Training session to set up MESH for the Bow River at Banff catchment

To download this repository on the $HOME directory of your Graham account:

foo@gra-login1:~$ git clone https://github.com/kasra-keshavarz/community-modelling-workflow-training.git ./github-repos/community-workflows

Library requirements

General

Certain libraries and binary executables are necessary to run the workflows in this repository. Below necessary libraries for general usage are mentioned:

1. CDO (Climate Data Operators >=v2.2.1),
2. ecCodes (>=v2.25.0),
3. Expat XML parser (>=v2.4.1),
4. GDAL (>=3.5.1),
5. GEOS (>=3.10.2),
6. HDF5 (>=1.10.6),
7. JasPer (>=2.0.16),
8. libaec (>=1.0.6),
9. libfabric (>=1.10.1),
10. libffi (>=3.3),
11. libgeotiff (>=1.7.1),
12. librttopo (>=1.1.0),
13. libspatialindex (>=1.8.5),
14. libspatilite (>=5.0.1),
15. netcdf-fortran (>=4.5.2),
16. netcdf (>=4.7.4),
17. postgresql (>=12.4),
18. proj (>=9.0.1),
19. python (>=3.10.2),
20. sqlite (>=3.38.5),
21. udunits (>=2.2.28)

Each of the above libraries and binaries may need further dependencies. It is up to the user to assure all requirements are satisfied. Most GNU/Linux distributions should be able to offer all the libraries above through their remote package repositories. If not, it is recommended to compile and store them for future reference.

Digital Research Alliance of Canada (DRA) Graham HPC

Fortunately, all the above requirements are available on the DRA's Graham HPC. You may load the modules with the following command:

foo@bar:~$ module load StdEnv/2020
foo@bar:~$ module load gcc/9.3.0
foo@bar:~$ module load \
  sqlite/3.38.5 postgresql/12.4 gdal/3.5.1 \
  udunits/2.2.28 cdo/2.2.1 gentoo/2020 \
  imkl/2020.1.217 openmpi/4.0.3 libfabric/1.10.1 \
  jasper/2.0.16 freexl/1.0.5 geos/3.10.2 \
  libaec/1.0.6 mpi4py/3.1.3 \
  libffi/3.3 hdf5/1.10.6 \
  libgeotiff-proj901/1.7.1 librttopo-proj9/1.1.0 \
  proj/9.0.1 eccodes/2.25.0 netcdf-fortran/4.5.2 \
  mii/1.1.2 ucx/1.8.0 python/3.10.2 \
  netcdf/4.7.4 cfitsio/4.1.0 \
  libspatialite-proj901/5.0.1 expat/2.4.1 \
  yaxt/0.9.0 libspatialindex/1.8.5 arrow/13.0.0 \
  scipy-stack/2023b ipykernel/2023b;

Note

Both scipy-stack/2023b and ipykernel/2023b need to be loaded at the end to assure the sys.path addresses in Python sessions are ordered as expected.

It is recommended to save all load modules as a list to be able to restore them whenever needed. Using the LMOD features, you may save them with:

foo@bar:~$ module save scimods # you can change "scimods" to anything!

And, you may restore the list with:

foo@bar:~$ module restore scimods

Note

Please note that some of the libraries and binary programs are necessary for the Python environment to run smoothly (see below).

Python requirements

General

The following list of Python packages are required to run much of the workflows in this repository. The requirements.txt file describes the packages necessary to run the workflows.

Please refer to DRA's manual for necessary information on how to create a Python virtual environment using the requirements.txt file mentioned above.

The installation process needs to be done in the login node of the Graham cluster, so let's switch to a login node:

foo@bar:~$ ssh gra-login1 # user your username and password

Once you login, your sheel will look like the following:

foo@gra-login1:~$ 

Whenever you change a node, make sure you load all the necessary modules:

foo@gra-login1:~$ module restore scimods

Then, you may create Python virtual environments (after assuring all the modules are loaded) on Graham HPC, to isolate the environment to execute the workflows. on Graham, it is recommended to use your $HOME directory, so a path like the following is recommended:

foo@gra-login1:~$ python -m virtualenv $HOME/virtual-envs/scienv

After creating the virtual environment, you can activate the environment with:

foo@gra-login1:~$ source $HOME/virtual-envs/scienv/bin/activate
(scienv) foo@gra-login1:~$ # this is how your Graham sheel will look

After the activation of the virtual environment, you may install any Python package within the environment. To install those we need for the modelling workflows:

(scienv) foo@gra-login1:~$ pip install -r ~/github-repos/community-workflows/0-prerequisites/requirements.txt

Once the scienv is ready, you may add the virtual environment to the Jupyter Lab as a kernel using the following command:

(scienv) foo@gra-login1:~$ python -m ipykernel install --name "scienv" --user

Important

If you face any errors by executing the command above, make sure jupyter and ipykernel packages are installed properly. Similarly, you may again use pip to install these packages.

Once added as a kernel, you should your virtual environment within your Jupyter sessions.

Additional datasets necessary

1. MERIT-Basins vector hydrography Dataset (v0.7/v1.0, minor bug fix for coastaline pixels): https://www.reachhydro.org/home/params/merit-basins

   MERIT-Basins is available on Graham HPC under the following directory:

   /project/rrg-mclark/data/geospatial-data/MERIT-Basins # rpp-kshook (GWFO) allocation
   /project/rpp-kshook/Climate_Forcing_Data/geospatial-data/MERIT-Basins # rrg-mclark allocation

2. Datatool (version v0.5.1-dev): https://github.com/kasra-keshavarz/datatool

   Download with:

   foo@gra-login1:~$ git clone https://github.com/kasra-keshavarz/datatool.git ./github-repos/datatool

3. GIStool (version v0.1.7-dev, commit ff2a6da): https://github.com/kasra-keshavarz/gistool

   Download with:

   foo@gra-login1:~$ git clone https://github.com/kasra-keshavarz/gistool.git ./github-repos/gistool

4. EASYMORE (v2.0.0-dev): https://github.com/ShervanGharari/EASYMORE

   Download with:

   foo@gra-login1:~$ pip install git+https://github.com/ShervanGharari/EASYMORE.git ./github-repos/easymore

Last edited: March 30th, 2024