.

Right now, your SLURM template must include the followning lines (unmodified).

#SBATCH --job-name={name}
#SBATCH --output={slurm_log}
#SBATCH -N {nodes}
#SBATCH --time={time}

Optional environmental variables

This program uses six environmental variables to set important paths and options. Setting these variables is optional, as default default directories are defined or variables are not used if they are not set.

env variable default
WRFTAMER_HOME_PATH $HOME/wrftamer
WRFTAMER_RUN_PATH $HOME/wrftamer/run
WRFTAMER_ARCHIVE_PATH $HOME/wrftamer/archive
WRFTAMER_make_submit False
OBSERVATIONS_PATH N/A
WRFTAMER_DEFAULT_POI_FILE N/A

  • The envirnomental variable OBSERVATIONS_PATH must point to the directory with the observations you want to display with WRFplotter. See WRFplotter.

  • WRFTAMER_DEFAULT_POI_FILE is used to add points of interest to your map plots It must point to a csv file with the following structure:

Name;lat;lon
Station1; 54.015; 8.545
Station2; 53.154; 9.123
...

It will be used by map plots of WRFplotter. If this variable is not set, no points will be plotted.

Setting environmental variables with conda

You may set and unset these variables in the environment. If these files don't exist, just create them.

In the file: ~/.conda/envs/<env_name>/etc/conda/activate.d/env_vars.sh put the lines

set env WRFTAMER_HOME_PATH <your/desired/path/>
set env WRFTAMER_RUN_PATH <your/desired/path/>
set env WRFTAMER_ARCHIVE_PATH <your/desired/path/>
set env WRFTAMER_make_submit <False or True>
set env OBSERVATIONS_PATH <your/desired/path/>
set env WRFTAMER_DEFAULT_POI_FILE </absolute/path/to/poi_file.csv>

In the file ~/.conda/envs/<env_name>/etc/conda/deactivate.d/env_vars.sh add

unset env WRFTAMER_HOME_PATH
unset env WRFTAMER_RUN_PATH
unset env WRFTAMER_ARCHIVE_PATH
unset env WRFTAMER_make_submit
unset env OBSERVATIONS_PATH
unset env WRFTAMER_DEFAULT_POI_FILE

This way, these variables are set each time you load your conda environment and are unset on deactivation.