Show pageOld revisionsBack to top This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong. ====== WRF on DARWIN====== Currently WRF is not installed on DARWIN by the system administrators. It can be installed by users for personal use or for use by their workgroups by following the below directions. These directions will walk users through all the steps required to install and compile WRF and WPS. It will also explain how to add the WRF installation into your home or workgroup's VALET directory. Lastly, these steps will walk you through a trial run. These directions are loosely based on the directions from NCAR's [[https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php|"How to compile WRF: The Complete Process"]] page. Our directions differ from NCAR's in order to get better performance for WRF on DARWIN. The main difference is the default MPI software WRF leverages, MPICH, whereas Open MPI is the MPI software tightly-integrated with Slurm on DARWIN. These directions will take you through the steps of installing WRF to use Open MPI instead of MPICH for parallel processing. ===== Installing WRF ===== The directions below will walk you through the download, installing, compiling WRF, and creating a WRF VALET package. **Make sure you follow the directions CAREFULLY and make the necessary changes to reflect your personal environment**. The directions and examples below are based on user ''traine'' in workgroup ''it_css''. You will need to change all references to reflect your username and workgroup. ==== Set up the installation environment ==== Start by joining your workgroup <code bash> [traine@login00.darwin ~]$ workgroup -g it_css [(it_css:traine)@login00.darwin ~]$ </code> The next commands load the required VALET packages and set the necessary environment variables. Use the VALET commmand ''vpkg_require'' to load the appropriate version of the ''openmpi'' package. <code bash> [(it_css:traine)@login00.darwin ~]$ vpkg_require openmpi/4.0.5:intel-2020 Adding dependency `intel/2020u4` to your environment Adding package `openmpi/4.0.5:intel-2020` to your environment </code> Set the environment variables needed for compiling and installing the WRF packages. <code bash> [(it_css:traine)@login00.darwin ~]$ WRF_PREFIX=$WORKDIR/sw/wrf/20210121 ## MAKE SURE THAT THIS IS CHANGED TO YOUR WORK GROUP DIRECTORY OR HOME DIRECTORY ## [(it_css:traine)@login00.darwin ~]$ mkdir -p "$WRF_PREFIX" [(it_css:traine)@login00.darwin ~]$ WRF_SRC="${WRF_PREFIX}/src" [(it_css:traine)@login00.darwin ~]$ WRF_BIN="${WRF_PREFIX}/bin" [(it_css:traine)@login00.darwin ~]$ WRF_INC="${WRF_PREFIX}/include" [(it_css:traine)@login00.darwin ~]$ WRF_LIB="${WRF_PREFIX}/lib" [(it_css:traine)@login00.darwin ~]$ WRF_LIBRARIES_SRC="${WRF_SRC}/LIBRARIES" [(it_css:traine)@login00.darwin ~]$ mkdir -p "$WRF_LIBRARIES_SRC" [(it_css:traine)@login00.darwin ~]$ WRF_TESTS_SRC="${WRF_SRC}/TESTS" [(it_css:traine)@login00.darwin ~]$ mkdir -p "$WRF_TESTS_SRC" </code> <code bash> [(it_css:traine)@login00.darwin ~]$ export CC=icc [(it_css:traine)@login00.darwin ~]$ export CFLAGS="-xHost" [(it_css:traine)@login00.darwin ~]$ export CXX=icpc [(it_css:traine)@login00.darwin ~]$ export CXXFLAGS="-xHost" [(it_css:traine)@login00.darwin ~]$ export CPPFLAGS="-I${WRF_INC}" [(it_css:traine)@login00.darwin ~]$ export FC=ifort [(it_css:traine)@login00.darwin ~]$ export FCFLAGS="-I${WRF_INC} -xHost" [(it_css:traine)@login00.darwin ~]$ export F77="$FC" [(it_css:traine)@login00.darwin ~]$ export LDFLAGS="-L${WRF_LIB}" [(it_css:traine)@login00.darwin ~]$ export PATH="${WRF_BIN}:$PATH" [(it_css:traine)@login00.darwin ~]$ export MPICC="mpicc" [(it_css:traine)@login00.darwin ~]$ export MPICXX="mpicxx" [(it_css:traine)@login00.darwin ~]$ export MPIFC="mpifort" </code> ==== Download and install required libraries ==== WRF requires several libraries for the installation. First, we use the environment variable ''WRF_LIBRARIES_SRC'' to change into the ''LIBRARIES'' directory, then we will use the ''wget'' command to download the libraries, and finally use the appropriate commands to install each library. === Download libraries === <code bash> [(it_css:traine)@login00.darwin ~]$ cd "$WRF_LIBRARIES_SRC" [(it_css:traine)@login00.darwin LIBRARIES]$ wget https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compile_tutorial/tar_files/netcdf-4.1.3.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ wget https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compile_tutorial/tar_files/jasper-1.900.1.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ wget https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compile_tutorial/tar_files/libpng-1.2.50.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ wget https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compile_tutorial/tar_files/zlib-1.2.7.tar.gz </code> === Install netcdf library === <code bash> [(it_css:traine)@login00.darwin LIBRARIES]$ tar -xf netcdf-4.1.3.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ cd netcdf-4.1.3 [(it_css:traine)@login00.darwin netcdf-4.1.3]$ ./configure --prefix="$WRF_PREFIX" --disable-dap --disable-netcdf-4 --disable-shared [(it_css:traine)@login00.darwin netcdf-4.1.3]$ make -j 4 [(it_css:traine)@login00.darwin netcdf-4.1.3]$ make install [(it_css:traine)@login00.darwin netcdf-4.1.3]$ cd .. </code> === Install zlib library === <code bash> [(it_css:traine)@login00.darwin LIBRARIES]$ tar -xf zlib-1.2.7.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ cd zlib-1.2.7 [(it_css:traine)@login00.darwin zlib-1.2.7]$ ./configure --prefix="$WRF_PREFIX" [(it_css:traine)@login00.darwin zlib-1.2.7]$ make [(it_css:traine)@login00.darwin zlib-1.2.7]$ make install [(it_css:traine)@login00.darwin zlib-1.2.7]$ cd .. </code> === Install libpng library === <code bash> [(it_css:traine)@login00.darwin LIBRARIES]$ tar -xf libpng-1.2.50.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ cd libpng-1.2.50 [(it_css:traine)@login00.darwin libpng-1.2.50]$ ./configure --prefix="$WRF_PREFIX" [(it_css:traine)@login00.darwin libpng-1.2.50]$ make -j 4 [(it_css:traine)@login00.darwin libpng-1.2.50]$ make install [(it_css:traine)@login00.darwin libpng-1.2.50]$ cd .. </code> === Install jasper library === <code bash> [(it_css:traine)@login00.darwin LIBRARIES]$ tar -xf jasper-1.900.1.tar.gz [(it_css:traine)@login00.darwin LIBRARIES]$ cd jasper-1.900.1 [(it_css:traine)@login00.darwin jasper-1.900.1]$ ./configure --prefix="$WRF_PREFIX" [(it_css:traine)@login00.darwin jasper-1.900.1]$ make -j 4 [(it_css:traine)@login00.darwin jasper-1.900.1]$ make install [(it_css:traine)@login00.darwin jasper-1.900.1]$ cd .. </code> ==== Test library compatibility ==== Use the environment variable ''WRF_TEST_SRC'' to change into the ''TESTS'' directory. Next download the test file, compile them using the libraries you installed and verify the results. Remember all executables should typically be run on compute nodes. However these tests require compiling which must be done either on the login (head) node or by using ''salloc %%--%%partition=devel'' to run on compute nodes, and since these tests will run quickly and only use minimal resources to verify the correct installation of the libraries, we will use the login (head) node for this demonstration. <code bash> [(it_css:traine)@login00.darwin LIBRARIES]$ cd "$WRF_TESTS_SRC" [(it_css:traine)@login00.darwin TESTS]$ </code> === Download and extract test file === <code bash> [(it_css:traine)@login00.darwin TESTS]$ wget https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compile_tutorial/tar_files/Fortran_C_NETCDF_MPI_tests.tar [(it_css:traine)@login00.darwin TESTS]$ tar -xf Fortran_C_NETCDF_MPI_tests.tar </code> === Compile and run test (serial) === <code bash> [(it_css:traine)@login00.darwin TESTS]$ $FC $FCFLAGS -c 01_fortran+c+netcdf_f.f $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $CC $CPPFLAGS -c 01_fortran+c+netcdf_c.c $CFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $FC -c 01_fortran+c+netcdf_f.f $FCFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $CC $CPPFLAGS -c 01_fortran+c+netcdf_c.c $CFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $FC 01_fortran+c+netcdf_f.o 01_fortran+c+netcdf_c.o $FCFLAGS $LDFLAGS -lnetcdff -lnetcdf [(it_css:traine)@login00.darwin TESTS]$ ./a.out </code> <note tip>Verify your test results. This can be done by comparing them to the test results on the NCAR site. The results are in the [[https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php#STEP3| Library Compatibility Tests]] section **Test #1**</note> === Compile and run test (MPI) === <code bash> [(it_css:traine)@login00.darwin TESTS]$ $MPICC $CPPFLAGS -c 02_fortran+c+netcdf+mpi_c.c $CFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $MPIFC -c 02_fortran+c+netcdf+mpi_f.f $FCFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $MPICC $CPPFLAGS -c 02_fortran+c+netcdf+mpi_c.c $CFLAGS $LDFLAGS [(it_css:traine)@login00.darwin TESTS]$ $MPIFC 02_fortran+c+netcdf+mpi_f.o 02_fortran+c+netcdf+mpi_c.o $FCFLAGS $LDFLAGS -lnetcdff -lnetcdf [(it_css:traine)@login00.darwin TESTS]$ mpirun -np 2 ./a.out </code> <note tip>Verify your test results. This can be done by comparing them to the test results on the NCAR site. The results are in the [[https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php#STEP3| Library Compatibility Tests]] section: **Test #2** </note> ==== Build WRF ==== === Download WRF source code === <code bash> [(it_css:traine)@login00.darwin TESTS]$ cd "$WRF_SRC" [(it_css:traine)@login00.darwin src]$ wget https://www2.mmm.ucar.edu/wrf/src/WRFV4.0.TAR.gz [(it_css:traine)@login00.darwin src]$ tar -xf WRFV4.0.TAR.gz [(it_css:traine)@login00.darwin src]$ cd WRF [(it_css:traine)@login00.darwin WRF]$ </code> <note important>The compile steps can take quite a long time to complete. Use caution and make sure that you select the correct options as directed below. </note> === Configure WRF serial build == Configuring serial build. You will want to select option ''64'' and nesting option ''0''. <code bash> [(it_css:traine)@login00.darwin WRF]$ export NETCDF="$WRF_PREFIX" [(it_css:traine)@login00.darwin WRF]$ ./configure [(it_css:traine)@login00.darwin WRF]$ ./compile -j 4 em_real [(it_css:traine)@login00.darwin WRF]$ mkdir -p "$WRF_BIN" [(it_css:traine)@login00.darwin WRF]$ install --target-directory="$WRF_BIN" --mode=0775 main/*.exe </code> === Configure WRF parallel build === Configuring parallel build. You will want to select option ''66'' and nesting option ''1''. <code bash> [(it_css:traine)@login00.darwin WRF]$ ./clean -a [(it_css:traine)@login00.darwin WRF]$ ./configure </code> === Apply required patch for OpenMPI flags === The patch command with the use of ''EOT'' does not display the bash prompt after it is entered, but instead ''>'', because it is waiting for the patch lines to be entered at the terminal until ''EOT'' is typed. **You may find it best to copy and paste the lines below to avoid making any typing mistakes as spacing is very important and it must be typed exactly done this way.**. <code bash> [(it_css:traine)@login00.darwin WRF]$ patch -p1 <<EOT --- A/configure.wrf 2020-12-10 14:06:01.907649095 -0500 +++ B/configure.wrf 2020-12-10 14:40:00.791338460 -0500 @@ -118,8 +118,8 @@ SFC = ifort SCC = icc CCOMP = icc -DM_FC = mpif90 -f90=\$(SFC) -DM_CC = mpicc -cc=\$(SCC) +DM_FC = mpif90 +DM_CC = mpicc FC = time \$(DM_FC) CC = \$(DM_CC) -DFSEEKO64_OK LD = \$(FC) @@ -140,7 +140,7 @@ BYTESWAPIO = -convert big_endian RECORDLENGTH = -assume byterecl FCBASEOPTS_NO_G = -ip -fp-model precise -w -ftz -align all -fno-alias \$(FORMAT_FREE) \$(BYTESWAPIO) -xHost -fp-model fast=2 -no-heap-arrays -no-prec-div -no-prec-sqrt -fno-common -maxvx2 -FCBASEOPTS = \$(FCBASEOPTS_NO_G) \$(FCDEBUG) +FCBASEOPTS = \$(FCBASEOPTS_NO_G) \$(FCDEBUG) -assume nocc_omp MODULE_SRCH_FLAG = TRADFLAG = -traditional-cpp CPP = /lib/cpp -P -nostdinc EOT </code> === Compile WRF parallel build === <code bash> [(it_css:traine)@login00.darwin WRF]$ ./compile -j 4 em_real [(it_css:traine)@login00.darwin WRF]$ mkdir -p "$WRF_BIN" [(it_css:traine)@login00.darwin WRF]$ for exe in main/*.exe; do # Install each exe to WRF_BIN with the prefix "mpi_" on it # to differentiate from the serial variants: WRF_ROOT="$(echo $exe | sed -e 's/main\///')" install --mode=0775 "$exe" "${WRF_BIN}/mpi_${WRF_ROOT}" done [(it_css:traine)@login00.darwin WRF]$ </code> ==== Build WPS ==== === Download WPS source code === <code bash> [(it_css:traine)@login00.darwin WRF]$ cd "$WRF_SRC" [(it_css:traine)@login00.darwin src]$ wget https://www2.mmm.ucar.edu/wrf/src/WPSV4.0.TAR.gz [(it_css:traine)@login00.darwin src]$ tar -xf WPSV4.0.TAR.gz [(it_css:traine)@login00.darwin src]$ cd WPS [(it_css:traine)@login00.darwin WPS]$ </code> <note important> * WPS unlike WRF only has a serial build and currently does not have parallel build. * The compile steps can take quite a long time to complete. Use caution and make sure that you select the correct options as directed below. </note> === Configure and compile WPS serial build == Configure serial build. You will want to select option ''17''. <code bash> [(it_css:traine)@login00.darwin WPS]$ export JASPERLIB="$WRF_LIB" [(it_css:traine)@login00.darwin WPS]$ export JASPERINC="$WRF_INC" [(it_css:traine)@login00.darwin WPS]$ ./configure [(it_css:traine)@login00.darwin WPS]$ ./compile [(it_css:traine)@login00.darwin WPS]$ install --target-directory="$WRF_BIN" --mode=0775 *.exe </code> Now select option ''19'' <code bash> [(it_css:traine)@login00.darwin WPS]$ ./clean -a [(it_css:traine)@login00.darwin WPS]$ ./configure [(it_css:traine)@login00.darwin WPS]$ ./compile [(it_css:traine)@login00.darwin WPS]$ for exe in *.exe; do # Install each exe to WRF_BIN with the prefix "mpi_" on it # to differentiate from the serial variants: install --mode=0775 "$exe" "${WRF_BIN}/mpi_${exe}" done </code> ==== Create a VALET package for WRF ==== If you are planning on being the only person running WRF or are installing a custom version of WRF it is suggested that you add the WRF VALET package to the ''$HOME/.valet'' directory. If you are installing WRF and planning for multiple people in your workgroup to run it, then you should add the WRF VALET package to the ''$WORKDIR/sw/valet'' directory. You **do not** have to add it in both places. <note warning> If you have done this before then you just need to append the new version to the extant VALET package definition file.</note> === Adding WRF VALET package in HOME directory === <code bash> [(it_css:traine)@login00.darwin WRF]$ mkdir -p ~/.valet #This step is only required if you don't have a .valet directory already created [(it_css:traine)@login00.darwin WRF]$ cat <<EOT > ~/.valet/wrf.vpkg_yaml # # WRF - Weather Research and Forecasting model # wrf: prefix: $(dirname "$WRF_PREFIX") description: Weather Research and Forecasting model url: "https://www.mmm.ucar.edu/weather-research-and-forecasting-model" versions: "$(basename "$WRF_PREFIX")": description: 2020-12-10 build with Intel compilers and Open MPI dependencies: - openmpi/4.0.5:intel-2020 EOT [(it_css:traine)@login00.darwin WRF]$ </code> === Adding WRF VALET package in workgroup directory === As mentioned at the top of this page these directions are based on user ''traine'' in workgroup ''it_css''. In these next directions we are going to add the WRF VALET package in the ''it_css'' workgroup directory. The location of the VALET directory might differ in your workgroup directory, or you might not yet have a VALET directory in your workgroup. If you do not have a valet directory, we suggest that you follow the directory structure we use in the ''it_css'' workgroup. Not only will it make it easier to follow these directions, but it will also make it easy to add more software packages and versions in the future. <code bash> [(it_css:traine)@login00.darwin ~]$ cd $WORKDIR/sw [(it_css:traine)@login00.darwin sw]$ mkdir -p valet #This step is only required if you don't have a valet directory already created [(it_css:traine)@login00.darwin sw]$ cat <<EOT > valet/wrf.vpkg_yaml # # WRF - Weather Research and Forecasting model # wrf: prefix: $(dirname "$WRF_PREFIX") description: Weather Research and Forecasting model url: "https://www.mmm.ucar.edu/weather-research-and-forecasting-model" versions: "$(basename "$WRF_PREFIX")": description: 2020-12-10 build with Intel compilers and Open MPI dependencies: - openmpi/4.0.5:intel-2020 EOT [(it_css:traine)@login00.darwin ~]$ </code> ==== Check installation ==== If all the prior steps were followed and successfully completed you are now ready to check if you can successfully load WRF with VALET and then find the WRF executables. === Clear environment and load WRF VALET package === <code bash> [(it_css:traine)@login00.darwin ~]$ vpkg_rollback all [(it_css:traine)@login00.darwin ~]$ vpkg_require wrf Adding dependency `intel/2020u4` to your environment Adding dependency `openmpi/4.0.5:intel-2020` to your environment Adding package `wrf/20210113` to your environment </code> === Check WRF executables === <code bash> [(it_css:traine)@login00.darwin ~]$ which real.exe /work/it_css/sw/WRF/20210113/bin/real.exe [(it_css:traine)@login00.darwin ~]$ which mpi_real.exe /work/it_css/sw/WRF/20210113/bin/mpi_real.exe </code> ===== Run WRF and WPS outside SRC directories ===== By default WRF and WPS are setup to run in their respected ''SRC'' directories. This is not the best idea and can cause serious problems to the WRF installation and also block the WRF software from being run by multiple people at the same time. The following steps will show you how to set up a safe environment to run WRF and WPS from a directory outside the ''SRC'' directories. ==== Getting the example_wrf.tar file ==== By default WPS/WRF is set up to run from it's installed ''SRC'' directory. This is not ideal, so running WPS/WRF from another directory outside it's install directory is much safer and can help protect against accidental modifications or deletions of the originally installed source files. To set up the WPS and WRF to run their respected programs outside their installed ''SRC'' directory we have created an example to test called ''example_wrf.tar'' available in ''trainf'''s home directory to copy into the directory location in which you plan to download the data files and run all the WRF programs. For this demo, the ''example_wrf.tar'' file will be copied into the location ''$WORKDIR/traine/wrf''. <code bash> [(it_css:traine)@login00.darwin wrf]$ cp /home/1200/wrf/example_wrf.tar . </code> ==== Set up WPS ==== Now that you have the ''example_wrf.tar'' file where you want it, now we can extract it and and start to set up the WPS programs. <code bash> [(it_css:traine)@login00.darwin wrf]$ tar -xvf example_wrf.tar example_wrf/ example_wrf/wrf_job/ example_wrf/wrf_job/Makefile example_wrf/wrf_job/wrfmpi.qs example_wrf/wrf_job/namelist.input example_wrf/data/ example_wrf/data/WPS_GEOG/ example_wrf/data/GFS/ example_wrf/data/GFS/modelDataDownload.sh example_wrf/data/wps_geo_download.sh example_wrf/wps_job/ example_wrf/wps_job/namelist.wps example_wrf/wps_job/Makefile </code> Now you need to download the **Statis Geography Data** and uncompress it by using the provided script ''wps_geo_download.sh'' provided in this example. This is a fairly large set of files (~29GB) and could take some download and uncompress. <code bash> [(it_css:traine)@login00.darwin wrf]$ cd example_wrf/data [(it_css:traine)@login00.darwin data]$ ./wps_geo_download.sh ...MANY LINES AND MINUTES LATER... </code> If this is successful, then it is suggested that you remove the ''geog_high_res_mandatory.tar'' file. <code bash> [(it_css:traine)@login00.darwin data]$ rm geog_high_res_mandatory.tar </code> Now we will download some GFS model data. You will need to update the ''modelDataDownload.sh'' script provided in the example to the appropriate dates you want for the data available from NOAA's [[ftp://ftpprd.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/| FTP site]]. The directory should be formatted something like ''gfs.YYYYMMDD''. In this example, we will be using ''gfs.20210107''. <code bash> [(it_css:traine)@login00.darwin data]$ cd GFS </code> In the ''modelDataDownload.sh'' file you will want to update the ''yyyy'', ''mm'', ''dd'', ''hh'', ''fff'', and ''model'' to match the date and model data that you are trying to download. This example used ''01/07/2021'' for the ''00Z'' run for the GFS model. You also need to put in your email address from <note important>This data is not longer available on NOAA's FTP website, so the example below will not work if you copy it directly and do not change the dates. </note> After making the necessary changes, you can run ''modelDataDownload.sh'' script. To do this you will want to use ''sbatch'' and the ''download.qs'' script. The ''download.qs'' will run the ''modelDataDownload.sh'' script on a compute node to download all the requested model data. <code bash> [(it_css:traine)@login00.darwin GFS]$ sbatch download.qs [(it_css:traine)@login00.darwin GFS]$ ls gfs_20210107_00z_000f gfs_20210107_00z_078f gfs_20210107_00z_156f gfs_20210107_00z_234f gfs_20210107_00z_312f gfs_20210107_00z_003f gfs_20210107_00z_081f gfs_20210107_00z_159f gfs_20210107_00z_237f gfs_20210107_00z_315f gfs_20210107_00z_006f gfs_20210107_00z_084f gfs_20210107_00z_162f gfs_20210107_00z_240f gfs_20210107_00z_318f gfs_20210107_00z_009f gfs_20210107_00z_087f gfs_20210107_00z_165f gfs_20210107_00z_243f gfs_20210107_00z_321f gfs_20210107_00z_012f gfs_20210107_00z_090f gfs_20210107_00z_168f gfs_20210107_00z_246f gfs_20210107_00z_324f ... ... ... ... ... gfs_20210107_00z_072f gfs_20210107_00z_150f gfs_20210107_00z_228f gfs_20210107_00z_306f gfs_20210107_00z_384f gfs_20210107_00z_075f gfs_20210107_00z_153f gfs_20210107_00z_231f gfs_20210107_00z_309f </code> To prevent an issue later down the road, we will need to move the ''modelDataDownload.sh'' script up one directory into the ''data'' directory. <code bash> [(it_css:traine)@login00.darwin GFS]$ mv modelDataDownload.sh download.qs ../ </code> Now we have the required metadata files and sample GFS data downloaded to run WPS and WRF. ==== Example WPS run ==== To avoid running executable on the login node, we are going to request a interactive compute node session using ''salloc''. <code bash> [(it_css:traine)@login00.darwin GFS] salloc --partition=it_css salloc: Granted job allocation 11715984 salloc: Waiting for resource configuration salloc: Nodes r00n49 are ready for job [triane@r00n49 GFS]$ </code> Now we are ready to process the downloaded GFS data with the WPS scripts for our example. To do this we will need to change directories into the the ''wps_job'' directory, then use of the ''make'' command to gather the require files that you will need to run WPS. <code bash> [(it_css:traine)@r1n00 data]$ cd ../../wps_job/ [(it_css:traine)@r1n00 wps_job]$ make cp /work/it_css/sw/WRF/20210113/src/WPS/geogrid/GEOGRID.TBL GEOGRID.TBL cp /work/it_css/sw/WRF/20210113/src/WPS/metgrid/METGRID.TBL METGRID.TBL cp /work/it_css/sw/WRF/20210113/src/WPS/ungrib/Variable_Tables/Vtable.GFS Vtable cp /work/it_css/sw/WRF/20210113/src/WPS/link_grib.csh link_grib.csh </code> After that is completed you will need to update the ''namelist.wps'' file to reflect the model data that you downloaded in the prior steps. Use ''nano'' or ''vim'' and make the necessary changes noted with ''#EDIT THIS LINE'' and ''#ADD THIS LINE''. <code bash> &share wrf_core = 'ARW', max_dom = 1, start_date = '2021-01-07_00:00:00','2006-08-16_12:00:00', #EDIT THIS LINE end_date = '2021-01-23_23:00:00','2006-08-16_12:00:00', #EDIT THIS LINE interval_seconds = 10800 #edit this line io_form_geogrid = 2, / &geogrid parent_id = 1, 1, parent_grid_ratio = 1, 3, i_parent_start = 1, 31, j_parent_start = 1, 17, e_we = 10, 112, #EDIT THIS LINE e_sn = 10, 97, #EDIT THIS LINE ... geog_data_res = 'default','default', dx = 10000, #edit this line dy = 10000, #edit this line map_proj = 'lambert', ref_lat = 38.72, #edit this line ref_lon = -75.08, #edit this line truelat1 = 38.72, #edit this line truelat2 = 38.72, #edit this line stand_lon = -75.08, #edit this line geog_data_path = '../data/WPS_GEOG/', #EDIT THIS LINE opt_geogrid_tbl_path = './' #OPTIONAL EDIT THIS LINE ... &metgrid fg_name = 'FILE' io_form_metgrid = 2, opt_metgrid_tbl_path = './' #ADD THIS LINE / </code> Now you are ready to run ''geogrid.exe'' <code bash> [triane@r00n49 wps_job]$ geogrid.exe Parsed 28 entries in GEOGRID.TBL Processing domain 1 of 1 Processing XLAT and XLONG Processing MAPFAC Processing F and E Processing ROTANG Processing LANDUSEF Calculating landmask from LANDUSEF ( WATER = 17 21 ) Processing HGT_M Processing SOILTEMP Processing SOILCTOP Processing SCT_DOM Processing SOILCBOT Processing SCB_DOM Processing ALBEDO12M Processing GREENFRAC Processing LAI12M Processing SNOALB Processing CON Processing VAR Processing OA1 Processing OA2 Processing OA3 Processing OA4 Processing OL1 Processing OL2 Processing OL3 Processing OL4 Processing VAR_SSO Optional fields not processed by geogrid: LAKE_DEPTH (priority=1, resolution='default', path='../data/WPS_GEOG/lake_depth/') URB_PARAM (priority=1, resolution='default', path='../data/WPS_GEOG/NUDAPT44_1km/') FRC_URB2D (priority=1, resolution='default', path='../data/WPS_GEOG/urbfrac_nlcd2011/') IMPERV (priority=1, resolution='default', path='../data/WPS_GEOG/nlcd2011_imp_ll_9s/') CANFRA (priority=1, resolution='default', path='../data/WPS_GEOG/nlcd2011_can_ll_9s/') EROD (priority=1, resolution='default', path='../data/WPS_GEOG/erod/') CLAYFRAC (priority=1, resolution='default', path='../data/WPS_GEOG/clayfrac_5m/') SANDFRAC (priority=1, resolution='default', path='../data/WPS_GEOG/sandfrac_5m/') !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ! Successful completion of geogrid. ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! </code> Next run the ''link_grib.csh'' script and make sure to pass the correct path to where you are storing the downloaded GFS data. <code> [(it_css:traine)@r1n00 wps_job]$ ./link_grib.csh ../data/GFS/ </code> This script will create soft links to the GFS data files into the current directory. You should see something like this <code bash> [(it_css:traine)@r1n00 wps_job]$ ls -l GRIB* lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAA -> ../data/GFS/gfs_20210107_00z_000f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAB -> ../data/GFS/gfs_20210107_00z_003f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAC -> ../data/GFS/gfs_20210107_00z_006f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAD -> ../data/GFS/gfs_20210107_00z_009f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAE -> ../data/GFS/gfs_20210107_00z_012f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAF -> ../data/GFS/gfs_20210107_00z_015f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAG -> ../data/GFS/gfs_20210107_00z_018f lrwxrwxrwx 1 traine it_css 33 Jan 26 13:59 GRIBFILE.AAH -> ../data/GFS/gfs_20210107_00z_021f ... </code> Next run ''ungrib.exe'' <code bash> [triane@r00n49 wps_job]$ ungrib.exe </code> Depending on the amount of data, this could take several minutes. If successful, then you should get a message like this at the end. <code bash> ... #MANY LINES & MINUTES LATER ********** Done deleting temporary files. ********** !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ! Successful completion of ungrib. ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! </code> The last step is to run ''metgrid.exe'' <code bash> [(it_css:traine)@r1n00 wps_job]$ metgrid.exe ... #MANY LINES & MINUTES LATER Processing 2021-01-13_21 FILE !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ! Successful completion of metgrid. ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! </code> Now you can end the salloc session and release the compute node <code bash> [(it_css:traine)@r1n00 wps_job]$ exit exit [(it_css:traine)@login00.darwin wps_job]$ </code> Now you are ready to set up WRF. ==== Set up WRF ==== If you have not followed all the prior steps leading up to this point **YOU MUST go back and complete them**. The WPS steps to download and process files in them are required for the WRF steps to work. After you have completed the prior steps, then you can go into the ''wrf_job'' directory. <code bash> [(it_css:traine)@login00.darwin wps_job]$ ../wrf_job [(it_css:traine)@login00.darwin wrf_job]$ pwd /work/it_css/sw/WRF/example_wrf/wrf_job </code> Now run the ''make'' command. This will copy all the required files needed to run WRF. <code bash> [(it_css:traine)@login00.darwin wrf_job]$ make cp /work/it_css/sw/WRF/20210113/src/WRF/run/GENPARM.TBL GENPARM.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/HLC.TBL HLC.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/LANDUSE.TBL LANDUSE.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/MPTABLE.TBL MPTABLE.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/SOILPARM.TBL SOILPARM.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/URBPARM.TBL URBPARM.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/URBPARM_UZE.TBL URBPARM_UZE.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/VEGPARM.TBL VEGPARM.TBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/ozone.formatted ozone.formatted cp /work/it_css/sw/WRF/20210113/src/WRF/run/ozone_lat.formatted ozone_lat.formatted cp /work/it_css/sw/WRF/20210113/src/WRF/run/ozone_plev.formatted ozone_plev.formatted cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTM_DATA RRTM_DATA cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTM_DATA_DBL RRTM_DATA_DBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTMG_LW_DATA RRTMG_LW_DATA cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTMG_LW_DATA_DBL RRTMG_LW_DATA_DBL cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTMG_SW_DATA RRTMG_SW_DATA cp /work/it_css/sw/WRF/20210113/src/WRF/run/RRTMG_SW_DATA_DBL RRTMG_SW_DATA_DBL </code> Next copy the ''met_em'' files that were created in the WPS steps. <code bash> [(it_css:traine)@login00.darwin wrf_job]$ cp ../wps_job/met_em* . </code> Now you should be ready to run WRF. ==== Run WRF example ==== In this example, we are running the WRF ''real.exe'' and ''wrf.exe'' in ''WRF/run''. These steps should be similar to the steps required to run WRF in ''/WRF/test/em_real'', but there might be some differences that are not covered in these directions. First, changes need to be made to the ''namelist.input'' file by using ''nano'' or ''vim''. These changes should reflect the information from the model data that you downloaded earlier. See all lines marked with ''#EDIT THIS LINE'' and change accordingly. <code bash> &time_control run_days = 0, run_hours = 12, run_minutes = 0, run_seconds = 0, start_year = 2021, 2000, 2000, #EDIT THIS LINE start_month = 01, 01, 01, #EDIT THIS LINE start_day = 07, 24, 24, #EDIT THIS LINE start_hour = 00, 12, 12, #EDIT THIS LINE end_year = 2021, 2000, 2000, #EDIT THIS LINE end_month = 01, 01, 01, #EDIT THIS LINE end_day = 13, 25, 25, #EDIT THIS LINE end_hour = 21, 12, 12, #EDIT THIS LINE interval_seconds = 10800 #EDIT THIS LINE input_from_file = .true.,.true.,.true., history_interval = 180, 60, 60, frames_per_outfile = 1000, 1000, 1000, restart = .false., restart_interval = 7200, io_form_history = 2 io_form_restart = 2 io_form_input = 2 io_form_boundary = 2 / &domains time_step = 180, time_step_fract_num = 0, time_step_fract_den = 1, max_dom = 1, use_adaptive_time_step = .true., e_we = 10, 112, 94, #EDIT THIS LINE e_sn = 10, 97, 91, #EDIT THIS LINE e_vert = 33, 33, 33, p_top_requested = 5000, num_metgrid_levels = 34, #EDIT THIS LINE num_metgrid_soil_levels = 4, dx = 10000, 10000, 3333.33, #EDIT THIS LINE dy = 10000, 10000, 3333.33, #EDIT THIS LINE grid_id = 1, 2, 3, parent_id = 0, 1, 2, i_parent_start = 1, 31, 30, j_parent_start = 1, 17, 30, parent_grid_ratio = 1, 3, 3, parent_time_step_ratio = 1, 3, 3, feedback = 1, smooth_option = 0 / </code> To run WRF, we will use the ''wrfmpi.qs'' file. This job script is based on the script template found in ''/opt/shared/templates/slurm/generic/mpi/openmpi/openmpi.qs''. The ''openmpi.qs'' job script file is updated regularly, so it is suggested that you also check and compare the contents of the ''openmpi.qs'' files to see if there has been any major changes to it. If there has been, then you might want to consider adding those changes to the ''wrfmpi.qs'' job script. To run WRF, you will use ''sbatch'' to submit the job to Slurm. <code bash> [(it_css:traine)@login01 wrf_job]$ sbatch wrfmpi.qs </code> After a couples minutes this script should finish running. Once the job has completed you can check the ''slurm-<<//JOB_ID//>>.out'' file to check if the run was successful. If it was a successful run, then the last couple lines of the file should say something similar to the lines shown below <code bash> Timing for main (dt= 80.00): time 2021-01-07_11:58:40 on domain 1: 0.00617 elapsed seconds Timing for main (dt= 80.00): time 2021-01-07_12:00:00 on domain 1: 0.00617 elapsed seconds Timing for Writing wrfout_d01_2021-01-07_12:00:00 for domain 1: 0.01661 elapsed seconds d01 2021-01-07_12:00:00 wrf: SUCCESS COMPLETE WRF Finished Running mpi_wrf.exe </code> ===== References ===== Here is a list of references that might help you with errors that might be encountered while working with WRF. * [[https://www2.mmm.ucar.edu/wrf/OnLineTutorial/compilation_tutorial.php | How to compile WRF]] * [[https://www2.mmm.ucar.edu/wrf/users/downloads.html | UCAR's WRF Downloads Page ]] * [[https://www2.mmm.ucar.edu/wrf/users/pub-doc.html | WRF Users Page ]] * [[https://www2.mmm.ucar.edu/wrf/users/download/get_sources_wps_geog.html | WPS V4 Geographical Static Data Downloads Page]] * [[https://cpb-us-w2.wpmucdn.com/sites.uwm.edu/dist/8/663/files/2019/05/wrfv4guide.pdf | Using the WRF-ARW on the UWM mortimer HPC Guide for WRF-ARW Version 4.0.3 ]] * [[https://github.com/wrf-model/WRF | WRF GitHub Page ]] software/wrf/darwin.txt Last modified: 2024-04-02 14:39by anita