This repository accompanies the McMillan et al. (2023) paper in Geophysical Research Letters (https://doi.org/10.1029/2023GL103996) and mainly consists of deal.II paramter files for the ASPECT geodynamic modelling software. The source code for all of the models discussed in the journal article are contained in this repository.
Model variants are contained in the models directory and are broken into two phases.
While v0.prm is a complete parameter file, the others are written to be merged together with v0.prm in order to produce complete models.
The base model (both phases) can be run as follows:
mpirun -np 10 /home/dealii/aspect/aspect-release models/v0.prm
wait
mpirun -np 10 /home/dealii/aspect/aspect-release models/v0-phase-2.prmThe run-tag script is provided for running variant models.
Basic usage is: ./run-tag <model variant tag> <path to ASPECT> <low-res flag?>
As an example, model v2.1 can be run as follows:
./run-tag v2.1 /home/dealii/aspect/aspect-release
This will run both phases of the model. The debug build of ASPECT can be used by passing it as the path.
Note that the run-tag script was originally written to be used by the SLURM scheduler, which does not support the -np argument to mpirun. Therefore, the scripts will only run on 1 CPU process by default. To utilize more processes, edit run-tag manually by changing both instances of mpirun ... to e.g., mpirun -np 20 ... to use 20 cores.
For running on personal machines with few cores, a low-res.prm parameter file is provided, which reduces the spatial resolution of the model. This can be utilized by passing the -l flag, as follows:
./run-tag v2.1 /home/dealii/aspect/aspect-release -l
This will automatically merge the low-res.prm file in with whatever model variant is passed as a tag. Conversely, a high resolution version is provided with high-res.prm through the -h tag, which doubles the spatial resolution.
The run-all script simply loops through all the model variant tags v0...v10.3 and passes them to run-tag.
Note that it is necessary to modify the file permissions for each script to enable them to be executed, e.g., chmod +x ./run-tag. If using Docker, this should be done on the host machine (or inside of WSL in Windows) before building the Docker container.
The file generate-particcle-data.sh generates a text file with particle coordinates located throughout the crust, but particles were not used in the published models.
The Dockerfile that was originally used to build the models is provided. This file is based on a specific version of deal.II and builds ASPECT from the tag 08b6a15, so the models are fully reproducible.
For VS Code users, a devcontainer.json is provided.
All output files are stored in the model's corresponding output-*\ subdirectories, e.g., output-v2.1\ or output-low-res-v2.1\
The ParaView state file used to create the visualizations for the paper is provided (drip.pvsm). To use it, launch ParaView and choose File > Load State... and navigate to the drip.pvsm file, select it, and click OK. Unfortunately, ParaView only supports absolute paths to locate the solution.pvd file. To work around this, select Choose file names from the drop-down box, and browse to the output folder you wish to visualize, e.g., output-v2.1, and select the solution.pvd file within it. After selecting OK, the solution file and several data filters should be displayed in the pipeline browser.
The Slurm script used to run the models on SciNet/Niagara cluster is provided as slurm-run-tag. This script takes the model variant tags from the Slurm task array, allowing the user to choose which model variants to run on the cluster. For example, to run models v2.1, v2.2, and v2.3, one can pass --array=21,22,23 to Slurm. The script defaults to v0 if no array jobs are specified.
These files were copied and cleaned up from a working repository. I may have missed something or introduced minor typos. Please open an issue in this repository in the event of bugs or issues running the code.