Page Comparison

Software Requirements

• Fortran 90/95 and C compiler

• MPI or OpenMP libraries

• netCDF4 library linked with HDF5 and zip libraries and extended by the Fortran netCDF package (The netCDF4 package comes with the programs ncdump and nccopy)

• UNIX utilities: make, ksh, uname, sed, awk, wget, etc.

• For post-processing: Climate Data Operators (CDO) and netCDF Operators (NCO)

...

The starter package is written for the HRLE–3 Mistral HRLE-4 Levante at DKRZ. Additional changes need to be applied for unsage on other machines. On the Mistral At DKRZ the Simple Linux Utility for Resource Management SLURM ist is installed.
On mistral please use OPENMPI by setting

module load intel-oneapi-compilers/172022.0.1-gcc-11.2.0
module load openmpi/4.1.2-intel-2021.0.2p1_hpcx-intel14

export OMPI_MCA_pml=cm
export OMPI_MCA_mtl=mxm
export MCM_RDMA_PORTS=mlx5_0:15.0

Unpack and Configure

1. Copy the starter package from RedC🔑 Downloads.
2. Unpack the starter package:

$ tar -xzvf cclm-sp-45.0.tgz
$ mv cclm-sp-45.0 yourpath/cclm-sp

yourpath/cclm-sp is named SPDIR from here on.

3. Copy and unpack the supplementary data files for testing the starter package (the program wget needs to be installed on your system)

$ cd yourpath$SPDIR/cclm-sp/data
$ ./get_sp_ext.sh

4. Change to the directory yourpath$SPDIR/cclm-sp/configure_scripts
5. Adjust the settings in the file system_settings to your computer system
6. Type the following command to create a default test experiment

...

The source code for INT2LM, CCLM, and the auxiliary program packages CFU and fortran-csv-lib need to be compiled for running a simulation with the regional climate model. INT2LM and CCLM needs need to be compiled by yourself. CFU and fortran-csv-lib are automatically compiled when you run the config.sh script.

...

$ cd /$SPDIR/src/int2lm

Open the Fopts script with any text editor and change the fortran Fortran options according to your computer platform. Compile INT2LM:

$ make

In case of problems, try module unload python3 before the make-command (conda might have installed Fortran and destroyed the Intel-Fortran). If compilation is successful an executable $SPDIR/src/int2lm/bin/int2lm.exe is created.

...

$ cd /$SPDIR/src/cclm

Open the Fopts script with any text editor and change the fortran Fortran options according to your computer platform. The Makefile might be edited for setting the right MACHINE (e.g., levante.atos.local). Compile CCLM:

$ make

...

this results in a much faster compilation. Replace N by the number of requested processors.

Run the test examples

There are two tests, one for testing CCLM with GCM or reanalysis data as initial and boundary conditions (${SPDIR}/chain/gcm2cclm/sp001) and one for testing CCLM with coarse grid CCLM data as initial and boundary conditions (${SPDIR}/chain/cclm2cclm/sp002). Actually, sp001 creates the necessary input data for sp002.

Before you start the experiment look for the following environment variables in the job_settings of settings of sp001 and sp002 file and adopt adapt them to your needs.

PROJECT_ACCOUNT=  # your project account
EMAIL_ADDRESS=    # your email address if you want to get information when your job crashes or finishes

Now you should be ready to start the first experiment:

...

This experiment is a two month simulation, 50 km / Europe / driven bei by ERAInterim

The scripts are called in the order prep.job.sh, int2lm.job.sh, cclm.job.sh, arch.job.sh, post.job.sh.  If your job crashes in one of the scripts you do not have to run all the successful scripts again, but can start this script again after you made the corrections by submitting the appropriate command from the following list:

$ ./subchain prep
$ ./subchain conv2icon
$ ./subchain iconcclm noprep
$ ./subchain arch
$ ./subchain post

After successful completion of the  SP001 experiment adopt the sp001 experiment adapt the scripts in  SP002 sp002 and start this experiment:

$ cd ${SPDIR}/chain/icon2iconcclm2cclm/sp002
$ ./subchain start

This experiment is a two-month simulation, 2 km resolution, around Hamburg, driven by the results from the sp001 experiment.