WRF

From gfi
Revision as of 10:40, 28 February 2024 by Tsp065 (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

This page will feature some help on getting started with WRF at GFI.


Login to Cyclone

Create folder for playing with WRF with initjet

In that folder do: git clone https://git.app.uib.no/Clemens.Spensberger/initjet.git

Go into the initjet folder

Command: module load netCDF-Fortran/4.4.4-foss-2018b OpenBLAS/0.3.1-GCC-7.3.0-2.30

Go into the folder src and type: make main.exe

Using: vi namelist.initJET you can edit the namelist for your WRF simulation

For example: 201 by 101 grid points, 10k by 5k domain, sst_diff = 0, p-pert, P_nx=51, P_ny=51, P_amp=6.d2

If you want to view output from initJET first do: module load ncview

Then you can use ncview to view the initial state

ssh uib-username@fram.sigma2.no

wget https://github.com/wrf-model/WRF/releases/download/v4.5.2/v4.5.2.tar.gz

tar -xzf v4.5.2.tar.gz

module load WRF/4.4-foss-2022a-dmpar (software needed to compile WRF)

export NETCDF=/cluster/software/netCDF-Fortran/4.6.0-gompi-2022a (setting environment variable that points WRF to the right places)

Execute command: ./configure

Correct answer is choice 34 (dmpar with NGU)

And just use basic (default 1)

One can start a terminal-in-terminal with command: tmux (control B and then D detaches terminal. Command to get back: tmux a)

Command to compile: ./compile em_real

Make directory for run: mkdir run

Instead of compiling own WRF, one can use existing copy: cp /cluster/projects/nn8124k/wrf_intro/* run/.

qos gives priority on queue: devel or short give higher priority

Adjust namelist to fit correct grid etc.

Need to also copy met_em files from cyclone to Fram before running WRF

scp ./met_em.d01.2000-01-0*nc username@fram.sigma2.no:~/run/.

In jobscript change: name, account, time, nodes, mail-user, qos

In namelist change:

To run the job: sbatch jobscript.sh

To check the job queue: squeue -u username

You can also check job directly: scontrol show job JOBID

If run crashes, you can look at "slurm" file, alternatively go to run directly that you defined and look at rsl.error files.

To cancel a job: scancel <job-id>