Parflow
- Simon Alexander
Overview
Homepage: ParFlow hydrologic model: Modelling surface and subsurface flow on high-performance computers
Using
ParFlow is currently not available as a module, but has been installed individually within a project (due to extending the code base).
If you would like the standard ParFlow installed as a module please request via our Service Portal:
Multicore
Parflow is build upon OpenMPI and thus can run across multiple nodes.
Local Install Process:
ParFlow has a number of dependencies that are typically provided by pre-installed modules (such as hypre). We have not been able to use our module version of silo, but instead have had to install this locally, as demonstrated in the process below.
This example also uses specific versions/releases of ParFlow and silo. ParFlow releases can be found at the link at the start of the this page, silo releases can be found here: Silo Releases
You will need to replace <project-name> with your own project folder name, or update the folder paths with you own locations..
Testing:
After the make install step, you can run make test to verify the installation.
Due to the nature of these tests, you MUST NOT run these on the login nodes as they will utilize the majority of cores and over nine hours to complete. This will affect ALL users on that login node.
Instead, submit a job onto the cluster using a Slurm script of the form below that calls make test from within the parflow build folder:
Our testing has observed:
We assume ParFlow’s test suite must have a job timeout limit of 3600s built within it. But due to the percentage of successful tests, we’re satisfied that this installation works.