WRF

Weather Research and Forecasting (WRF) model is a well-known atmospheric modeling system developed by NCAR. It is suitable for both meteorological research and operational weather prediction.

Official website: https://www2.mmm.ucar.edu/wrf/users/

Updated: Aug 2023

Modules

Module name	Description	Note
WRF/4.4.2-DMSM-cpeCray-23.03	Standard WRF model	Aggressive optimization
WRFchem/4.5-DM-cpeIntel-23.03	WRF model with chemistry, including WRF-Chem tools	Standard optimization
WPS/4.4-DM-cpeCray-23.03	WRF pre-processing system	for WRF 4.4.X
WPS/4.5-DM-cpeIntel-23.03	WRF pre-processing system	for WRF 4.5.X

1. Input file

1.1 To run the WRF model, time-dependent meteorological data (global model output/background state) is required. It can be downloaded from, for example, WRF - Free Data and NCEP GFS / GDAS.

1.2 To configure the domain and simulation time, the namelist.wps is needed. A brief description of it can be found here. It is recommended to use the WRF Domain Wizard or the GIS4WRF plug-in for QGIS to define WRF domains.

Some static datasets, such as geog, Global_emissions_v3 and EDGAR, are readily available on LANTA at /project/common/WPS_Static/. They can be utilized by, for example, specifying geo_data_path = '/project/common/WPS_Static/geog' in the namelist.wps.

1.3 To run the WRF model, the namelist.input is required. A concise description of the essential parameters can be found here, while the full description is available in chapter 5 of the WRF user's guide.

2. Job submission script

Below is an example of a WRF submission script. It can be created using vi submitWRF.sh.

To ensure whole node allocation, please verify that 128 = (Number of MPI processes per node) x (Number of OpenMP threads per MPI processes) – this is recommended to avoid potential issues.

#!/bin/bash
#SBATCH -p compute             # Partition
#SBATCH -N 1                   # Number of nodes
#SBATCH --ntasks-per-node=32   # Number of MPI processes per node
#SBATCH --cpus-per-task=4      # Number of OpenMP threads per MPI process
#SBATCH -t 5-00:00:00          # Job runtime limit
#SBATCH -J WRF                 # Job name
#SBATCH -A ltxxxxxx            # Account *** {USER EDIT} *** 

module purge
module load WPS/4.4-DM-cpeCray-23.03
module load WRF/4.4.2-DMSM-cpeCray-23.03

export OMP_STACKSIZE="32M"
export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK}

ulimit -s unlimited

# *** {USER EDIT} *** #
# Please check that namelist.wps and namelist.input are in this same directory
link_grib /--Path-to-your-meteorological-data--/
link_vtable /--Name-of-Vtable-to-parse-the-above-met-data--/

# -- WPS -- #
link_wps
srun -n${SLURM_NTASKS} ./geogrid.exe
srun -N1 -n1 ./ungrib.exe
srun -n${SLURM_NTASKS} ./metgrid.exe
unlink_wps

# -- WRF -- #
link_emreal
srun -n${SLURM_NTASKS} ./real.exe
srun -n${SLURM_NTASKS} -c${SLURM_CPUS_PER_TASK} ./wrf.exe
unlink_emreal

Additional information about ThaiSC support commands for WPS and WRF can be found by using commands such as link_wps --help, man link_emreal or module help WRF.

Some options do not support hybrid run (such as running WRF-Chem). If this is the case, do the following:

use #SBATCH --cpus-per-task=1 and export OMP_NUM_THREADS=1,
increase the total number of tasks, for example, #SBATCH --ntasks-per-node=128
specify the number of tasks for each executable explicitly; for instance, use
srun -n16 ./real.exe
srun -n128 -c1 ./wrf.exe

3. Job submission

To submit jobs to the SLURM queuing system on LANTA, execute

sbatch submitWRF.sh

Two complete examples are available at /project/common/WRF/. To run it within a directory, use cp /project/common/WRF/Example1/* . or cp /project/common/WRF/Example2/* . and follow the instructions in the README file.

Users should check the slurm-xxxx.out file regularly because an abnormal exit from an MPI task/process may NOT cause the entire job to terminate. Setting an appropriate job runtime limit -t is also helpful.

4. Post-processing

NCL, NCO, CDO, netcdf4-python, Ncview, wrf-python, pyngl, pynio and cartopy are available for processing NetCDF files. They are installed in Conda environment netcdf-py39 of Miniconda3/22.11.1-1.

To use NCL, for instance,

module load Miniconda3/22.11.1-1
conda activate netcdf-py39

# For NCL only
export NCARG_ROOT=${CONDA_PREFIX}
export NCARG_RANGS=/project/common/WPS_Static/rangs
export NCARG_SHAPEFILE=/project/common/WPS_Static/shapefile  # (If used)

# Commands such as 'ncl xxx' or 'python xxx' for serial run

Please abstain from doing heavy post-processing tasks on LANTA frontend/login nodes.

5. Advanced topics

Contact Us
ThaiSC support service : thaisc-support@nstda.or.th