Docker image containing the parallel version of MODFLOW 6 with application examples

Modflow on parallel processing seemed to lie in the far-off future, but now it’s possible to take advantage of most of your cores to run parallelized versions of MODFLOW 6. Since most MODFLOW modelers work on Windows we wanted to bring something more than a “packed” solution; our goal was to bring an enriched “packed” experience.

Because parallel is more than the solver, there are other dependencies such as OpenMPI and Petsc that need to be installed as well as some C++ and Fortran compilers. 

We can’t change the fact that most modelers won’t use the parallel version of MODFLOW-6, for the same reasons that they still use MODFLOW-2005, but at least we can provide a framework where they can run successfully with the minimum steps and reduce the occurrence of failures and example of the solver over an applied model and compare the simulation with the serial MODFLOW-6.   

This is an applied case where MODFLOW-6 serial and parallel version have been compiled and set as executables on a Docker image that runs Jupyter Lab. The image also has an example model and the tutorial shows the procedure to run the parallel MODFLOW-6 under OpenMPI with the selection of the core number.

For this tutorial you need to have Docker Desktop installed on your computer:

docs.docker.com/desktop/install/windows-install/

Tutorial

Commads

To run the Docker image on Powershell:

docker run -p 8888:8888 -e JUPYTER_TOKEN=hatarilabs hatarilabs/modflow_parallel:0.0.1

To tun the parallel version of MODFLOW-6:

mpiexec -np 2 mf6_parallel mfsim.nam