hpc-testkit/WRF-Benchmark-EastAsia3km
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|
Repository files navigation
(1) Introduction This is the benchmark case for a 3km simulation. The simulation area of the WRF model is primarily located between 80E-130E and 15N-55N, which covers most of the East Asia. Files for benchmarking were created with WPSV4.3.1. And they can be used with any V4* WRF code. The benchmark case was a 6 hour simulation from 2024-06-17_00 to 2024-06-17_06 (in UTC). The forcing Data are 0.25 degree ERA5 re-analysis: https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview https://cds.climate.copernicus.eu/datasets/reanalysis-era5-pressure-levels?tab=overview The benchmark test was completed for both an OneAPI/ifx and GNU/gfortran build of WRF. (2) Input Files The benchmark test include the following files: -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- File name | Usage | Site --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- namelist.input | set up to run the 6 hour simulation | this page namelist.wps | the namelist used for the WPS processes | this page diffwrf.py | script to compare two wrfout files | this page met_em.d01.2024-06-17_00_00_00.nc | input files for the real program | https://huggingface.co/datasets/Benchmark-research/wrf-benchmark-eastasia3km/tree/main met_em.d01.2024-06-17_03_00_00.nc | input files for the real program | same as above met_em.d01.2024-06-17_06_00_00.nc | input files for the real program | same as above ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- (3) HOW TO RUN Either build the code with OneAPI or GNU. Make sure the following files are in the test/em_real directory: met_em*nc, and namelist.input. Run real.exe first to generate wrfinput_d01 and wrfbdy_d01. After the successful completion of real.exe, run wrf.exe. When the simulation completes, there should be several wrfout_d01* files. (4) HOW TO COMPARE OUTPUT Make sure python is installed on the system. Issue the command: > python diffwrf.py filename1.nc filename2.nc (modify the command as needed for your comparison) Below is a sample of the differences between a simulation using an OneAPI compiler and a simulation using a GNU compiler. >python diffwrf.py wrfout_d01_2024-06-17_06_gfortran.nc wrfout_d01_2024-06-17_06_ifx.nc Found wrfout_d01_2024-06-17_06_gfortran.nc Found wrfout_d01_2024-06-17_06_ifx.nc Differences of the output from two WRF model simulations for a few important fields Variable Name Minimum Maximum Average Std Dev Skew ========================================================================================= U -31.2153 28.2783 -2.51248e-05 0.260264 0.146371 V -35.8924 39.5658 -4.60322e-05 0.260513 0.144535 W -24.0051 31.4639 -2.96735e-06 0.106599 0.0616054 T -19.269 16.4696 -6.673e-05 0.0878128 -0.693905 PH -489.502 514.829 -0.0256624 3.22575 -0.526888 QVAPOR -0.00764939 0.0081012 3.43118e-08 5.89051e-05 0.112762 TSLB -3.31134 5.33344 1.76842e-05 0.0360554 1.16776 MU -611.835 482.146 0.00453943 5.38994 -0.392063 TSK -23.1584 26.8756 0.000616455 0.508289 -0.0718302 RAINNC -50.9517 37.1058 2.01039e-06 0.4238 -2.34324 Correlations of the output from two WRF model simulations for a few important fields Variable Name Correlation RMSE ======================================================== U 0.999793 1.84029 V 0.999218 1.84211 W 0.929244 0.761044 T 0.999999 0.620747 PH 1 23.0304 QVAPOR 0.999933 0.000416399 TSLB 0.999996 0.0720895 MU 0.999958 5.38836 TSK 0.99911 0.50814 RAINNC 0.996917 0.423676