WRF Runner

• Run a complete WPS/WRF experiment in one go
• Make a dedicated directory for each experiment
• The original WRF source code is untouched
• Namelists under version control so you can always go back

How to use

It assumes you have a compiled WRF and WPS with the following folder structure

• $HOME/wrf-model/WPS # compiled clone of https://github.com/wrf-model/WRF
• $HOME/wrf-model/WRF # compiled clone of https://github.com/wrf-model/WPS
• $HOME/Urban-M4/misc/wrf-runner # clone of https://github.com/Urban-M4/misc
• $HOME/Urban-M4/experiments # destination for input/output of each experiment

Make sure the paths are set up correctly on your system and match with the paths in wrf_runner.sh. Modify (and commit) the namelists as seen fit, then call the runner:

# Run interactive job
bash wrf_runner.sh

# Or submit as batch job to slurm queue
sbatch wrf_runner.sh

This will create a new folder in $HOME/Urban-M4/experiments/<timestamp> where all output and intermediate files are stored, and run all steps for WPS and WRF.

Notes

f90nml is used to format and modify namelists from the command line. Beware that:

• paths must be in "'double quotes'" (see marshallward/f90nml#126)
• cannot update in place

To do

• Use MODIS_LCZ landuse with wudapt-to-wrf (checkout this branch)
• Use the "standard" heatwave case for amsterdam
• Use IFS analysis for initial/boundary conditions
• Use custom WUR landuse from summer in the city (checkout different branch of WRF?)
• Use custom river temperatures from rijkswaterstaat(?)
• Enable additional custom landuse for building albedo/emissivity
• Add some default plots of input/output for quick inspection of results